Agent Beat for October 24, 2026

This is Agent Beat. It's Saturday, October 24, 2026. Today we trace how AI finds brands and how agents choose what to buy, through the day's reporting on shopping agents, search traffic, and the systems that decide what gets cited and what gets bought.

Walmart is drawing the boundaries for agents that shop on its sites. The retailer rolled out three tools across Walmart and Sam's Club. Agent Profiles let shoppers authorize an agent, set spending limits and track purchases. An Agent Gateway gives agents a more direct connection to Walmart's shopping services. New purchasing rules limit what agents can do and what customer information they can reach. Walmart announced the changes on October 21 and said the goal is to make its experience the most trusted for shoppers who use agents. The company also said it is helping marketplace sellers prepare for AI-driven product discovery.

Trust is also a matter of traffic, and the traffic from ChatGPT remains small. Marie Haynes analyzed 12 months of GA4 data across 100 websites, covering more than 145 million sessions. ChatGPT sent 0.62% as many sessions as Google organic search. Adding referrals from other AI assistants raised that to 0.73%. Among 55 sites tracking conversions, 14% of ChatGPT visitors converted against 3.0% of Google organic visitors. But Google still drove 27 times as many conversions as all the AI tools together. Haynes, who has called GEO a distraction, also found that ChatGPT referral growth stalled across the sites.

Google is trying to make links more visible inside its AI answers. On October 15, Google said it is showing more links in AI Mode and AI Overviews, placing them higher and highlighting one or two for some queries, according to Abondance's account of the announcement. The same piece reports, citing Le Figaro, that Chartbeat data put the fall in news sites' Google traffic at 40% between July 2025 and July 2026. Google still does not report clicks from its AI answers, the article says. That leaves publishers and SEOs unable to measure what the new link formats deliver.

The citation cards around AI Overviews may be changing too. Search Engine Roundtable reports that Google appears to be testing or rolling out AI-written summaries inside the citation cards that sit beside AI Overviews. The cards, shown at the right or bottom of an overview, can now carry an AI summary instead of a plain description.

On the advertising side, OpenAI has described how its ChatGPT Ads system works. Search Engine Journal's write-up says OpenAI's account describes a custom large language model orchestrator that reads multi-turn conversations, nominates ads for the auction and drops placements that miss the deadline for the response. The piece sets out how relevance, bids and what the conversation remembers interact when an ad is chosen and where it appears.

Semrush says Google's information agents and Universal Cart stand to change how people discover and buy products online. The post walks through what the two features do for discovery and for checkout.

In France, the picture for publishers is getting harder. French media sites lost 23% of visits in a year as Google's AI Overviews spread, according to the article. The piece reports that Google's preferred sources feature turns brand recognition into a search lever of its own. It argues that being picked as a preferred source depends on users already knowing the brand, which changes how publishers and brands plan search work.

Duane Forrester makes a related argument about indexing. In a piece published by Search Engine Journal, Forrester argues that being indexed no longer buys a visit, because in the answer layer it only makes a page eligible to be cited. The distinction is between being in the index and being cited in an answer.

Semrush sets out an AI search strategy in four layers. It starts with a diagnosis of whether AI systems can reach a site and whether they describe the brand correctly. The post says rewriting content will not help if those first problems are left in place. The order matters: the diagnosis comes before content changes.

An agency writer ran a test of its own citations. The writer put the questions its prospects type into Google to several AI assistants in late September, then measured how ChatGPT, Claude and Perplexity answered them. The post sets out what the test showed about the agency's own citations in the three assistants.

A wider look at small-business sites adds a plain finding. An analysis of 474 small and mid-sized company websites around Metz, with 118 answers recorded from ChatGPT and Perplexity, found that AI assistants can reach the sites but often cannot tell who the company is, according to the article. It says small businesses can earn a place in AI answers if their sites state plainly who they are and what they do.

Thanks for listening. For the full written brief, go to agent-beat.com.