September 18th-September 24th // Estimated Reading Time: 10 minutes
In This Edition!
Meta finally has an AI hit with Muse!
New and improved (and cheaper) models keep on coming…
Top Headlines🔥
After spending billions on AI, Meta has a hit with Muse
Source: Forbes
September 22nd, 2026
Summary: Two weeks after release, Meta's Muse AI agent passed ChatGPT to become the top free app in Apple's App Store. The agent handles tasks like buying products, planning trips, and completing online paperwork, so it needs access to sensitive areas like email and bank accounts. Meta shares have climbed 30%+ this month, the stock's strongest run since 2013, driven largely by Muse's popularity. That's a relief for investors, who have been nervous about Meta putting as much as $145B into capital spending this year, most of it tied to AI. Amazon cut off Muse's access to its store and says Meta never sought approval. However, other retailers are opening their doors. Shopify struck a deal to allow Muse to buy from its merchants using Shop Pay. Meta says Walmart, Best Buy, Sephora, Gap, and many more have also signed on. At Meta Connect on Wednesday, Zuckerberg tied Muse to Meta's hardware, positioning the company's smart glasses as the best way to use it. He also announced the launch of a new, voice-controlled, Tamagotchi-style AI companion gadget called Muse Charm, which will ship in December.
Zuckerberg said that Muse is free with a large token allowance, and over time Meta expects to earn money by taking a small fee on the transactions Muse handles. There are $20 and $100 monthly subscription tiers for heavier use. Zuckerberg also said that Muse lives in a secure environment where even Meta can’t see its data.
Opinion: Muse puts a user-friendly AI agent into the hands of consumers for the first time. Unlike traditional agentic offerings, Muse pushes hard to learn about you and then surfaces important tasks it can do for you. It does so intuitively and seamlessly. Investors are clearly loving the early user adoption of Muse and rewarding the stock. Just goes to show that if a company is going to run up its AI capex, it needs to show results in the form of AI infrastructure revenue or user adoption. Until now, Meta had neither, just an abstract story around how AI is helping its ads business. Perhaps Meta has finally landed on AI product market fit.
Marketers, keep an eye on how Muse affects behavioral patterns, like how users research and shop. Muse could create all sorts of new earned and paid marketing opportunities (with a lot of that money going directly into Meta’s pocket), while closing the door further on legacy channels. How long before users see sponsored products in Muse? How long before Muse data starts feeding the Meta ad beast?
Meet U of Digital at BRXND NYC (and save 30%)
The U of Digital team will be at BRXND NYC on November 5th for a day focused on AI + marketing. BRXND is giving our readers 30% OFF with code UOFDIGITAL30.
New Products & Features 🚀
What It Does: Anthropic's new Opus 5.5 model performs about as well as Fable 5.1, its strongest public model, on most work. Typical jobs are roughly 40% cheaper than on Opus 5, and it generates responses 30% faster. Opus 5.5 also writes more clearly and is less likely to hallucinate. Anthropic raised the usage limits for paid Claude plans. New versions of Sonnet and Haiku, Anthropic's smaller and cheaper models, will come in the next few weeks.
Quick Take: Companies have been saving Anthropic's priciest model, Fable, for tough jobs and sending routine work to cheaper ones. Opus 5.5 looks like Anthropic's answer: near top-tier quality at a lower price. Marketers need to constantly be evaluating the models they’re using for various tasks with every new release to ensure they’re on the efficient frontier of AI usage.
What It Does: OpenAI launched GPT-6 Sol and GPT-6 Luna, lower-cost versions of GPT-6 Astra, its top model released earlier this month. Developers will pay half what they paid for the previous Sol and Luna models. OpenAI says Sol makes about half as many factual errors as the last version. Both models also write more clearly. They're available now in ChatGPT Work and Codex for paid users, but not yet in the standard ChatGPT chat.
Quick Take: OpenAI and Anthropic released cheaper models within hours of each other, with nearly identical pitches: lower prices, fewer mistakes, and better writing. Another idea: marketers should consider using AI routers that can quickly pick up on which models are good for which tasks, and route AI work accordingly.
What It Does: Google is giving advertisers more visibility into AI Max, its set of AI tools designed to help marketers run Google Search ads more effectively. A new Google Ads report will show the full path of each search ad in one place: the keyword that triggered it, the ads people saw, and the page they landed on. Google hasn't said when it will launch, only that more details are coming later this year.
Quick Take: The AI black box opening up helps marketers better understand the value of their search ads. But be wary, because the reporting is all still being generated by the media provider.
What It Does: Ask Neura lets advertisers ask questions about fraud, brand safety, viewability, location, and attention across social, streaming TV, and the open web. It can explain sudden spikes or drops in KPIs, and give a quick rundown of how a campaign is performing before linking to the DV dashboard with the full breakdown. DV data can also be used in outside AI tools like Claude. On the horizon is an agent that will adjust campaigns on its own, but only in ways advertisers have approved in advance.
Quick Take: This is the “AI playbook” for ad tech companies as of now: build in-platform AI capabilities while also enabling marketers to use their AI tool of choice from the outside. There’s no clear market consensus on what the “right” answer is, as different customers prefer different approaches. Ultimately, we believe outside AI tools will win.
What It Does: Pinterest's new ads will appear at the top of search results and on the page that opens when someone taps a Pin. Many Pinterest searches start with a photo instead of words. Pinterest's AI reads the image, including product type, color, and style, and combines it with any keywords to match people with relevant ads. Advertisers bid for the top spot. The beta opens in all Pinterest ad markets in the coming weeks.
Quick Take: Having high-quality product imagery, and product imagery that closely resembles images frequently pinned on Pinterest, just become much more important! A picture is worth a thousand words. That’s Pinterest’s hope with Visual Search Ads.
AI Use Case of The Week💡
Coleman turns a fake AI ad into a 105% sales jump🔒
The Setup: In June, an AI-generated, fake ad showed a 200-foot Coleman backyard lazy river selling at Costco for $999.99. The product was fake, but the content spread quickly. Soon, copycat sellers started pushing lookalike versions. Coleman had two options: set the record straight, or have fun with it and risk linking its name to an AI hoax.
The AI Solution: To evaluate the situation, Coleman's team used the OODA (observe, orient, decide, act) loop, a quick decision framework developed by the US military for combat pilots. After concluding that audiences would continue talking about the lazy river, the team used Claude and ChatGPT to create new lazy-river assets for its website, TikTok, and Instagram. The company's lawyers and AI oversight group were able to review them quickly. Coleman rolled out lazy river-themed pages on its website that pointed shoppers to products it really sells, like coolers and lawn chairs. On TikTok and Instagram, the brand posted videos and images featuring the fake lazy river.
The Results:
• 150M social impressions and 3M engagements
• Direct-to-consumer sales up 105% year-over-year
• One in five direct-to-consumer dollars in June traced back to lazy river content
• 1.5M views on a behind-the-scenes video with the team about how it handled the incident
Why This Matters: The fake ad was free market research, showing real interest in a backyard lazy river. AI quickly helped Coleman steer demand to real products on its shelves.
Your Action: Build a fake content playbook. Set up social listening alerts for your brand name paired with phrases like "is this real," "fake," "AI," and "scam." Also flag sudden spikes in brand mentions, which often signal something viral before you know what it is. Agree with leadership and legal now on when you'll debunk, play along, or stay quiet, so your reaction time can be fast. Then run a drill. Ask AI to create a realistic fake ad for a product your brand doesn't sell. Treat it like the real thing, and track how long it takes your team to go from spotting it to having an approved response ready to post. If it takes more than a day, identify what’s slowing you down.
Other Notable Headlines📌
WPP opens an AI-powered production hub in London - The 34,000-square-foot space, called “Devon’s Point”, combines studios, AI production tools, and areas for working with creators. WPP plans to open similar hubs around the world.
Scope3 rebrands as Apostra, trading sustainability for AI🔒 - The company built to track the ad industry's carbon footprint has now pivoted to AI agents that buy media across Meta, Google, ChatGPT, and other platforms. Its original carbon measurement work lives on as a side business.
ChatGPT's grip on AI chatbots is loosening🔒 - ChatGPT’s share of total AI prompts dropped from 70% to 50% in the first half of 2026, according to Comscore. Gemini climbed to 30% and Claude to 11%.
Sam's Club's retail media network now predicts who will buy next - Its new AI targeting uses membership data and 400+ shopping signals to find likely first-time buyers for brands like P&G.
Microsoft insider called AI scraping "the largest theft of labor in human history," court filings show - Newly unsealed documents in The New York Times' copyright lawsuit against OpenAI and Microsoft include internal admissions that AI chatbots threaten publishers' businesses.
OpenAI is enlisting an influencer army to make it look 'good for the world'🔒 - Sponsored Instagram posts promoting ChatGPT jumped from 61 in June to 141 in August, per HypeAuditor. AI doesn’t have a great reputation. OpenAI is now recruiting parenting, fitness, and small business creators alongside tech influencers to make ChatGPT look ‘good for the world’ and ‘cool’.
That’s It For This Week 👋
The U of Digital AI Edge Newsletter is intended for AI Edge Pro subscribers, but occasional forwarding is okay!
To subscribe visit Uof.Digital/Newsletters/ai-edge or AI Edge Pro, or contact us directly for group subscriptions.
And remember, U of Digital helps teams drive better outcomes through structured education on critical topics in marketing / advertising like AI, mar tech, ad tech, CTV, commerce media, and more. Interested in learning more about how we can supercharge your team? Let’s talk!
Thanks for reading!
