Your weekly guide on everything AI marketing. Covering everything from AI marketing news, tips, deep dives, events, podcasts, jobs and much more. Never miss a beat with our 5-minute newsletter.
In today’s email:
Snap loses the right to mine chatbot prompts for ads in Germany
ChatGPT sees its share of AI prompts fall from 70% to 50%
Scope3 becomes Apostra and bets the business on agents that buy media
Plus: monday.com on why AI is not a condiment, Samsara’s CMO on 113 agents in production, testing a post on 10,000 AI readers with Jevtown and why Amazon shut out Meta’s shopping agent.
The Top 3 Stories
1. A German court stops Snap mining chatbot prompts for ads
The Regional Court of Cologne has barred Snap from using what people type into its My AI chatbot to choose ads without a legal basis. The same ruling outlawed pre-ticked ad consent boxes. Snap now faces penalties of up to €250,000 per breach.
Why it matters: If your brand runs its own chatbot, what customers type into it can’t feed your ad targeting in Europe without explicit consent. 📰 PPC Land
2. ChatGPT’s grip on AI discovery is slipping fast
Comscore’s latest AI Intelligence Report shows ChatGPT’s share of prompt volume fell from 70% to 50% between January and June 2026 as rival assistants gained ground. Being used is not the same as being credited. Tripadvisor supplies 61% of travel sourcing for AI platforms but receives just 21% of the citations.
Why it matters: Optimising for one assistant now covers half the audience at best, so track citations per assistant, not just whether you were read. 📰 Digiday
3. Scope3 renames itself Apostra and bets on agents that buy media
Scope3, which made its name measuring the carbon cost of ads, has rebranded as Apostra and made agentic buying its core business. It is already running automated buys in the US, UK, Brazil and France. The total so far is around $1.2 million over three months, for some “very large” brands.
Why it matters: Agents buying media for you means fewer ad tech middlemen taking a cut, but at $1.2 million so far this is still a pilot, so test it with a small budget first. 📰 Adweek
More trending AI marketing news from last week
YouTube is turning Ask YouTube into a product-comparison shopping assistant.
SQREEM pitches a behavioural model that predicts what people do, not say.
93% of GTM teams run AI agents, yet nearly a third can’t count them.
A third of shoppers regret a purchase an AI recommended to them.
A 10% AI boost to ROAS could add £80bn in UK sales over ten years.
Together With PeerPush
Launch your product on PeerPush. People find it, AI assistants discover it.
Most launch sites give you one big day, then the feed resets. PeerPush keeps your product discoverable long after launch day, by people and by the AI assistants they now ask for recommendations.
Paste your URL and its AI builds a structured product page in minutes, with use cases, audiences, pricing and platforms that both humans and machines can read. In the last 30 days, PeerPush pages were read 10.8 million times by 33 AI systems, including ChatGPT, Claude, Perplexity and Gemini.
Over a million monthly visitors, a weekly newsletter and a community of 48,000+ builders do the rest. Listing is free.
Quote of the week
Stop treating AI like peanut butter. [Marketers] are spreading it on top of their work. They’re treating it as a condiment… not treating as a key ingredient. You need to bake it in.
Trending AI tools for marketers
Here are 4 trending tools to explore this week. Quick, practical upgrades to your AI marketing arsenal.
Jevtown lets 10,000 AI personas react to your post before you publish it.
Clueso MCP turns a chat with your AI assistant into an editable, on-brand product video.
AI Creative Insights by Decode predicts attention and emotion on an ad before media goes live.
thestory.run coaches employees to write their own LinkedIn posts, never ghostwriting them.
TV, podcasts & streaming
Finite Podcast, “The CMO with 300 AI agents”
Jodi Norris talks to Meagen Eisenberg, Samsara’s CMO, whose team runs 113 AI agents in production marketing workflows. Her argument is that agentic marketing changes how the work gets done rather than adding a channel. For marketing leaders that means hiring differently, watch it here.
AI training
How to pressure-test a post before you publish using Jevtown
The overview:
Jevtown shows your draft to 10,000 AI personas, and this workflow tells you who it lands with, who it annoys and what to fix before a real audience sees it.
Step-by-step:
Paste your draft
Post a headline, social post, listing or product description, no sign-in needed. English copy is read by 10,000 English-speaking personas within seconds.Watch how far it travels
The text starts with the 600 people it suits best and only spreads while they are glad to see it. The one-sentence result tells you which wave stopped it.Find who stopped and who got annoyed
Break the reactions down by interest, job, age, budget and city. Then read the voices of the crowd to hear how individual personas responded.Edit and run it again
Use Versions to rewrite the text and show it to the crowd again. Both versions sit side by side, so you can see exactly what changed.Test your price if you are selling
For a product, set a ladder of prices and see how many would buy at each. Jevtown also shows which price earns the most.
Pro tip: Move in a resident modelled on your ideal customer and check how often Jevtown guesses their reactions correctly before you trust it with bigger decisions.
The weekly deep dive
Who owns the customer when an agent does the buying?
Amazon has blocked Meta’s new Muse agent from shopping on its site, and the fight is really about who keeps the customer, and the ad money, once software does the browsing.
The block
Amazon’s Conditions of Use require agents to identify themselves in every request, and Muse does not. Shoppers who try it on Amazon now see a warning that unauthorised agent access breaks those terms. Amazon’s own Buy for Me agent does similar work, but it identifies itself and lets brands opt out.
The money
Mark Zuckerberg says Muse will take a very small cut of each transaction, potentially paid by the businesses people buy from. Amazon’s advertising business made more than $68 billion last year, and it depends on people browsing its pages. An agent that goes straight to checkout never sees a sponsored listing.
The legal route
Amazon lost its anti-hacking argument against Perplexity’s shopping agent in August, when the Ninth Circuit ruled that the user, not the AI company, was accessing its computers. The ruling left contract claims open. This week Amazon added one against Perplexity, and its Muse block leans on the same Conditions of Use.
Our takeaway: If agents do the buying, sponsored listings lose their audience, so work out how much of your marketplace revenue still depends on a human scrolling.
📖 Read the full article at GeekWire
