Google poisons third-party SERP trackers as poisoned SERPs surge to 8.0% of scraped results

Google's poisoned SERPs break keyword tracking at the data layer

Google appears to be returning altered results pages to the third-party crawlers that power rank-tracking software. A keyword-tools vendor's study describes a "poisoned SERP" as a normal-looking results page where the domains that usually rank for a keyword are missing, with no error or CAPTCHA to flag the problem. Data the vendor published October 12 shows that 8.0% of scraped Google SERPs were corrupted or poisoned.

The same vendor notes that Google's latest bot-mitigation defenses are deprecating older scraping approaches, removing the workarounds practitioners have leaned on. The practical effect, as framed in a video recap of the day's search forum discussion, is that keyword tracking is being broken at the data layer rather than at the reporting layer.

AI Overviews claim 40% of local results while most businesses have no playbook

A separate thread on the same day reports that Google's AI Overviews now control 40% of local search results, citing data from LocalFalcon. The framing is blunt: most businesses have no strategy for it. For local SEO practitioners, the tracker poisoning story and the AI Overview share story arrive together, which means two unrelated measurement problems land in the same reporting week. A reader running a local campaign can no longer trust either the per-keyword rank pull or the surface-level share-of-voice number without manual verification.

Gemini launch and "personal concierge" Discover reshape the AI search surface

The day's video coverage also covers the official launch of Gemini inside Google surfaces, alongside a separate framing of Google Discover as a "personal concierge." Read together, the two signals describe a search experience where the answer, the follow-up and the recommendation funnel all sit inside Google's own AI surface rather than on a publisher page. That surface shift is the reason poisoned third-party tracking carries higher stakes than it would have a year ago: if fewer users click through, the rankings being measured are also the rankings fewer humans see.

Anthropic tightens source rules as OpenAI, Google and Reddit close the same gap from the other side

Coverage from a podcast on AI search reports that Anthropic has banned fake sources for its AI search product. The same item notes that OpenAI, Google and Reddit are tightening the same gap from the other direction, "where content gets found, so a fake presence now has two referees watching for it." For SEO practitioners, the takeaway is that provenance checks are now a two-sided problem: earning citation inside the answer engine, and not being filtered out by the answer engine's own anti-fake-source layer.

What to confirm in the next reporting cycle

Three checkpoints are worth running this week. First, re-pull a known keyword set from both the affected tracker and a manual browser session to confirm whether the 8.0% figure published on October 12 still matches what is observed on October 17, since bot-mitigation defenses are being deprecated in parallel. Second, audit local landing pages against the AI Overview surface, since the 40% share figure is local-specific and may not generalize to national campaigns.

Third, review AI-bot crawler directives, since AI answer engines, scrapers and Google's own defenses are all moving at once; a quick pass with an AI Bot Robots.txt Checker and a Serp Snippet Preview refresh will catch the pages most likely to be misrepresented inside the new surface.