5 minutes.
Smarter marketing.
Every weekday.

Every few months a new model launch promises to change everything and mostly moves a benchmark score nobody outside the lab reads. This launch reads differently, mostly because of who already switched and why.

Anthropic released Claude Fable 5.1 on September 1, available immediately across Claude.ai, Claude Code, Claude Cowork, the API, and every major cloud platform. It handles long, multi-step projects with far less hand-holding, hands back a finished deliverable instead of needing constant check-ins, reads diagrams, charts, and tables buried in PDFs, and writes with noticeably better clarity according to Anthropic's early partners.

The clearest theme in Anthropic's own customer quotes is how much less oversight the model needs on long jobs. An engineer at MongoDB said Fable 5.1 built a complex prototype over three days, running for hours unattended with strong verification loops. A machine learning engineer at Ramp described a 38-hour unattended run that diagnosed a data labeling bug, corrected it, and came back with results and next steps. Hebbia said Fable 5.1 produced the best PowerPoint decks of any model they've tested, and Canva said the writing itself was more understandable and better followed their style guidance than any earlier Claude model.

One example stands out for what it says about how the model works, not just how well. An engineer at the investment firm Millennium described a crash in their internal systems that happened about once in a million runs and had stumped their own engineers, plus every AI model they'd tried, for four to five years. Fable 5.1 was the first to find it, disassembling an external vendor library and matching it against the crash data to trace the bug.

Cache-read pricing dropped 75%, which Anthropic says works out to about 25% cheaper for typical workloads and up to 45% cheaper for heavy, agentic work, at the same per-token price as before. Multiple companies said they're moving work they used to reserve for the pricier Opus model onto Fable 5.1 instead. If you already use Claude for anything, this model is already live, nothing to buy or upgrade. Try handing it something you'd normally break into five small steps and see if it can run the whole thing in one pass.

Source: Anthropic

For a decade, "Hey Google, find a [x] near me" ran through a system built to do one job well: pull a short, structured answer from Maps and Google Business Profile data. That system starts disappearing this week.

Starting September 4, 2026, Google will begin permanently retiring Google Assistant from Android phones, tablets, Wear OS watches, connected earbuds, and Android Auto, replacing it with Gemini. The switch rolls out automatically over several weeks, no setting change required, and once it reaches a device there is no way back to Assistant. Cars with Google built-in, Google TV, and smart speakers are not part of this phase and keep Assistant for now.

The bigger change is what's actually answering. Assistant did narrow, structured lookups against Maps and Business Profile data and handed back a short list. Gemini is a full conversational model that reasons across more sources, holds context through a conversation, and can be more editorial about which businesses or answers it actually surfaces instead of a plain list. That's the same shift already playing out with ChatGPT and Google's own AI Overviews, about to land as the default assistant on the phone in most people's pocket.

Local business owners have already been asking whether they show up in AI Overviews or get cited by ChatGPT. This is that same question, about to apply to voice search on the most common phone operating system in the US. What actually gets a business surfaced hasn't changed as much as the interface is about to: an accurate, complete Google Business Profile, the right categories, real hours, current photos, and reviews with actual detail and recent dates.

Google isn't asking anyone's permission here. It's about to change the default interface on the majority of the world's Android phones over a matter of weeks, with no opt-out once your device gets reached. Pull up your own Google Business Profile this week and make sure it's actually complete, then literally ask Gemini a question a customer might ask about your business or industry and see what it says back.

Source: Search Engine Land

For twenty-plus years, the software industry's whole model has been build the dashboard, then get the customer to live inside it. Salesforce just admitted that model might be ending, and it's the one doing the admitting.

Salesforce in Claude launches with 37 prebuilt skills covering the actual work a salesperson does all day: prep for a meeting, review a deal's health, check pipeline status, all pulled from live Salesforce data instead of a static report someone built last week. Salesforce says more skills are coming later this year, and the onboarding reads your existing Salesforce, Slack, and any other connected tool to build a dashboard tailored to your actual accounts the first time you open it.

This is a bigger shift than one product update. Every major ad platform, X, TikTok, Snapchat, Meta, has already built a version of this for reporting: connect your account, ask a plain-English question, get an answer without touching the dashboard. Salesforce just applied that same idea to the core CRM itself, the software millions of sales teams live inside every single day.

Salesforce has spent decades defending its own interface, building features specifically to keep customers inside its dashboard instead of exporting data elsewhere. CNBC reported the move comes as Salesforce responds to what some are calling a "SaaSpocalypse," the fear that AI agents make traditional software interfaces optional. Rather than wait to get bypassed by outside AI tools reading its data anyway, Salesforce built the exit ramp itself.

Salesforce didn't do this out of generosity. It did it because the safer bet is being the CRM that lives inside the AI assistant your team already trusts, instead of being the CRM everyone quietly stops opening once that assistant can do the job anyway. If you or a client run Salesforce, get on the beta list this week and test it against one real pipeline review before deciding if it's hype. If you don't run Salesforce, ask whatever CRM or ad platform you do use whether something like this is on their roadmap, because it's coming for all of them one way or another.

Source: VentureBeat

This case started in 2023, when California Attorney General Rob Bonta and a bipartisan group of state attorneys general sued Meta, arguing internal research showed the company knew its engagement features were harming teenagers and shipped them anyway. This week, on the eve of trial, Meta settled instead of letting a jury see the rest of that internal research.

The product changes are specific and they're shipping now, not someday. Teens get a two-hour daily cap across Facebook and Instagram combined, overridable by a parent but on by default. Night Mode blocks posting and viewing between midnight and 6am. School Mode mutes notifications from 8am to 3pm. Teens also get prompted at 15 minutes of continuous use, then again at 60 and 90 minutes, and can opt into a chronological feed that skips Meta's recommendation engine entirely.

About $12.2 billion of the settlement goes to states over the next ten years regardless of what anyone else does, earmarked for youth mental health and social media addiction programs. The rest, roughly $5 billion, only unlocks if TikTok, YouTube, and Snapchat each adopt their own one-hour daily cap and night mode and put up matching money. Meta framed that as a challenge to the industry. It's also a smart way to make sure Meta doesn't lose teen attention to platforms that didn't have to change anything.

Either way, expect the rest of the industry to feel this pressure. State attorneys general now have a signed template that says these features work in court. TikTok, YouTube, and Snapchat don't need to take Meta's money to end up facing the same lawsuit with the same demands attached.

Meta didn't do this because it had a change of heart. It did this because a jury was about to see internal documents the company spent years keeping sealed, and a settlement was cheaper than that trial. That doesn't make the product changes fake. Instagram is the biggest teen platform there is, and a real two-hour daily ceiling on that audience is going to show up in engagement numbers whether Meta meant it as a fix or a legal exit. If you or a client run content or ads that lean on a teen or young Gen Z audience on Facebook or Instagram, pull your audience insights this week and check what share of your reach actually comes from under-18 accounts.

Source: Washington Post

Good Good built its name as one of the biggest success stories in YouTube golf, over 2 million subscribers, a roster of personalities with real individual followings, and this fall it became the title sponsor of its own PGA Tour event. Callaway has partnered with the channel since 2023. This week, that partnership produced an ad that never should have made it past a first read of the script.

The video, posted to Callaway's own social channels, showed Good Good co-founder Garrett Clark charging at fellow Good Good personality Alexis Miestowski as she reached for his new Callaway driver. He tackles her to the ground, stands over her, and says, "Do not touch my new driver."

The backlash was immediate and came from across the golf world, not just outside critics. Both companies deleted the ad within the day, though clips had already been saved and kept circulating. The PGA Tour, which now has a direct business relationship with Good Good through the sponsored event, issued its own statement distancing itself: "We do not condone violence in any form, and the since-deleted video does not reflect our values or commitment to inclusivity and respect."

Good Good's CEO said the concept was meant as a parody of a scene from a horror film and that "the execution missed the mark." Callaway said it was "disappointed by the content that was posted" and is "appreciative of Good Good for addressing this." Callaway's CEO later confirmed the company had approved the video and said it has since changed its internal approval process specifically to prevent a repeat.

Strip away the apologies and look at what actually had to happen for this ad to exist. Someone wrote the concept. Someone approved the script. Two actors filmed it. Someone edited it. Someone at Good Good signed off on sending it to a partner brand. Someone at Callaway, a company with real legal and marketing infrastructure, watched the finished video and approved posting it to their official channels. That's not one bad call. That's an entire chain of people who either didn't think it through or didn't feel safe enough to raise their hand and say this is a problem.

If you work with an outside creator, agency, or freelancer who posts under your brand's name, ask yourself honestly whether anyone on that team would push back on a bad idea before it goes live, or whether you're trusting a relationship instead of a process. A partner who won't tell you no is not a partner you can trust with your name.

Source: Yahoo Sports / The Independent

Neil Patel posted something on Facebook this week that's worth taking seriously even if some version of it sounds familiar: TikTok isn't the platform you can keep treating as a side project for younger customers.

The numbers back him up. TikTok now leads every major platform in average daily usage time per user, ahead of YouTube, Instagram, Facebook, WhatsApp, and X. That lead has grown steadily since 2019, when the average TikTok session added up to just 27 minutes a day. It's now more than tripled.

Patel's post pushed back on the idea that TikTok is a Gen Z-only platform, and he's not wrong to push back. He compared it to Snapchat, once dismissed as a high schooler app, whose earliest heavy users are now adults with jobs and buying power. TikTok's earliest users are aging up the same way, and its audience gets a little older every year.

There's a real business case underneath the attention numbers too. TikTok Shop keeps posting bigger commerce numbers each quarter, and the platform is actively building more in-app purchase and payment tools to close the gap between watching and buying.

One honest caveat: not every tracker agrees on the full picture. Per-user daily minutes clearly favor TikTok now, but because YouTube's total global user base is still massive, some total-volume measures still show YouTube ahead in combined worldwide watch time. Both things can be true at once: TikTok wins on attention per person, YouTube still wins on total reach.

Pull the real numbers for your own industry and audience this week instead of going on a gut feeling that might be two years out of date. If your customers skew even a little younger than retirement age, there's a real chance more of them are watching video on TikTok than you think.

Source: Exploding Topics

X just admitted something out loud that most platforms only imply: their own ad dashboard isn't where you actually want to be managing your campaigns anymore.

An MCP server, short for Model Context Protocol, works like a translator. It sits between your X Ads account and whatever AI assistant you're already using, and it lets that assistant read your real campaign data and answer questions about it directly. X said the connector works with Grok, Claude, Claude Code, or "a custom agent built on the MCP SDKs," so this isn't locked to X's own Grok chatbot.

Once it's connected, you can ask in plain language how a campaign performed last week, which audience segment is actually converting, or whether a bid change would help, and get an answer pulled from your live account instead of digging through X's own reporting screens.

X isn't first here. Meta launched its own MCP connector back in April, and TikTok, Pinterest, and Snapchat have all added similar options since. Every major ad platform is now racing to make itself readable by outside AI tools instead of forcing advertisers to live inside their dashboard.

For a small business or solo marketer running X ads, this means you no longer need to become fluent in X's specific ad interface to get a straight answer about your account. You can just ask.

Connect your X ads account to whatever AI tool you already trust and ask it one real question, like which ad set is actually driving results. What comes back will tell you fast whether this is worth building into your regular routine.

Source: Social Media Today

Testing has always been the advice everyone gives and almost nobody actually does, because setting up a proper Google Ads experiment used to require real technical know-how: control groups, split percentages, statistical significance, all of it. Google just removed most of that friction.

The new experiments creation flow lives right in Google Ads Manager. You pick an existing campaign from a list, choose what you want to test, bidding strategy, budget, creative, whatever variable you're curious about, and Google handles the comparison structure behind the scenes. You don't need to know what a proper test setup looks like anymore, Google builds it for you.

Performance Planner got a companion upgrade. It now models how a specific change, say a higher budget target or a different bidding strategy, would likely affect your existing campaign's performance before you commit to it. If you like what you see, one click applies the change directly to the live campaign.

Google also confirmed the updated testing process supports comparison tests involving AI Max elements, meaning you can now measure whether Google's AI-driven ad features are actually helping your specific account, instead of taking Google's word for it.

This matters most for small accounts that never had the bandwidth to run formal experiments. Big agencies with dedicated analysts have been running structured tests for years. This closes that gap for anyone managing their own Google Ads account without a testing background.

Pick one open question you've been guessing at in your Google Ads account, a bidding strategy switch, a budget increase, whether AI Max is actually pulling its weight, and when this rolls out to your account, run it as an actual test instead of a gut call.

Source: Social Media Today

Since July 27, Meta's Advantage+ Creative Text Generation has quietly leveled up. It's not the old, obviously-AI image generation that looks nothing like your brand. This feature takes your actual uploaded image, the one with your real design, real colors, real layout, and generates new versions where only the main headline text changes.

Here's how it works in the ad setup flow. Alongside your primary text and headline options, Meta now shows a third column of Advantage+ Creative suggestions: typically three alternate messaging angles, each paired with at least two new image variants, for up to eight total options. Everything on the image stays identical except the headline, meaning the font style, colors, top banners, icons, and overall composition are preserved exactly.

These options are pre-selected by default. Unless you either turn individual variants off creative by creative, or set guardrails through the Branding section under Identity in Ads Manager (logo, fonts, colors, tone, restricted words), Meta's AI is free to generate and potentially serve these alternate headline versions in your campaigns.

The quality has genuinely improved, which is part of why this matters now. AI-generated ad text used to be bad enough that most advertisers ignored it. That's changing, which means more of these auto-generated variants are actually good enough to get served, whether or not you reviewed them first.

Even with Branding guardrails set, enforcement isn't airtight. Meta still occasionally adds an emoji you wouldn't choose, or slips in a word from your restricted list. Set the guardrails anyway, they reduce the problem even if they don't eliminate it, and review your active ad creative for headline variants you didn't write.

Open Ads Manager, go to the Identity section, and fill out your Branding profile, logo, fonts, colors, tone, and restricted words, even if you never plan to use AI-generated creative on purpose. Then check your currently running ads for headline text you don't recognize. If you find any, that's Meta's AI already making decisions for you.

Source: Jon Loomer Digital

YouTube announced via its Creator Insider account that it's extending its "exposure-based" view-counting model to every format on the platform: long-form videos, podcasts, and live streams, not just Shorts. The change takes effect August 24, 2026.

Here's the actual mechanic. Right now, a view on a long-form video generally requires a few seconds of intentional watch time before it counts. After August 24, a view registers the instant the first frame plays, whether someone watched three seconds or three minutes, whether it autoplayed in a feed or someone clicked with intent to watch. This is the exact same standard Shorts has used since March 2025, just now applied everywhere.

Your historical view counts on old videos won't get retroactively recalculated. But once August 24 hits, any new views on old content start counting under the new rule too, so you'll likely see your numbers tick up faster than usual across your whole channel, not just new uploads.

The part that actually matters for anyone running ads or getting paid: monetization is untouched. Ad revenue and Partner Program eligibility still run on "Engaged views" and "Engaged Watch Hours," a separate, stricter metric that requires real watch time. You can track that specific number inside YouTube Studio Analytics under Advanced Mode, and it's the number worth watching, not the headline view count.

If you or a client report YouTube view counts anywhere, in a proposal, a monthly report, a case study, add a line now explaining that the counting method changes August 24. Get ahead of the "why did our views suddenly spike" question before it lands in your inbox.

Source: Dexerto, via YouTube's official Creator Insider community post

Google first tested Search Profiles back in June, positioning them as a social-style, followable presence for brands and creators sitting directly inside search results, competing for attention against Google's own AI Overviews. This weekend, Google made the feature dramatically more accessible.

The new thresholds: 35,000 subscribers or followers on YouTube, Instagram, or X, or 100,000 followers on TikTok. That's down from 100,000 and 300,000 respectively, a roughly two-thirds cut across the board. If eligible, you set it up directly at Google for Creators, and your profile picks up a header image, a scrolling feed of your recent articles, videos, and social posts, and a follow button that lets people track you without ever visiting your site or following you on the platform where you originally posted.

Google also quietly added something arguably more useful for anyone doing SEO: new reports inside Google Search Console showing the actual search terms people used to find your content, plus ranking insights and search volume data. That's real visibility into how people are finding you through Google, not just referral traffic numbers after the fact.

Separately, YouTube is expanding copyright violation alerts so channel managers, not just channel owners, get notified when a claim or removal request comes in. If you have someone else running your YouTube channel day to day, they'll now see these issues in real time instead of waiting for you to forward an email.

The interesting part isn't the feature itself, it's the direction. Google spent years sending people away from search results as fast as possible. Now it's building follow buttons and dedicated profiles directly into the results page, competing with the very platforms it used to just link out to. That's a meaningful shift in how Google wants people to discover and stick with a brand.

Check your current follower and subscriber counts on YouTube, Instagram, X, and TikTok against these new thresholds. If you're within striking distance on any platform, that's now a legitimate near-term goal instead of a six-figure fantasy, and if you already qualify, go claim your profile at Google for Creators today.

Source: Social Media Today

Anthropic confirmed this week that Claude will start embedding an invisible watermark directly into the text it generates. This isn't a stamp added after the fact. It's woven into how the model chooses its words as it writes, which means the watermark is part of the text itself, not metadata that can be stripped off.

Because the mark lives inside the words, it travels wherever the text goes, including when you copy and paste it into an email, a blog post, or a client deliverable. Anthropic says it "may persist through some editing," though the company hasn't said exactly how much editing breaks it. For files like images, Claude attaches a different kind of mark: signed provenance metadata using the C2PA standard, the same open standard being adopted industry-wide.

The driver here is regulation, not user complaints. The EU AI Act's Transparency Code took effect August 2 and requires AI companies to mark generated content in a way other systems can detect. Anthropic signed on, and the commitment isn't limited to Europe. It applies everywhere Claude is used, including the API, Claude.ai, Claude Code, Claude Cowork, Claude Tag, and Claude accessed through AWS, Google Cloud, or Microsoft Foundry. Only models released on or after August 2 have it at launch, with older models getting it retrofitted later.

There are real limits worth knowing. A detected watermark doesn't prove Claude originated the ideas, since plenty of people use Claude to proofread, translate, or summarize someone else's writing. And the absence of a detected mark doesn't mean something wasn't AI-written either, since heavy editing, paraphrasing, translation, or short passages can all wipe out the signal. Public detection tools aren't live yet, Anthropic calls them "forthcoming."

This probably isn't really about catching any one person. It reads more like Anthropic checking a regulatory box it's legally required to check. But the reaction has been fast and telling, with follow-up coverage already surfacing Claude users worried this could get them caught using AI at work or in school. That tells you the real story isn't the watermark itself, it's that a lot of people have been operating on an assumption of invisibility that was never actually guaranteed.

Decide now, before it's decided for you, how transparent you want to be about using AI in your content. If you use Claude or any AI tool to draft client work, social posts, or emails, figure out your own honest answer to "am I comfortable if someone finds out," because the era of assuming nobody could tell is quietly ending.

Source: Claude Help Center

Adobe interviewed 850 small business owners and solopreneurs in the U.S. for a new report tackling a question a lot of us have been dancing around: how does a business stand out when every competitor has access to the same AI tools, the same suggestions, and the same shortcuts?

The headline finding is almost uncomfortable in how obvious it sounds once you read it: "when everyone has access to the same digital tools, the edge goes to businesses that apply their taste and point of view to their branding and marketing." Nearly four-fifths, 79%, of the small business owners surveyed agreed that having a clear point of view isn't just good branding, it's literally what makes people follow them. And 56% agreed that when everyone has the same tools, the competitive edge goes to whoever has the best taste.

Here's the part that should give every business owner pause. About 1 in 3 respondents said they'd rather post nothing at all than publish something that doesn't sound like them. That's a real instinct, and it's a good one. But the same report found that 63% of these owners second-guess their own design and content decisions, and almost half said they heavily edit AI-assisted content before it goes out. Translation: business owners already know when something sounds off, they just don't always trust that gut check enough to act on it.

Adobe's recommendation is refreshingly unglamorous: write down your actual purpose statement, the specific reason your product or service matters to your customers, and use it as the filter for everything you publish. Not a mission statement for the wall. A one or two sentence gut-check you can hold any AI-assisted draft up against before it goes live.

Write one sentence explaining why your business specifically, not businesses like yours in general, matters to the people who use it. Then go back through the last five pieces of content you published and ask honestly whether each one sounds like that sentence or like it could belong to anyone.

Source: Adobe

On August 11, Manus, the AI agent tool that builds websites, runs research, and completes multi-step tasks on command, announced it's unwinding its acquisition by Meta and returning to independent operations. To comply with regulatory requirements in specific jurisdictions, data generated by certain users since the December 29, 2025 acquisition date will be permanently deleted starting August 23.

Here's where it gets serious. Manus splits affected users into tiers. If you created your account or changed your account email on or after acquisition day, your entire account, subscription, and task history gets fully deleted, and any remaining balance is automatically refunded to your original payment method. If your account predates the acquisition and stayed unchanged, you keep your account, but any task data generated since December 29, 2025 still gets wiped, and you lose access to Manus entirely for two days during the deletion window (August 23-24).

Manus built backup and restoration tools, and it's genuinely trying to soften the blow: free credits, a pro-rated refund for anyone whose subscription gets cancelled, and a "welcome back bonus" once you restore your data on August 25. To its credit, Manus is also being transparent about why this is happening. But none of that changes the core lesson for small business owners: a tool you were routinely using for real output could vanish, get restructured, or get sold out from under you with essentially no notice beyond an in-app message and an email you might have skimmed past.

This isn't really a story about Manus specifically. It's a story about a pattern. AI agent tools are still young companies, many are venture-funded, several have already been acquired or folded into bigger platforms this year, and none of them owe you permanence. If a tool generates a website, a research doc, or client-facing content you actually rely on, treat "it lives in their cloud" as temporary by default, not as a backup.

List every AI or cloud-based tool you use regularly for your business. For each one, confirm you have a local copy or an export of the output you'd actually miss, not just faith that the company will still exist in the same form next year.

Source: Manus

Mark Zuckerberg published a 6,500-word essay this week laying out his vision for "personal superintelligence," where everyone gets an AI agent that manages their health, career, finances, and hobbies. It reads like an inspirational manifesto. It also reads, according to Social Media Today's own writeup, "like a court defendant's plea more than an inspirational treatise for the future."

Here's the tension nobody at Meta is eager to point out. Zuckerberg's essay insists AI "isn't about automation or replacing human labor." That directly contradicts his own past statements about AI replacing staff at his company, and it sits awkwardly next to the fact that Meta cut 20% of its employees earlier this year while pouring roughly $1 trillion into AI development. The essay was written, in part, to get ahead of the backlash against AI and to argue against the kind of regulation that might slow Meta down.

None of that makes the underlying AI tools useless. Meta's ad-creative AI, its Business Agent, its Muse image tools genuinely save small business owners real time when used well. But there's a difference between a company building useful tools and a company writing a philosophical essay to soften you up before the next price increase or the next round of "AI-first" product pushes. Zuckerberg has done this exact move before, on child safety, on free speech, on the metaverse: present a values-driven case for whatever happens to also be good for Meta's business.

If you run a small business and you're being nudged, by every platform at once, to hand more of your marketing over to an AI agent, that pressure is real and it isn't going away. The right response isn't to refuse AI on principle. It's to stop letting the platform's enthusiasm set your adoption pace. You get to decide what earns a spot in your workflow, and the bar should be results you can measure, not a well-written essay from someone who profits every time you say yes.

Pick one AI feature Meta, or any platform, has been pushing into your account lately, something you haven't fully adopted yet. Run it on one real task this week, then look at the actual result before deciding whether it earns a bigger role in your marketing. Let the output convince you, not the sales pitch.

Source: Social Media Today

On August 5, 2026, Walt Disney Company and TikTok announced a global content-sharing deal that lets a curated group of TikTok creators use characters and scenes from hundreds of Disney films and shows, spanning Pixar, Marvel, Star Wars, and FX, inside their own short-form videos. It's the first agreement of its kind between a major traditional studio and TikTok.

The partnership runs through a new Disney Creator Ambassador program, and it moves in both directions. A curated selection of the resulting videos won't just run on TikTok, they'll also stream on Disney+ under a new vertical-video tab built for mobile viewing, marking the first time TikTok-style content has appeared on Disney+ at all. A US pilot starts in the coming months before the program expands to more markets.

For years, studios treated fan-made content built on their IP as a legal risk to police, not a channel to invest in. Disney flipping that stance, and building a formal pipeline that pushes creator content back onto its own flagship streaming service, is a signal that short-form video has crossed a credibility threshold. When a company this protective of its characters decides the safest move is to officially license the remix instead of chasing takedowns, that tells you where the audience, and the leverage, has shifted.

Here's the part every small business owner needs to hear clearly: this deal covers a select group of creators in an official ambassador program, not the general public. Nothing here gives you or your business permission to use Marvel, Pixar, or Star Wars characters in your own content. Unlicensed use of copyrighted characters and logos is exactly as risky as it was last week.

What you can actually borrow is the format. Disney is essentially building a farm system for a very specific kind of storytelling, short, character-driven, fan-voiced content, and studying what those ambassador videos look like once they start appearing is a free creative brief on what resonates right now, minus the trademarked cast. The part that stands out most isn't the characters, it's that Disney decided the fastest way to reach younger audiences was to hand its story to someone else's voice and someone else's platform. That's the same trade a lot of small businesses are still scared to make, letting a customer, an employee, or a local creator tell your story in their own words instead of your polished version.

This week, find one piece of content a customer or employee has already made about your business, unscripted and unpolished, and repost it or build a follow-up around it. You don't need a license from Disney to borrow the lesson: someone else's voice often earns more trust than your own.

Source: Deadline

As of August 5, 2026, Google closed the password-only door on the Ads API. If you log into Google Ads yourself through the browser, none of this touches you today.

Passkeys are the login method built to replace passwords, a cryptographic key tied to your device instead of something you type and something a phishing email can steal. Google is not asking you to change how you log into the Google Ads website in your browser. This rule targets a narrower, more technical door: the OAuth refresh tokens that let outside tools, ad management scripts, and reporting dashboards connect to your account and pull real numbers without you typing a password every time.

Here's the part that matters for keeping your sanity this week: tokens that already exist keep working exactly as they have been. Nothing breaks on August 5 itself. The exposure shows up the first time one of your connected tools has to mint a brand new token, whether that's a reporting platform you just signed up for, a script your agency wrote, or a Google Ads Editor bulk upload that hasn't run in a while.

If you log into ads.google.com yourself and manage campaigns by hand, this almost certainly does not touch you. If an agency, a freelancer, or a reporting tool pulls your numbers automatically, ask them one question this week: is a passkey already set up on the account that authorizes your API access? Google also flags up to a seven day trust delay before a newly created passkey becomes fully usable, so waiting until a token actually fails is the expensive way to find out.

This is exactly the kind of update that never makes it into a client meeting because it sounds like IT plumbing. But plumbing is where your reporting either keeps flowing or backs up on a random Tuesday. If you pay someone else to pull your Google Ads numbers, this is a five minute question that saves you a very confusing phone call in a few weeks.

Ask your agency or whoever built your reporting connection this exact question this week: is a passkey already set up for the account issuing our Google Ads API access, or are we waiting until something breaks to find out? Get the answer in writing, not a shrug.

Source: PinMeTo

If you run Snapchat ads, there's a new door into your account this week, and it doesn't lead to a person.

As of August 4, 2026, Snapchat launched its Ads MCP Server, an official connection point that lets outside AI tools like Claude, ChatGPT, and Gemini pull directly from a business's real Snap Ads account data instead of working off a vague text description. As Snapchat put it in its own announcement, AI tools "can only work with the information it can access," and without a direct feed, Snap performance was simply invisible to the AI assistants advertisers already use every day.

MCP, short for Model Context Protocol, is the connector standard Anthropic introduced in late 2024 that lets outside AI models talk directly to a platform's real data instead of relying on whatever a person types into a chat box. Snapchat is not first here and won't be last. Google shipped an MCP server for its Ads API last October, Amazon Ads followed in February, Pinterest and Meta both opened theirs earlier this summer, and TikTok has its own version too. Snapchat's entry means the four biggest self-serve social ad platforms in the US now expose campaign data to outside AI tools through the same connector standard.

The access controls matter more than the headline. Snapchat requires organization admins to authorize each AI agent separately and set its access level, and at launch every approved connection is read-only. When write access rolls out in a future release, admins will individually decide which agents stay read-only and which get permission to actually make changes inside the ad account. Right now these tools can only look. Soon, if you let them, they'll be able to touch.

For a small business owner or a one-person marketing team, the upside is real. Ask Claude or ChatGPT how your Snapchat spend compared to last week without opening five different ad dashboards. But the same convenience means your real budget numbers, targeting choices, and performance data are now sitting behind whatever permissions someone on your team clicked "approve" on, possibly without fully reading what they were approving.

Go check every AI tool currently authorized on your ad accounts this week, Snapchat, Meta, Google, wherever you spend money. If you don't recognize a connection or can't explain what it can see, revoke it. Read-only today doesn't mean read-only forever.

Source: Social Media Today

As of August 2, 2026, a new European Union law is officially in force. You don't sell into the EU. Keep reading anyway, because the fingerprints of this law are already all over apps you use every single day.

Article 50 of the EU AI Act now requires any business whose AI tools reach people in the EU to clearly disclose when content is AI-generated or AI-altered, whether that's a chatbot conversation, a synthetic image, or an edited video of a real person. The European Commission finalized its guidelines on July 20, and the penalties are steep: up to €15 million or 3% of a company's worldwide revenue, whichever is larger.

Here's why that matters if you've never sold a single thing in Europe. Meta, Google, LinkedIn, and Microsoft don't build separate products for separate regulators when they can help it. It's cheaper and safer to build one AI-disclosure system and turn it on everywhere at once. That's the real story behind a summer full of unconnected-seeming updates: LinkedIn's new "AI slop" flag button, Meta's mandatory AI labels on Facebook and Instagram ads, Instagram's "AI Info" tag rolling out across profiles and posts.

The labeling infrastructure being built right now doesn't check where your customers live before it applies. If you use AI to write a caption, generate a product photo, or clean up a video, there's a real chance the platform eventually flags it, automatically, whether you disclosed it yourself or not.

That's starting to matter to buyers, too. People are getting sharper at spotting AI content and warier of brands that hide it. A business that owns its AI use, here's how we use it, here's what a human still checks, reads as more trustworthy than one that gets outed by a label buried in a dropdown menu.

Go look at your last ten posts or ads this week. If AI touched any of them, the image, the caption, the video, ask yourself honestly whether you'd be fine with a visible label on it. If the answer is no, that's the thing to fix now, before the label shows up for you automatically.

Source: European Commission, Digital Strategy

Every local business owner running lead ads knows this exact pain. You get the lead, you're excited, and then nothing happens. They don't answer the follow-up text. They ghost the confirmation email. Somewhere between "submitted the form" and "showed up for the appointment," half your leads just evaporate.

Meta just gave you a way to close that gap. Inside Instant Forms, there's now a "Book Time" option in the ending section. You paste your Calendly or HighLevel scheduling link, Meta automatically detects which one it is, and it builds a live calendar right into the thank-you screen. The lead's name, email, and phone number carry over automatically so nobody re-types anything. HubSpot support is rolling out through early August, with more calendar providers expected after that.

Compare that to the old flow: fill out a form, wait for a redirect, load a new page, re-enter information you already gave, maybe give up before you get there. Every one of those steps bleeds leads. This collapses the whole thing into one uninterrupted screen.

For home services, contractors, healthcare, and local B2B, this matters more than almost any other ad update this year. An appointment booked at the moment of highest interest is a fundamentally stronger lead than a form sitting in a spreadsheet waiting for someone to call back. Speed to lead is one of the biggest predictors of whether a prospect ever becomes a customer, and this removes the single biggest speed bump in the process.

The catch: it's currently strongest on Facebook's mobile app placements, and not every advertiser has the option yet since Meta is rolling it out in phases. If you run lead ads and use Calendly or HighLevel, check your Instant Form editor this week. Under the Ending section, look for "Book time" as an Additional Action. If it's not there yet, check back in two weeks.

Source: Meta for Business

Picture this: you just opened a cafe, you're juggling three little kids and a hundred moving pieces, and you need window signage fast. So you generate some AI food photos and tape them up as a placeholder. Reasonable. Human. Exactly what an overwhelmed owner would do.

That's what happened at Grind & Unwind, a new Haight Street cafe in San Francisco. Owners AJ and Lyndsey Lozano put up AI-generated images of sandwiches and pastries in their window, meant to hold the space while they focused on getting the business open.

A Reddit thread titled "Yum, slop" went viral, with commenters comparing the uncanny food to loofahs, lungs, and mushrooms from a horror show. The mockery escalated into graffiti on their awning, and the couple spent roughly $700 repainting it.

In 2025, only 20% of consumers said heavy AI use would decrease their trust in a favorite brand. In 2026, that number is 40%, according to a Fractl consumer trust study. Distrust of AI-heavy marketing roughly doubled in a single year.

Here's the part worth sitting with. This isn't an "AI is bad" story. A restaurant that microwaves a frozen dinner and plates it fancy doesn't get bad reviews because it owns a microwave. It gets bad reviews because it served something bad and hoped nobody would taste the difference. AI is the microwave. It's a tool. The failure is skipping the taste test.

Nearly half of small business owners say AI helps most with marketing content, and that instinct is correct. The mistake at Grind & Unwind wasn't reaching for a fast tool while drowning in a new business, it was publishing the first draft without a single human glance before it went in the window. Before you publish anything AI made this week, run the "would I serve this to a customer's face" test. If you'd hesitate even a second, don't ship the first draft. Ship the one you'd actually stand behind.

Source: SFist

Most people think of Google Search Console as a website tool. Check keywords, check clicks, check if Google can find your pages. Google just quietly expanded that into territory that has nothing to do with having a website at all.

Search Console rolled out a new property type called platform properties, gradually since a July 7 announcement. It gives you a dedicated performance report for how your Instagram, TikTok, X, and YouTube content actually shows up in Google Search and Discover, complete with the specific keywords bringing people from a Google search straight to your social profile.

Here's the part worth sitting with. You don't need a website to use this. It was explicitly built for creators and accounts that live entirely on social media. That means even a business running Instagram-only or TikTok-only can now see, in plain numbers, what people are typing into Google before they land on a post.

This is separate from another long-standing SEO practice called sameAs schema, which links your website to your social profiles so Google understands they're the same business. Both matter, and both are the kind of unglamorous technical setup that either gets done in the first week of working with an SEO person, or never gets done at all.

That's really the point of today's issue. This tool is free, it takes minutes to connect, and Facebook and LinkedIn support is coming. If you've got someone managing your SEO and they haven't mentioned this to you yet, it's a fair question to ask them this week. If you don't have someone handling this for you, reach out to Sales Funnel Marketing and we'll walk you through exactly how to connect it.

Source: SearchScore Research

For most of this year, advertising inside ChatGPT meant one thing: money most small businesses do not have. Big deals with big minimums, run by a handful of national brands. That gate just came down.

As of the week of July 22, OpenAI's Ads Manager is open to any business willing to enter a credit card. There's no massive minimum spend anymore, and campaigns run on a cost-per-click auction instead of a flat rate, so you only pay when someone actually clicks. Early advertisers already running campaigns include names like Best Buy, Lowe's, and VistaPrint, but the barrier that kept the door shut to everyone smaller is now gone.

The targeting works differently than anything you're used to. Instead of picking keywords, your ad gets matched based on the context of the actual conversation someone is having with ChatGPT. Ask ChatGPT to help plan a kitchen remodel, and a relevant ad can appear at the bottom of the response, clearly labeled as sponsored and visually separated from the answer itself.

Before you get excited, know the limits. Ads only show to people on ChatGPT's free and Go tiers, so anyone paying for Plus or higher never sees them. Inventory is still capped while OpenAI protects the user experience, and advertisers report getting far less reporting data back than they're used to from Google or Meta. Analysts still expect total chatbot ad spend industry-wide to stay under a billion dollars this year, so this is not a channel about to replace your Google or Facebook budget.

That's exactly why now is the moment to test it, not skip it. A new ad channel is cheapest and least crowded before everyone else figures it out. A hundred dollars in ChatGPT ads this month teaches you more about whether this channel works for your business than any article, including this one, ever will.

Source: B2The7

In the early 2010s, hashtags were basically free advertising. Slap enough of them on a post and strangers would find you. That era is over, and Instagram just made it official.

In December 2025, Instagram capped posts at five hashtags and said using too many irrelevant or generic tags can actually hurt a post's performance. Instagram's own leadership has said hashtags don't meaningfully move reach anymore. Discovery now runs almost entirely on the algorithm reading behavior signals like watch time, saves, and shares, not the tags you typed in.

That doesn't mean hashtags are dead, it means sloppy hashtags are dead. The brands still winning with them use a handful of specific tags. Skincare brand KraveBeauty built an entire community around its own branded hashtag, #PressReset. Snack brand SmartSweets pairs its branded tags with a single relevant community hashtag instead of a wall of generic ones.

For a small business, that means throwing thirty hashtags at every post isn't a growth strategy anymore. A better move: pick one branded hashtag you own, one or two niche tags your actual customers search, and skip the rest. Put real keywords in your caption and bio instead, since that's increasingly what surfaces you in search now, not a pile of hashtags.

Test what's actually working before you lock in a hashtag list forever. Type the # symbol while writing a caption and Instagram will suggest relevant tags directly. Then track which posts using which tags actually bring in engagement or followers over a few weeks, instead of guessing and repeating the same list forever.

Source: Shopify

While everyone was still arguing about ChatGPT, Google quietly shipped three new Gemini models in a single day, and one of them is already working inside Google Search.

Google introduced 3.6 Flash, its new workhorse model for everyday tasks, 3.5 Flash-Lite, its fastest and cheapest model yet, and a specialized cybersecurity model. The headline number: 3.6 Flash uses 17% fewer tokens than its predecessor while performing better, which in plain English means faster answers at a lower cost to run.

"Developers and customers building production AI agents need higher token efficiency, lower latency, and more reliable performance," said Tulsee Doshi, Senior Director of Product Management for the Gemini team at Google.

Here's the part that actually touches your business today. 3.5 Flash-Lite isn't staying locked inside developer tools, it's rolling out directly into Google Search right now. That's the model doing more of the thinking behind AI Overviews and AI-generated answers, the same answers increasingly showing up above your website in search results.

If you use the free Gemini app to draft social captions, summarize a contract, or brainstorm content ideas, you're getting a meaningfully faster, more capable assistant starting today, at no extra cost. Google also confirmed it has started training Gemini 4, calling it their "most ambitious pre-training run yet." Translation: Google isn't standing still, and neither is anyone else in the AI race.

Source: Google

Ask ten business owners whether they should spend on Facebook ads or a local radio spot, and you will get ten confident opinions. Look at the actual data, and the real answer is not which channel wins. It is that both are losing to something free.

Stop asking whether Facebook ads beat a radio spot. Both are paid, and paid is losing to something free: a referral from a real customer. Put a referral system next to whatever you are already paying for, and you will out earn business owners still stuck picking sides.

The budget fight is already settled. Small business marketing dollars have shifted hard toward digital and social, and when a paid social campaign is run well, a 5 to 1 return is the standard industry benchmark, according to Sprout Social's 2026 ROI research.

Here is the part nobody selling ad space wants you to see. Nielsen data shows more than 9 in 10 consumers trust a referral from a friend or family member over anything else, while only 36% trust an online video ad, the exact format most social budgets get poured into.

Referral marketing converts at 3 to 5 times the rate of other channels, and businesses running a real referral program saw revenue jump by an average of 86%. For small businesses specifically, referral programs return 3.6 times their cost, a real number, not a vibe.

So the fight was never social versus traditional. It is paid versus earned. Every paid dollar, social or traditional, should be paired with a system that turns customers into referrers: a review ask after every job, a simple "who do you know," an incentive for sharing. The channel is just the delivery truck. Trust is the cargo.

Source: Sprout Social

One business owner watched 4,651 real reviews collapse to 63 in about a day, and Google admits its own system did it.

Screenshot your current review count and a handful of your best reviews today. And if you spot a fake review on your own listing this week, think twice before hitting report, that action looks like exactly what triggers the wipe.

Since around July 1st, the Google Business Profile forums have filled with complaints from owners watching reviews vanish overnight. Some lost a handful, some lost thousands, and several report their star rating dropping straight to zero with new reviews blocked from posting afterward.

"We had approximately 4,651 legitimate customer reviews collected naturally over many years from real customers. Suddenly, within about 24 hours, our public review count dropped to only 63 reviews," one owner wrote in the forums. Google confirmed to Search Engine Land this is intentional, not a random bug: "When our systems detect suspicious reviews, we take a range of actions including removing reviews and temporarily pausing reviews on the profile to prevent further abuse. We are investigating the issue and will restore any reviews that were incorrectly removed."

A Google Product Expert tracking the complaints says the pattern often starts right after an owner reports a fake or spam review on their own listing. The report itself seems to trip the spam system, which then hides the whole review history and blocks new ones from posting. No fix timeline has been given, and reviews are one of the biggest local ranking and trust signals a small business has.

Until Google says more, hold off on mass reporting fake reviews on your own listing, and keep proof of your current numbers on hand in case you need it for a restoration request.

Source: Search Engine Land

For three years, using ChatGPT meant asking a question and doing the work yourself with whatever came back. That deal just changed.

Give it a real project this week, not a quick question. Hand ChatGPT Work a task you've been avoiding, a report, a spreadsheet, a client deck, and see how far it gets before you have to step in. That's the actual test of whether this is worth your time.

On July 9, OpenAI released GPT-5.6, its new model family, in three tiers: Sol, the strongest, built for heavy lifting; Terra, everyday work at half the cost of the last flagship; and Luna, the fastest and cheapest, still surprisingly capable. The rollout had been delayed at the request of the U.S. government and went broadly available the same day.

The bigger deal for a small business owner shipped alongside it: ChatGPT Work. It's an agent that takes an outcome you want, not just a question, gathers what it needs from your connected files and apps, breaks the job into steps, and stays on it for hours without you re-prompting every five minutes. Ask for a client deck or a quarterly report and it hands back something finished, not a rough draft you still have to assemble yourself.

Access is rolling out first to Pro, Enterprise, and Edu plans, with Plus and Business getting it within days. The desktop app is already live on Mac and Windows, even on the free plan.

Source: OpenAI

For twenty years, good product photography meant hiring a photographer and paying the bill before a single ad ever ran. Meta just handed every advertiser on its platform a version of that photographer that works for free, and it lives right inside the tools you already use.

On July 7, Meta launched Muse Image and previewed Muse Video, the first media generation models built by its Superintelligence Labs research group. Muse Image is available today in the Meta AI app, on meta.ai, in Instagram Stories in the US, and on WhatsApp in limited countries, with Facebook coming soon.

The advertiser-facing version is what should get your attention. Meta says Muse Image can restyle ad images inspired by your existing creative, and it can generate still images pulled straight from your video assets. The headline feature: it lets Meta AI take a photo of a shopper's actual room and show your product staged inside it, pulled from your real product catalog.

This is already at real scale. Meta says more than 8 million advertisers are already using at least one of its generative AI ad creative tools, and Muse Image currently ranks No. 2 on Arena for text-to-image and both single and multi-image editing, based on human preference rankings as of July 5.

Every image Muse Image generates in the Meta AI app or on meta.ai carries Content Seal, an invisible watermark that survives cropping, compressing, and resizing, even a screenshot, so there's a way to check whether an image was AI made.

Source: Social Media Today

More of your future customers are asking ChatGPT for recommendations instead of Googling. So who does the AI actually recommend, and why? Semrush just bought us all the answer.

AI doesn't recommend the business with the best website. It recommends the business that's described consistently across reviews, directories, and community sites. This week, check that your name, services, and story match everywhere you're mentioned online, and start stacking reviews.

The company's new AI Visibility Index analyzed 126 million real AI prompts to see which sources chatbots pull from when they build an answer. The headline finding: chatbots cite specialists over generalists, and they lean hard on third-party sources, review platforms, community sites like Reddit and Quora, directories, and reference pages, rather than a brand's own website.

Each AI tool also has its own favorite sources, so where you show up matters as much as showing up at all.

Here's the part that should change how you think: the brands AI describes most consistently aren't the ones publishing mountains of content. Per Semrush, they "publish the right content on the right surfaces, owned and third-party, so AI systems receive a consistent, coherent signal about who they are and what they do."

Source: Social Media Today

Ever wonder why a post from a business you've never followed lands in your feed like it belongs there? That's not an accident. That's the whole strategy now.

Every public post you publish now gets tested on people who have never heard of you, on every platform, not just Facebook. Write for the stranger, not the regular. And know this: posts set to "friends only" are exempt from the whole system, for better and for worse.

For years, social media ran on the "follow graph": you saw posts from people and pages you chose to follow. That era is over. Every major platform, Facebook, Instagram, TikTok, LinkedIn, YouTube, has shifted to an "interest graph," where AI decides what you see based on what you watch, save, and share, not who you follow. Meta's own engineers describe it as recommendation systems pulling from "tens of billions" of pieces of content outside your network. In practice, your post gets an audition: it's shown to a small batch of non-followers first, and if they watch, save, or send it to a friend, the platform widens the audience. Perform again, and it widens again.

Why did every platform make the same move at once? Because TikTok proved that interest beats friendship for holding attention, TikTok never needed you to follow anyone to keep you scrolling for an hour. The others copied the playbook to survive. And there's a business reason the platforms don't advertise: recommendations give brand-new content from small accounts a fighting chance (Meta literally says every new piece of quality content gets a shot at "a large, relevant audience"), which keeps small businesses and creators posting, and keeps the feed full of free inventory to sell ads against.

Now the question Scroll readers keep asking: "If my posts are set to friends-only, does the algorithm still push them to strangers?" No, full stop. Recommendation systems only pull from PUBLIC content. Friends-only posts, private accounts, and limited-audience posts are not eligible to be shown to non-followers, your privacy setting overrides the algorithm every time. Which cuts both ways: your family posts stay private, but if you're running your business from a personal profile locked to friends, you are invisible to the single biggest growth engine on social media. Your business Page and public account are where the auditions happen, that's where your best work needs to live.

Source: Meta AI

Google just rewrote the rules on its own AI, and it did it quietly.

Stop assuming Google's automated tools are hands off. The new terms spell out that you're on the hook for reviewing, approving, or removing anything Google Ads generates automatically, headlines, sitelinks, whole asset groups. If an AI written line makes a claim you can't back up, that's now on your account, not Google's.

The old Google Ads terms, last updated back in 2018, treated automation like an optional feature you could opt into or ignore. The new terms, effective July 1, treat it as the system you're already running inside. Whatever you type into AI Brief, Ask Advisor, or any of the conversational campaign tools can now be used to shape your account. So can the URLs you hand over and whatever Google crawls on your site once you give it permission.

Here's the part that should actually worry a small business owner. Your website isn't a curated ad library. It's got old blog posts, expired promotions, a careers page, legal disclaimers, all of it crawlable. Automation doesn't know the difference between your best offer and something you took down eight months ago. If it pulls the wrong thing into an ad and that ad runs, the terms are clear the responsibility sits with you, not Google.

Go check your Performance Max landing page report and your Search terms report filtered by landing page. See what pages Google's actually pulling from. If something shows up that was never meant to sell anything, exclude it with a page feed or a URL exclusion at the campaign level. Set a weekly review on new campaigns until you know what the automation is doing, then drop to monthly once it settles down.

Source: ZATO PPC Marketing

Here's the quiet change that's about to make a lot of marketing reports look like a mistake.

Before you trust this month's report, export or screenshot your current Reach numbers as a baseline. When the new “Viewers/Views” figure shows up, it will likely read higher, because it counts repeat views, not just unique people. Stop comparing this month to last month, and judge your content on what actually moves your business: saves, shares, DMs, calls, and clicks.

Meta is pulling its legacy “Reach” metric, along with old Video Impressions and Story Impressions numbers, out of the Graph API, the back-end pipe that feeds nearly every third-party tool you use. Metricool, Hootsuite, Buffer, Sprout, Iconosquare, the dashboard your VA pastes into your monthly report: they all pull from that pipe. In its place, Meta is rolling in a new “Viewers” metric meant to count how many people saw your content the same way across Facebook and Instagram.

Why should a busy shop owner care about an API? Because “Reach” was unique people, and “Views” counts every time your post got seen, same person, three scrolls, three views. Your headline number is going to jump, and it won't mean more humans saw you. Anyone glancing at a report and thinking “we're crushing it” is about to be fooled by a counting change, not a growth spurt. And anyone comparing June to May is comparing two different rulers.

So do two things this week. First, save a clean snapshot of your current numbers before they vanish, that's your only honest before-and-after. Second, pick the two or three outcomes that actually pay your bills, booked calls, form fills, DMs, foot traffic, and make those the line you watch. Reach was always a vanity number wearing a business suit. This is your excuse to stop bowing to it.

Source: Social Media Today

Quick story while it's fresh.

If any part of your business runs on a single AI tool, you just got a preview of how fast it can disappear. This wasn't a server crash or a billing snag. A federal directive ended the model overnight. Spend an hour this weekend writing down what AI you use and making sure you've got a working backup.

At 5:21pm Eastern last night, the US Commerce Department sent Anthropic an export control directive. The letter was signed by Commerce Secretary Howard Lutnick, written with help from the Bureau of Industry and Security. It told Anthropic to cut off any foreign national from using its two most powerful AI models, Claude Fable 5 and Claude Mythos 5. Anthropic said the only way to actually comply was to shut both models off worldwide, for every customer, foreign or American. Just gone.

Fable 5 came out Tuesday. Three days is how long the strongest publicly available AI model on the planet lasted before the government took it down.

Some background. Mythos 5 is the unrestricted version. Anthropic was only sharing it with about 50 vetted partners through a program called Project Glasswing. That list includes Amazon, Apple, Google, Microsoft, and CrowdStrike, and they were using it to hunt down security holes in their own software before the bad guys could find them. Fable 5 was the toned-down public version with guardrails baked in. The government claims Fable can be jailbroken to do the same security-hunting work. Anthropic basically said, sure, sort of, but every other big model can do that too, including OpenAI's GPT-5.5. They got verbal evidence only. No paper, no specifics, and the example they were shown was, in Anthropic's words, the model being asked to read a codebase and find software flaws. That's a thing cybersecurity defenders do every single day with regular ChatGPT.

The whistle blew yesterday, and the platforms had clearly been waiting for it.

For the next five weeks, your customers' feeds belong to the tournament. Stop posting like it's a normal week, attach your content to match days, watch parties and game-night energy instead of competing against them. And keep an eye on your ad costs around big matches, because the auctions are about to get loud.

Meta just announced World Cup features across every app it owns: live score alerts and a dedicated tournament community on Threads, a special search hub on Instagram with its own tournament button on football videos, a double-tap "Football Mode" on Facebook, and a Live Updates feature in Messenger that pipes goals and red cards straight into your group chats. WhatsApp got themed stickers and a tournament channel directory last week. Snapchat rolled out its own activations the day before, and TikTok signed up creator correspondents weeks ago. The tournament runs through mid-July, 48 teams, the biggest edition ever played.

Here's why this matters even if you couldn't care less about soccer: when every major platform builds dedicated surfaces for one event, the algorithm's attention follows. Tournament-adjacent content is about to get distribution help, and everything else gets squeezed. A business posting its usual Tuesday promo is now competing with the most-watched television event in human history, inside the same feed.

The play isn't to fake being a soccer brand. It's to borrow the moment: match-day hours posts, watch-party energy, "open during the game" angles, prediction contests. One caution, "World Cup" and "FIFA" are fiercely protected trademarks, so in paid ads say "the tournament" or "the big game" and you'll stay out of trouble.

Source: Social Media Today

For fifteen years, the deal was simple: you post, the algorithm decides who sees it, and nobody gets a vote. That deal just changed.

Stop posting random one-offs. Instagram now files every post under a topic, and users can add or delete those topics from their feed. If your content doesn't clearly belong to the topics your customers actually pick, you don't get demoted, you get removed. Pick your lane and make every post obviously about it.

Instagram announced it's extending “Your Algorithm”, the topic-control dashboard it tested on Reels last October and Explore in April, to the main feed. Users get a list of topics Instagram thinks they care about, and they can add the ones they want more of and delete the ones they don't. Whatever they choose directly rewires what shows up when they open the app.

Here's why this matters for a local business more than for a big brand: topics are now the currency. When a homeowner adds “home renovation” to their feed topics, every contractor whose content clearly fits that topic just got a standing invitation into that feed. The plumber posting a mix of memes, vacation photos, and the occasional water heater tip? The system can't file that account under anything, so it shows up nowhere.

What to do this week: audit your last ten posts and ask one question, could a stranger name your topic in two seconds? If not, tighten it. Use clear spoken keywords in your Reels, put your niche in your captions and on-screen text, and stay in your lane long enough for the system to learn what you are. Industry analysts are betting most users won't touch the settings, which means the ones who do are your highest-intent customers, literally raising their hands for your category.

Source: Social Media Today

Yesterday Anthropic flipped the switch on Claude Fable 5, and quietly changed what a small business can do with AI overnight.

Stop renting five different AI tools this week. The new top-tier model is included free for paid Claude users through June 22, so pick one real job you've been avoiding (a month of captions, your whole FAQ, a messy spreadsheet) and hand it the entire thing, not a sentence at a time.

On June 9, Anthropic released Claude Fable 5, the first model in a new tier that sits above its already-strong Opus line. It rolled out everywhere at once, the Claude API, Amazon Bedrock, and GitHub Copilot, and here's the part that matters for you: it's included free for Claude Pro, Max, Team, and Enterprise users during an introductory window from June 9 through June 22.

Translation for a business owner: the gap between "AI toy" and "AI employee" just closed. This model is built to take a whole task and run with it, reading your messy documents, working through multi-step jobs, and checking its own work before it hands it back. That's the difference between asking it to write one caption and asking it to plan, write, and organize a month of them while you go run your actual business.

What to do this week: don't overthink the tech. If you already pay for Claude, the upgrade is just there, no new signup. Block 30 minutes, grab the single most tedious recurring task on your plate, and give it the messy raw version (the export, the notes, the half-finished doc). The free window is the cheapest R&D budget you'll ever get. After June 22 the premium tier carries premium pricing.

Source: Anthropic

For three years Siri was the assistant everyone made fun of. Apple just stopped the joke cold.

When a customer asks an AI agent for a recommendation, it reads your reviews, hours, and website to decide whether to surface you, nobody scrolls to page two. Your job this week is boring but decisive: make sure your business name, hours, services, and reviews say the exact same thing everywhere online. Consistency is what AI trusts, and trust is what gets you recommended.

At its developer conference on June 8, Apple unveiled "Siri AI", a complete rebuild that can hold a real conversation, see what's on your screen, pull details out of your texts and emails, search the open web, and actually take actions across your apps. Point your camera at a plate of food and it estimates the nutrition; ask it to split a bill and it does it through Apple Cash. Under the hood, Apple is blending its own models with Google's Gemini, running on-device and through private cloud. It hits beta later this year, U.S. and English first (not the EU or China at launch).

Here's why a shop owner should care more than a tech blogger. For twenty years, "getting found" meant ranking in a list a human scrolled. Agentic assistants don't scroll, they read, judge, and hand back one or two answers. That means the businesses with clean, consistent, well-reviewed information online don't just rank higher; they become the recommendation. Everyone else becomes invisible, not because they're worse, but because the AI couldn't verify them.

So do the unglamorous work now. Audit your Google Business Profile, your website, and your social bios so the hours, services, address, and phone number match to the letter. Ask happy customers for reviews this month, volume and recency are exactly the signals an AI leans on. You can't optimize for Siri AI directly yet, but you can be the kind of clean, trusted listing every agent wants to recommend.

Source: Apple Newsroom

Every few years Google tries to become a social network. This time it might actually stick, because it's bolting the idea straight onto Search.

Search is turning into a feed. Ranking for a keyword used to be the finish line, now Google wants searchers to follow you and get your posts pushed back to them later. Stop treating your Google Business Profile like a phonebook listing and start running it like a channel: fresh posts, photos, and answered questions every single week. That habit is what earns the follow when this reaches you.

Google is testing "Search Profiles", social-style profiles that live right inside your Search results. They pull your latest articles, videos, and social posts into one branded spot, and they add a Follow button so the people who find you once get your content surfaced again later in their Discover feed. Social Media Today broke down the rollout this week.

Two shifts are hiding in this. First, Google's AI Overviews are now answering questions at the top of the page and starving the blue links underneath, so a "follow" relationship is Google's peace offering to anyone watching their organic traffic dry up. Second, the door isn't open to everyone yet: right now you need 100,000 followers on YouTube, Instagram, or X (300,000 on TikTok), US only. Most local owners don't clear that bar, yet.

Don't wait for the invite to act. The version of this you already own is your Google Business Profile: the posts, the photos, the Q&A, the reviews. That IS your search profile today. Feed it weekly like you'd feed a social account, because the follow button is going to come down-market, and the businesses already acting like a publisher will be first in line.

Source: Social Media Today

For years, "AI on the phone" meant a robot menu that made customers want to throw their device across the room. That era is over.

Stop sending calls to voicemail. The technology to staff your phone around the clock with an AI that reasons, books appointments, and speaks your caller's language just went live, but you reach it through a voice-agent platform, not from the company that built it. If you run a service business, missed calls are almost certainly your biggest unplugged leak.

OpenAI shipped three new realtime voice models that any app can plug into: one (GPT-Realtime-2) that reasons through a request like a sharp human and books or looks things up mid-conversation, one (GPT-Realtime-Translate) that translates speech across 70+ input languages into 13 output languages without the speaker pausing, and one (GPT-Realtime-Whisper) that transcribes a call live as it happens. They're not a demo. They're available right now.

Sit with that number. The leads are already calling you. The marketing already worked. The phone just rang into the void because you were under a sink, on a job site, or asleep. An AI that picks up on the first ring, qualifies the lead, and even handles a Spanish-speaking caller in their own language isn't a luxury anymore, it's plugging a hole that's been quietly draining your revenue for years.

Now the catch, because you deserve the truth: OpenAI released this as a developer tool, not an app you sign up for. You access it through voice-agent platforms, the Goodcalls, Synthflows, Retells and CloudTalks of the world, that wrap the models into something you can actually use. The economics are wild: the live-translation model runs about $0.034 a minute (roughly $2 an hour of perfect interpretation), and a full AI receptionist lands at $65, $499 a month versus $3,750, $4,000 for a human one. Real and bookable today, through a platform, not a download.

Source: OpenAI

For years the advice was “answer your DMs faster.” Meta just deleted that chore.

Turn it on, but train it before you trust it. Feed it your real prices, hours, FAQs, and the exact way you talk to customers, then watch the first 50 conversations like a hawk. The businesses that win here aren't the ones who automate the most; they're the ones whose agent sounds like a human who actually works there.

At its Conversations conference in London on Tuesday, Meta unveiled the Meta Business Agent, an AI that lives inside your WhatsApp, Messenger, and Instagram inboxes and actually does the work. It answers questions, pulls products from your catalog, qualifies leads, books appointments, and yes, closes sales. It speaks your customer's language, mimics your brand's tone, and knows when to tap you in for the human touch. There's even a morning briefing summarizing every overnight chat you missed.

Here's why this lands harder for a small operator than for a big brand: you don't have a night shift. A roofing company, a med spa, a bakery, the lead that DMs at 9pm usually goes cold by morning because nobody's there. This agent answers in seconds, every time, and Meta tested it with over a million small businesses in India, Mexico, and Brazil before flipping it on globally this week. It plugs into Shopify and Zendesk too, so it can take real action, not just chat.

The catch, and this is the line buried in the announcement, it's free now, but Meta says it “may become a paid subscription in the coming months.” Translation: get in while it's free, learn it cold, and build the muscle before there's a price tag on it. Start small, let it handle FAQs and hours first, keep your hands on anything involving money or a quote.

Source: Business Today

For years you had two ways to control how Google tracked your visitors. In 11 days, you'll have one, and it's the one most business owners have never touched.

Stop assuming a setting buried in Google Analytics is protecting your data. After June 15, Google Ads ignores those Analytics toggles entirely and obeys only the consent banner on your site. Pull up your banner this week, click through “Accept” and “Reject” yourself, and confirm it's actually telling Google the truth, because a broken banner now means broken numbers everywhere.

Google is rolling out what it calls “destination-specific controls” on June 15. In plain English: the master toggle inside Google Analytics that quietly let you throttle how ad data gets collected is going away. From that date, Google Ads stops looking at your Analytics dashboard and listens only to the consent (cookie) banner code on your website. If the banner says yes, Google goes all-in on linking and audience building. If it says no, or if it's misconfigured, Google goes dark.

Here's the part that should make you sit up: nothing in your dashboard will warn you. Your reports will keep loading. The numbers will just quietly get less accurate, fewer tracked conversions, smaller retargeting audiences, worse automated bidding, because the AI is now flying on whatever your banner feeds it. If you run any paid ads, or you pay someone to run them for you, this is your money on the line.

Three moves before June 15. One: open your own site and actually click “Accept,” then “Reject,” on the cookie banner to confirm it behaves. Two: make sure your privacy policy mentions associating data with Google sign-in information. Three: if you use an agency, email them today and ask one question, “Are we ready for the June 15 consent change?” If they hesitate, that's your answer.

Source: Launch

LinkedIn just put a clock on something a lot of business owners do without thinking: hitting “go live” on a whim.

If you ever tap “go live” to talk to customers, that button disappears in three weeks. Starting now, build the habit of scheduling a LinkedIn Event first, even one set for ten minutes out, so your team isn't scrambling on a launch day in late June.

As of June 22, 2026, you will no longer be able to start a live broadcast on LinkedIn without first attaching a scheduled Event, a page with a title, a description, and a start time. The instant, no-setup “go live” option is being removed entirely. You can still go live on short notice; you just have to spin up an Event first, even one scheduled for a few minutes from now.

Here's why this actually helps the little guy: a scheduled Event is promotable. It gets a landing page, a reminder system, and a window where people can register before you ever go on camera. The spontaneous stream got whoever happened to be scrolling. For a local business owner doing a product demo, a Q&A, or a behind-the-scenes look, that pre-event runway is the difference between three live viewers and thirty.

Do this before June 22: practice the new flow once this week. Create a test Event, schedule it for ten minutes out, and go live from it so the steps are muscle memory. Then build a simple repeatable title format for your broadcasts so setup never becomes the reason you skip going live.

Source: Social Media Today

A month ago, the smart move on ChatGPT ads was “interesting, sit tight.” That advice just expired.

Put a conversion pixel on your site this week, even if you have zero plans to advertise on ChatGPT yet. The early-access auction is thin and cheap right now, and the businesses with tracking already firing get smarter delivery before everyone else floods in and bids the price up.

In about two months OpenAI has gone from “we're testing ads” to a full self-serve ad platform: cost-per-click bidding, daily budgets, geo-targeting down to the ZIP code, a conversion pixel, a server-side Conversions API, and now cost-per-action bidding. The last piece, conversion-optimized campaigns that steer your ads toward the people most likely to actually buy, begins rolling out June 5. Until now you could only pay for impressions and clicks. You were buying attention and hoping it turned into something. That's over.

Here's why this lands on a small business owner's desk and not just an agency's: your customers already talk to ChatGPT before they ever fill out a form. “Is my current provider ripping me off?” “What should I pay for this?” “Do I even need this service?” Those are bottom-of-funnel conversations with real money behind them. The only reason to wait was that you couldn't prove the spend paid off. Now you can, recommended bids are sitting at just $3, $5 a click while the auction is still quiet.

What to do: get the pixel or Conversions API on your site, define a real conversion (a booked call, a quote request, a form submit, not a pageview), and confirm it's firing before you spend a dime. One warning from the people doing this daily: tracking set up badly is worse than none at all, because it quietly optimizes toward the wrong action and burns budget while looking busy. Measure twice, cut once.

Source: PPC Land

On May 5, OpenAI flipped a switch that changes who gets to play in AI advertising, and it's not just the big agencies anymore.

Stop writing broad “we're awesome” brand copy for this channel, it will burn your budget. ChatGPT ads reward one thing: a specific, intent-matched offer dropped in front of someone who's literally asking about your category. If you sell a clear service to a clear customer, you can test this with a few dollars and learn before your competitors even hear about it.

The pilot that launched in February required a $50,000 minimum spend. That number alone told you who it was built for: holding companies and national brands. As of May, that wall is gone. The ChatGPT Ads Manager is now self-serve with no minimum, open to any U.S. business that wants to test it. They also added cost-per-click bidding on top of the old CPM model, which is exactly what a performance marketer needs to take a channel seriously.

Here's why this matters for a small business: the ads appear in clearly labeled boxes at the bottom of AI answers, and targeting is contextual, based on what the person is actually talking about in that conversation, not the keywords you guessed at. The catch is that advertisers get zero user data back. No emails, no IPs, no precise locations. Just aggregated performance. That means the measurement game is different, and last-click thinking will leave you blind.

What to do this week: don't bet the budget. Pick your single sharpest offer, the one a customer would search for by intent, write three versions of it, and run a small CPC test. Bids under $3 tend to get no impressions, so plan for $3, $5 a click and treat it as paid research, not a scale play. Yet.

Source: Workshop Digital