Field Notes · Search

Google showed your rank tracker a different web.

Nozzle says 8.7% of the page-one results Google served to its collectors were missing most of the sites that normally rank. The baseline was 0.17%.

8.7%of page-one results served to Nozzle's collectors were missing most of the usual sites
Two hand mirrors side by side showing different reflections of the same object
Same query. Two answers. One of them is in your report. Art direction: Fiction Tribe, produced with AI.
· 4 min read

The short version

  • Nozzle says 8.7% of the page-one results Google served to its collectors since 13 September were missing more than half the sites that normally rank. The baseline was 0.17%.
  • Nozzle is a rank tracker with a commercial stake. AccuRanker measured 1.4% on a day Nozzle measured roughly 27%. Neither is audited.
  • Meanwhile 49% of marketing leaders in a Typeface survey plan to deprioritise SEO, on a sample with no published method.
  • Stop reporting one tracker's number as fact. Report a range, and name the method.

Your rank report is a photograph of the web taken by a robot. This month a rank tracker published evidence that Google has been handing the robot a different web.

Nozzle published a methodology and a dataset on 10 October and updated both on 21 October. Across 56.1 million judged search result pages in 53 countries, 8.7% of the page-one results served to its collectors since 13 September were missing more than half the sites that normally rank there. The baseline was 0.17%.

Is Google showing rank trackers different results?

Something changed, and nobody outside Google can say by how much. On Nozzle's wider measure, covering page one or two, the rate is 22.6%. It peaked at 43.5% on 30 September and was back up to 35.7% on 21 October. Ads survived the swap more reliably than organic results. Semrush confirmed it has seen inaccurate results returned to automated collectors since late summer. Google has said nothing.

8.7%
of page-one results served to Nozzle's collectors were missing most of the sites that normally rank. Baseline: 0.17%. Nozzle Research, self-published, October 2026

How far should you trust the Nozzle figures?

As a direction, yes. As a number, no. The figures are self-published by a rank tracker with commercial exposure to the problem. AccuRanker, running a different check, measured 1.4% on a day Nozzle measured roughly 27%. The two are not comparable and neither has been audited. Blaming Google is the author's inference from the timing of blocked and then swapped requests. Nobody has confirmed a cause.

What survives is uncomfortable enough. The index behaves differently for machines. Every rank report, share-of-voice chart and GEO dashboard reads the web through a collector. We sell GEO analysis, so that includes ours.

The question stops being where you rank. It becomes which version of the index you were shown.

Should you move budget from SEO to AI search?

Not on the evidence available this month. The Typeface Signal Report, published 14 October and covered by PPC Land, surveyed 500-plus US and UK marketing leaders. Forty-nine percent plan to deprioritise or stop investing in traditional SEO. Sixty-eight percent say they create content experiences for AI that humans will likely never see.

State the limits. Typeface sells AI content orchestration software, so the report measures appetite for its own product. There is no exact sample count, no fielding dates, no panel provider, no weighting and no margin of error. The vendor's pitch email says 63% where the PDF says 62% and 61%. Treat the direction as real and every decimal as marketing.

So a channel you can measure, imperfectly, is being dropped for one you cannot measure at all.

What is AI search actually doing to clicks?

It removes them, according to the one causal study on the table. Digital Content Next gathered the evidence on 5 October. The strongest item is a field experiment by Saharsh Agarwal and Ananya Sen that randomly assigned more than 1,000 Google users to search with or without AI Overviews. When an Overview appeared, outbound organic clicks fell 39.8%. Sponsored clicks did not move. Perceived search quality did not measurably improve.

39.8%
fewer outbound organic clicks when an AI Overview appeared. One randomised experiment, 1,000-plus users. Agarwal and Sen, cited by Digital Content Next

Limits again. DCN is a publisher trade body and the conclusion serves its members. The paper was posted on 3 April, so the October news is the round-up, not the finding. Secondary coverage quotes the effect as both 38% and 39.8%. And keep the scale in mind. In the Reuters Institute analysis of Chartbeat data that DCN cites, Google Search is 7.3% of publisher referrals. ChatGPT is 0.02%.

What should you do about search reporting?

Stop treating one tracker as the record. Run two with different methods. Report the range, and print each method next to its number.

Check a sample by hand. A short list of priority queries, a real browser, a real person, on a schedule. It is dull, and it is the one reading no collector can distort.

Put a kill date on AI visibility spend. State the hypothesis. State what would count as failure. Stop if it arrives.

A rank used to be a fact. Now it is one witness's account. Get a second witness before you move the budget.

Questions people ask

Is Google showing rank trackers different search results?

Nozzle reports that 8.7% of the page-one results served to its collectors since 13 September 2026 were missing more than half the sites that normally rank, against a 0.17% baseline. The figures are self-published and unaudited, and blaming Google is the author's inference from timing.

Can I still trust rank tracking data?

As a range, not as a single number. AccuRanker measured 1.4% on a day Nozzle measured roughly 27% using a different check, and Semrush confirmed inaccurate results returned to automated collectors since late summer. Report more than one tracker and name each method.

Do AI Overviews reduce clicks to websites?

In one randomised experiment with more than 1,000 Google users, outbound organic clicks fell 39.8% when an AI Overview appeared, with no effect on sponsored clicks. Google Search still supplies 7.3% of publisher referrals against 0.02% for ChatGPT.

Sources

  1. Poisoned Google SERPs, Nozzle Research, . Self-published by a rank tracker with commercial exposure to the problem. Updated 21 October. Not audited.
  2. Google feeds rank trackers swapped results on 8% of pages, Nozzle finds, PPC Land. Coverage of the same dataset.
  3. 49% of marketers plan to deprioritize SEO, Typeface survey finds, PPC Land, . Typeface sells AI content orchestration software. No exact sample count, fielding dates, panel provider, weighting or margin of error.
  4. The evidence is clear: AI search is costing publishers, Digital Content Next, . A publisher trade body reaching a conclusion that serves its members.
  5. Field experiment on AI Overviews and outbound clicks, Saharsh Agarwal and Ananya Sen, via SSRN. Posted 3 April, last revised 8 July. Secondary coverage reports the effect as both 38% and 39.8%.

See what AI tells your buyers before they call.

GEO Analysis captures the answers word for word, across five AI engines, in one week. Run one question free.

Run a question free

Keep reading

Buyers

The decision was made before you arrived.

6sense asked 4,611 B2B buyers how they choose. The favourite going in won 82% of the time, and AI use inside the evaluation went down.

· 4 min read
82%of deals go to the vendor favoured before first contact
Proof

You will lose the deal you cannot prove.

Forrester predicts two-thirds of B2B providers will lose deals because buyers cannot verify their claims. It discloses no method, which rather makes the point.

· 4 min read
2 in 3providers will lose deals because buyers cannot verify their claims, Forrester predicts
Every note