TL;DR
- The search data most SEO reports rely on has been distorted by bots, AI agents and a Google logging bug that ran for almost a year.
- Google is making scraping harder and more expensive, from mandatory JavaScript in January 2025 to removing the num=100 parameter in September 2025 and suing SerpApi.
- Third-party rank trackers now see less of the results page, at higher cost, and some AI tools that ground answers in scraped Google results face the same squeeze.
- At Storylake, we now report on clicks, conversions and AI answer visibility first, and treat impressions and tracked rankings as supporting signals.
- If your reporting still leads with impressions or average position, your baseline may be wrong.
Why your SEO dashboard may be misleading you
For more than a decade, SEO reporting rested on two numbers: how often you appeared in Google and where you ranked. Both have become unreliable over the past 18 months. Some of the distortion comes from machines searching Google at scale, some from Google’s own reporting errors, and some from Google deliberately cutting off the tools that collect ranking data.
The risk is practical. Marketing teams can end up making budget decisions from impression trends that were partly inflated, or from rank trackers that now see less of the page than they used to. Here is what changed, and how we adapt AI search reporting at Storylake.
What SERP data contamination means
SERP data contamination is any non-human or erroneous activity that ends up counted as search visibility. This has been standing out in three different ways:
Rank tracker bots
or years, trackers loaded 100 results per query with Google’s num=100 parameter. Each load could register an impression for every site on the page, including positions far below anything a person would scroll to. When Google stopped honouring the parameter around 10 September 2025, Search Console impressions dropped and average position rose for many sites, even though their real traffic stayed the same.
A logging bug
Google later confirmed that Search Console over-reported impressions from 13 May 2025 to 27 April 2026, almost a full year, while clicks were unaffected. Any year-on-year comparison that crosses that window needs a clear annotation.
AI agents searching on users’ behalf
In late 2025, site owners found full ChatGPT prompts appearing as queries in Search Console, which OpenAI described as a routing glitch. Long, conversational queries with impressions and no clicks are now a familiar pattern, and Google includes AI Overviews and AI Mode activity in the same Web performance report.
How Google is shutting out scrapers
Google has never allowed automated scraping of its results, but enforcement has tightened sharply since early 2025:
| Date | Change | Effect on SEO data |
|---|---|---|
| 9 Nov 2026 | Court dismisses Google’s amended complaint against SerpApi, with leave to refile | Legal route to stop scrapers still unresolved |
| 20 Jul 2026 | First dismissal of Google’s DMCA claims against SerpApi | Google told to show it acts for copyright owners |
| Dec 2025 | Google sues SerpApi, alleging it bypassed SearchGuard | Scraping becomes a legal risk as well as a technical one |
| 10 Sep 2025 | num=100 parameter stops working | Tracking 100 positions takes up to ten requests instead of one |
| Jan 2025 | JavaScript required to load results, with SearchGuard bot protection | Scrapers need full browsers, raising cost and slowing collection |
The lawsuit has not gone Google’s way so far. The judge found that Google had not shown the content SerpApi scraped was copyrighted, or that its owners authorised SearchGuard, and gave Google until the end of November to try again. The ruling does not say scraping is lawful, though. The technical barriers remain, and each one makes third-party ranking data slower, costlier and patchier.
There is a knock-on effect for AI search. Research from Peec AI found ChatGPT drawing on scraping providers such as Bright Data and Oxylabs for Google results, while OpenAI builds its own index in parallel. As Google’s results become harder to collect, AI platforms have a stronger incentive to rely on their own crawlers, which is why we covered ChatGPT’s Labrador index separately.
What changes for clients working with Storylake
Clean data comes before strategy. When we start an engagement, the first job in the Assess stage of our AI search approach is rebuilding the baseline so that every later decision rests on numbers that mean what they say.
| Area | What most SEO reports showed | What we report now |
|---|---|---|
| Headline metric | Impressions and average position | Clicks, conversions and pipeline from organic and AI referrals |
| Rank tracking | 100 positions for hundreds of keywords | Top positions for a focused set of commercial keywords, including AI Overviews |
| Year-on-year trends | Raw Search Console comparisons | Comparisons annotated for 10 September 2025 and the May 2025 to April 2026 bug window |
| AI search | Rarely measured | Mentions and citations for real buyer prompts across ChatGPT, Gemini and Perplexity |
| Query research | Search Console queries taken at face value | Long, prompt-style queries flagged and tested in AI assistants |
Two further habits help. Server log analysis separates crawler and agent activity from human visits, so we can see which AI bots reach a site and how often. We also connect analytics and CRM data in one attribution model, so AI-assisted visits are credited even when the first touch never shows up as a ranking.
This approach accepts that some numbers will look smaller. A client whose impressions fall after the bug correction has not lost visibility, and we say so plainly in reporting so that a corrected metric does not trigger a rushed change of strategy.
Conclusion
Google’s fight with scrapers and the noise from bots and AI agents will not settle soon, and the SerpApi case may yet change the rules again. Brands that keep measuring success through impressions and tracked positions risk optimising for a distorted picture. Those that anchor reporting in clicks, conversions and AI answer visibility will make better decisions whatever Google does next. If you would like us to check how clean your current baseline is, talk to our AI search team.
