The "Google is tagging AI Overviews links" story was a browser extension, not Google.

On 2 October Search Engine Roundtable reported that Tom Pool had seen `?st_source=ai_overview` added to links inside Google AI Overviews and AI Mode. Barry Schwartz wrote at the time that he could not reproduce it. The article now carries two updates. The first reads: "This was a browser extension - not Google." The second quotes Tom Pool: "Turns out it was a Chrome Browser Extension modifying the URLs to include these parameters. Presumably to sell the data on to brokers..." The headline still says Google is testing it, so anyone who reads only the headline will come away with the wrong idea. Google never made a statement, because there was nothing to state.

This was a watchlist line on 1 October, resting on a trade report. The source itself is now read. DemandSphere tracks a set of brand-name keywords once a day and counts whether a Google AI Overview appears. Its figures run from 26.12% on 1 September to 90.48% on 27 September. Search Engine Land's headline puts the move at "26% to 80%", and secondary reports give 80.23% for 28 September and 82.06% for 29 September (those two figures were not read first-hand). The limits are large. DemandSphere does not say how many keywords it tracks or how big the brands are. It counts an overview whenever one shows, whether or not the overview is about the brand. It covers "all markets, all devices" with no breakdown yet. And the jump on 26 and 27 September sits inside Google's spam update, which started on 24 September, so it may not hold.

Google's incident for the September 2026 spam update still carries only the 24 September launch post, on day eleven of a rollout Google said "may take up to two weeks". Perplexity's migration page is unchanged and names no model for re-routed Sonar requests, and its changelog has no October entry. Anthropic's table still shows `claude-sonnet-4-5-20250929` deprecated, with retirement on 30 November 2026 and `claude-sonnet-5-5` as the replacement.

A YouGov online survey of 1,000 Americans, run on 15 September 2026. Verbatim: "64% of Americans we surveyed say they never or rarely get their news from AI." "Over two-thirds (68%) of Americans who have used AI report it hallucinating information." "Just 5% of Americans reported trusting AI-generated news summaries more or somewhat more than traditional news outlets." On links: "Just 10% of Americans always click on news links", while 76% click at least sometimes. BuzzStream sells a digital PR tool, and the survey is about news, not about choosing a service.