AI search news · ·
Why does your rank tracker not match what you see in Google? Four vendors' own statements: Nozzle 8.0% of SERPs poisoned since September 13 (27% on September 29), Morningscore 60% of searches truncated on September 30, Sistrix top-10 still reliable, AWR a simplified page for suspected bots
Between 2026-09-18 and 2026-10-10 four rank-tracking vendors published their own accounts of Google serving automated requests a results page that differs from what a person sees. Nozzle, data through 2026-10-09: "Since September 13, 8.0% of SERPs were poisoned, against 0.17% in the baseline", peaking at 27.0% on September 29 across 27.5M SERPs. Morningscore: from 22:10 UTC on September 24 for almost six days, about 60% of searches on September 30 and October 1 returned fewer than 40 results, with 8 of the previous top 10 missing on average. Sistrix, 2026-10-07: divergent results since September, "top 10 rankings and the Visibility Index remain reliable". Advanced Web Ranking, 2026-09-18: a suspected bot "can be served a simplified page". None quotes Google. We read each document directly.
What the four vendors published, with their own dates and numbers
On 2026-10-12 Barry Schwartz at Search Engine Roundtable wrote that third-party rank trackers have been returning unreliable data for weeks and collected the vendor pointers; the article has no Google statement. We covered the first round on 2026-09-19. Since then four vendors have published documents with numbers; we read each on 2026-10-12.
| Vendor, document, date | What Google served them, in their words | Their numbers | What they changed |
|---|---|---|---|
| Nozzle Research, "Poisoned Google SERPs", 2026-10-10, data through 2026-10-09 | "Since September 2026, Google has been poisoning the SERPs it serves rank trackers": a full-looking page with the keyword's usual sites swapped out. "Requests to Google started failing on September 13; from September 20, the ones that got through came back poisoned." | 8.0% of judged SERPs poisoned since September 13 against 0.17% in the August 20 to September 9 baseline; daily 0 to 27%, peak 27.0% on September 29; poisoned pages keep 20.7% of the previous SERP's sites against 88.5% before; a project's own site misses the top 10 on 58.6% of poisoned SERPs against 2.7%; 77.9% have their usual sites back within three days; 27.5M SERPs, 660K+ keywords, 49 countries | Detection by each keyword's history: a SERP is poisoned when fewer than half of the keyword's usual top-10 sites appear on any page and at least 80% of them return within 14 days; the last three days are still filling in |
| Morningscore, "Google gave rank trackers truncated results for a week", 2026-10-08 | From 22:10 UTC on September 24 for almost six days, searches came back with fewer than 40 results instead of about 50; "8 of the previous top 10 sites were missing on average, sometimes including the #1". Manual Chrome searches at the same time returned complete results. | About 60% of searches truncated on September 30 and October 1 (US 40.4%, UK 45.1%, Denmark 61.2%) against 1.2 to 1.5% before September 24; two of five data providers never saw it at scale, on October 2 one provider returned truncated results for up to 72% of searches and another for 1 to 3%; 74% of missing sites back in the top 20 within three days; as of October 4, 1 to 2% of searches truncated | Truncated results discarded and other providers tried; if none returns full results the previous ranking is kept; rankings recorded from truncated results between September 24 and 30 corrected; a second-search confirmation before a steadily ranked site is marked as gone, in test mode |
| SISTRIX, "Why Google ranking data collection is changing", Johannes Beus, 2026-10-07 | "Since September 2026, we have observed that for some queries, Google returns results that differ significantly" from the normal results; on-demand updates switched off since September 2026; clicks routed through an internal /goto URL since June 2026; a maximum of 10 results per query since September 2025 | "The key point first: top 10 rankings and the Visibility Index remain reliable." Top-100 keyword counts can fall without a real visibility loss because result lists often end before position 100 | Extra verification checks filter divergent results, so some keywords refresh less often than the project settings say; the displayed last-update date is the thing to read |
| Advanced Web Ranking, "Why your tracked rankings no longer match what you see in Google", Iulia Lacriceanu, 2026-09-18 | "Google no longer serves one SERP": a suspected bot "can be served a simplified page - one where commercial and content-dense sites sit lower or drop out entirely"; country-wide queries get a country-neutral page; google.com/goto redirects on result links from late August 2026 | "any tracker, AWR included, can be served a page that differs from what your customers see"; precise-location engines land close to localized manual results while the country-wide engine drifts; no percentages published | Collection IP pool expanded, region-level tracking behind country engines, retries |
Where the four agree and where they differ
All four say the page Google serves an automated request can differ from the page a person sees, that it began in September 2026, and that it clears within days. They name it differently (poisoned, truncated, divergent, simplified) and date it differently: Nozzle puts failures from September 13 and poisoning from September 20, Morningscore's truncation began at 22:10 UTC on September 24, AWR wrote on September 18. The /goto click routing is dated June 2026 by Sistrix and late August 2026 by AWR. Nozzle states a cause, "It is Google's response to AI scraping, and rank trackers are caught in it"; that is the vendor's reading, Google has not said so. Two of Morningscore's five data providers never saw truncation at scale, and one of Nozzle's routes to Google returned 1 to 7% poisoned pages while another returned 9 to 30%, so the share a tool saw depends on how it collects.
One first-party number of our own belongs beside this. Search Engine Roundtable reported on 2026-10-11 that the Search Console Performance report was more than 64 hours behind. Our own property's table ended at 2026-10-09 when read at 16:40 UTC on 2026-10-11; at 16:43 UTC today it ran through 2026-10-11 in fresh data and 2026-10-10 in final data. The instrument the vendors point to as the check on their numbers was itself late for part of the same window, and caught up within a day on this property.
Why it matters if you watch AI visibility
- A ranking drop recorded between September 24 and October 1 may be a tracker artifact: Morningscore says 8 of the previous top 10 were missing on average from a truncated page, Nozzle that a project's own site missed the top 10 on 58.6% of poisoned pages. Read the same dates in Search Console, which is not collected by scraping, before acting on it.
- A tool that measures AI Overviews or AI Mode presence by collecting Google result pages collects under the same conditions. Nozzle finds social sites and forums fill the gap on poisoned pages, so a sources-cited read taken from a collected page can name sites a person would not have seen. Ask the vendor how it detects a divergent page; the four documents describe four different checks.
- Tools that ask ChatGPT, Perplexity, Gemini or Claude through the engines' own APIs do not collect Google result pages. AskedAbout's checks work that way, as our comparison of AI visibility tools states. Those reads have their own limit: an API sample is a sample of what the API returned that day, not proof of what every chat user sees.
How this was verified
We read the Nozzle, Morningscore, Sistrix and Advanced Web Ranking documents directly on 2026-10-12; every number in the table is the vendor's, with the vendor's date. The Search Engine Roundtable article of 2026-10-12 is the pointer; it also names SE Ranking and DataForSEO status notices, which render only by script in our reader and could not be read, so neither is cited. We run no rank tracker; the one first-party number here is the last date in our own Search Console table. We have not contacted any vendor.
Is Google blocking rank trackers in October 2026?
Google has not said so. Nozzle reports requests failing from September 13 and 8.0% of the SERPs that got through since then poisoned; Morningscore reports about 60% of searches truncated on September 30 and October 1, down to 1 to 2% by October 4; Sistrix says its top-10 rankings remain reliable.
Should I trust my rank tracker's late-September numbers?
Check them against Search Console for the same dates. A drop that reversed within days in the tracker and never appeared in Search Console is the pattern: Nozzle says 77.9% of poisoned SERPs had their usual sites back within three days.
Does this affect AI visibility tools?
It affects tools that measure Google's AI Overviews or rankings by collecting result pages. It does not affect tools that query ChatGPT, Claude, Gemini or Perplexity through their APIs.
See your number
See which businesses AI names when your client's buyers ask.
Running this for clients? The $249 agency 5-pack audits five businesses, white-labeled.
Who runs this
- Built and operated by Sensara LLC, Atlanta, Georgia — about us and how the audit works.
- See what the report looks like before you run anything — score per engine, the competitors AI names instead of you, and a fix plan.
- We run the same audit on ourselves every week and publish the result: in the latest run AI named AskedAbout in 1 of 144 answers. We report our own numbers the way we report yours.