A vendor changed only the "last updated" date on ten pages and AI citations rose 31.5%, and Google now spells out that strangers' edits to a business listing can go live after four days.

The help page now says, verbatim: "Once notified, you have 4 days to accept or reject the suggestion." And: "If you don't respond within 4 days, Google may automatically publish the update if it is supported by other publicly available information, such as your business website." It also says Google "may apply suggested user edits to your business information without prior review" in some cases. The page carries no revision date, so we know when it was reported, not when it changed.

Spotted by Lluc Penycate on X: for hotel searches, the AI Overview moves from the top of the page to the bottom-right, under the hotel's knowledge panel. The article links it to the 17 September test of an AI Overview inside a local business panel. No Google statement. A test, not a rollout.

Checked against the live repository with the token this morning: ten remote branches, none named for perplexity, sonar or agent; `main` still at `20b1172` dated 22 August; `src/adapters.ts` line 14 still `perplexity: "sonar"`, line 91 still `/chat/completions`. The Perplexity changelog adds Gemini 3.8 Flash promotional pricing and nothing new on Sonar.

Otterly took ten of its own older articles and, on 4 August 2026, changed only the visible "Last updated" date, leaving the text alone. Two unchanged articles served as controls. Tracked across 320 prompts on seven engines. Verbatim: "The 10 updated articles received 4,800 citations in the 30 days before the experiment and 6,312 in the 30 days after it. That is a 31.5% increase." The controls went "from 358 to 320 citations, a 10.6% decline." By engine the results split sharply: ChatGPT 37 to 293, Claude 1,135 to 2,118, Perplexity up 217%, Google AI Overviews up only 15%, while Google AI Mode fell 20% and Gemini fell 37%. Otterly's own caveat: "Citation trends were already moving before the date change, so we interpret the findings as directional rather than conclusive." Limits: one site, a software vendor's blog, two control pages, citations (links) counted rather than a brand being named, and no statement on whether the page's structured data changed with the visible date.

50 e-commerce queries, 1,950 test cases. Verbatim: on one open ranking model, "unsupported-rich variants show significant normalized rank gain over clean candidates (+0.065 to +0.092 across claim profiles, Holm-corrected)", i.e. product text stuffed with detail that no evidence backs ranked higher. The effect "is model-dependent (marginal on MiMo-v2.5, absent on GLM-5.3-Flash)". No commercial assistant was tested.

US PC opt-in panel, March to June 2026, counted by conversation. Verbatim: "Sponsored ads appeared in 47% of the ChatGPT conversations about video games and consoles"; toys and hobbies rose "from 5% to 24%". Sample size not given, and the denominator may include ad-free paid accounts. No service or professional category was reported.