Agent Beat for October 14, 2026

Welcome to Agent Beat, a daily briefing on how AI finds brands and how agents choose what to buy. I'm your host, and today is Wednesday, October 14, 2026. Let's get into the stories.

Research reported by Search Engine Journal, covering 1,650 responses, found that AI assistants recommend different brands depending on the language a question is asked in. Most AI visibility checks are run in English only. Marketers should test how their products show up in each language their buyers use, because an English-only check can hide brands that win in other languages.

That language finding sits alongside a new whitepaper from Abondance. It covers 17 generative engine optimization topics, built from 47 studies and 26 expert statements published between January and October 2026. It cites Conductor data that a page ranking first in Google is cited in an AI Overview 25 percent of the time against 10 percent for a page in position 10. The whitepaper sets that against Google's May 15 position that optimizing for AI search is still SEO. Marketers should keep ranking work as the base of AI visibility while budgeting separately for the work of getting third-party sources to describe them correctly.

On brand searches, Google now shows AI Overviews, and the sources in those summaries include pages the brand does not control, according to SEO Südwest. That gives competitors and third-party sites that write about a brand a way into the answer. Marketers should monitor what third-party pages say about their brand, because those pages can now shape the answer on their own name queries.

In citation patterns, a Semrush study covering January to July 2026 found that ChatGPT cites news media more than other AI assistants do, with the pattern growing in the UK and the US, according to Siècle Digital. The study says each assistant draws on a different mix of sources. Marketers should check citation sources assistant by assistant, because a page that gets cited by one may not be picked up by another.

Over in Bing Webmaster Tools, the platform added a Topics Portfolio view to its AI Performance report, which sorts the queries where a site is cited into themes. The view also lists queries where other sites are cited but the site is not, according to Search Engine Journal. Marketers can use it to find the topics where AI answers skip them and to decide which content to build next.

Separately, consultant Pierre Sauve measured how strongly 148 brands sit inside open models using log-probabilities and compared that with the brands ChatGPT and Gemini actually cited. He reports only 0.1 agreement on a 0 to 1 scale between the two rankings, and no measurable gain from a Wikipedia page or a heavy Reddit presence at equal web notability. He recommends audits that call the assistant interface rather than the API and that report presence and citation as two separate numbers, so teams should plan to defend two metrics instead of one visibility score.

Search Engine Roundtable also reports that Microsoft added Topics Portfolio to the AI performance reports in Bing Webmaster Tools. Bing says the view shows how a site's presence is spread across themes, how much demand each theme carries, how competitive it is, and where the site has room to grow. Marketers get a topic-level list of where AI answers cite them and where competitors are cited instead.

On the reliability of rank data, Search Engine Journal reports that Google is serving fake or altered search results to scrapers, which distorts the rank data third-party SEO tools collect. The practice follows weeks in which rank trackers could not report Search results reliably. Marketers should treat third-party rank numbers as unreliable for now and confirm important changes in Search Console or with manual checks.

For publishers of PDFs, Google's Martin Splitt and John Mueller posted a Search Off the Record episode on how PDFs are handled in Search, covering conversion for crawling and indexing, links, images, file limits, and whether PDFs can rank in search and AI search. The episode is Google's own description of that process. Marketers who publish reports, spec sheets or manuals as PDFs should treat them as content Google can index and cite, within the limits the episode describes.

In AI Mode, Google added Gemini 4 Argon for AI Pro and Ultra subscribers and made the model available to developers, enterprises and the Gemini app, according to Search Engine Journal. The change puts a newer model behind AI Mode answers for paying users. Marketers should watch whether the sources cited in AI Mode shift now that a different model writes those answers.

Turning to shopping agents, MarTech writes that consumer agents such as Meta's Muse can compare products, check policies and buy without the customer ever visiting a brand's site, which shifts who controls the customer relationship. It says brands now have to compete when agents mediate shopping, or lose customers to brands that do. Marketers should make product, price and policy data readable to agents and expect less direct site behavior to measure.

SEO Südwest adds another report on the same scraper practice. It reports that Google serves fake search results to scrapers, apparently to disrupt the collection of its content, and that SEO tools are picking up the false data. The site says this makes the data those tools report unreliable. Marketers who rely on third-party rank or SERP data should check it against their own observations before acting on it.

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