Agent Beat for September 22, 2026

Welcome to Agent Beat, a daily podcast briefing on how AI finds brands and how agents choose what to buy. Today is Tuesday, September 22, 2026.

We start with what chatbots tell shoppers about products. A Semrush survey of 2,338 U.S. adults says 58% of AI users have bought something a chatbot recommended. About the same share have decided against a purchase based on what a chatbot told them. Nearly 81% of people who shop with chatbots said they had dropped a purchase after AI guidance. More than 85% of AI users said negative or mixed reviews surfaced by a chatbot would make them less likely to buy. Marketers should watch what assistants say about their reviews and weak points. The same answer that introduces a brand can also talk a buyer out of it.

That influence over a purchase leads to a fight over who controls the shopping conversation. Amazon blocked Meta's Muse personal agent from browsing its marketplace and buying for users without identifying itself. Amazon cited account security and terms of use, according to FrenchWeb. The dispute is about who owns the customer when an AI agent shops. Marketers selling through agents should expect platforms to control agent access and identity. They should plan for agent authentication and platform rules.

The payment side of agentic commerce is also unproven. A column reviews the agentic commerce documentation published by Visa, Mastercard, PayPal and Stripe. It says none of the four names a live merchant, while Google names three. The column argues agent checkout is still unproven rather than a repeat of voice shopping. Marketers should treat payment vendors' agentic commerce claims as unverified until a merchant is named. They should keep product data agent-readable in the meantime.

The same readiness problem shows up in business buying. Gartner expects 90% of B2B purchases to run through AI agents by 2028. At the same time, 74% of procurement leaders still do not have data ready for them, the article reports. It names Google, Tech Mahindra and Uniphore among the vendors moving fastest into this market. Marketers who sell to businesses should make product, pricing and supplier data machine-readable now. Buying agents cannot shortlist what they cannot parse.

For marketers trying to be chosen, publishing is changing. The article describes a shift from SEO to Generative Engine Optimization, where marketers aim to publish in the sources AI tools cite. It reviews tools for producing GEO content. Marketers should treat getting cited in AI answers as a publishing goal. They should pick tools that help them land in the sources those answers use.

Finding those sources takes work. Trendos' Gintare Rimolaityte explains how to map the sources that AI answers cite in a given industry, engine by engine, and how to spot gaps competitors leave. The method turns that audit into content and outreach priorities. Marketers can use the same mapping to decide which third-party pages to pitch or write for when they want a brand named in AI answers.

Platform access can still be blocked. Amazon has blocked Meta's Muse agent from using Amazon.com, TechCrunch reports. TechCrunch notes that Amazon runs its own foundation models and inference platform and is under no legal obligation to give Muse access. Marketers selling through shopping agents should expect access to be granted platform by platform, not once and for all.

Even when access is allowed, agents need to find a site's resources. Lighthouse 13.5 adds an audit for AI agent resource discovery, and its checks differ from the latest ARD proposal. The audit runs inside a tool many teams already use. Marketers should run it on product and content templates to see whether agents can find the resources a site exposes. They should expect the checks to lag the current proposal.

In Europe, the search results themselves are changing. Google has removed free product listings and their carousels from search results across the European Economic Area. Google Ads spokesperson Ginny Marvin confirmed the removal follows the European Commission's decision. Data from Productrise and Semrush cited by Abondance shows 'popular products' carousels down 90% to 100% in Germany, France, Belgium, Sweden and the Netherlands. Comparison shopping services are taking the space. Merchants selling into those markets should stop counting on free listings for visibility. They should treat comparison shopping services as the route that remains.

That is today's briefing. Thank you for listening to Agent Beat. For the full written brief, go to agent-beat.com.