Do AI Visibility Tracking Tools Create Fake Queries in Google Search Console?

Can you see LLM queries in Google Search Console like traditional search queries?

Partly, depending on the assistant:

• Google's AI Mode and AI Overviews: yes. They are Google Search, and their queries appear in the Performance report like other web search queries. Google's documentation says AI Mode clicks count as clicks, positions follow the usual methodology, and a follow-up question in AI Mode counts as a new query.
• Google's generative AI report: shows impressions in AI Overviews and AI Mode by page, country and device, but no queries.
• ChatGPT, Claude, Perplexity and Copilot prompts: no. These happen outside Google. They reach Search Console only indirectly, if the assistant or a service it uses runs searches on Google.

Search Console also leaves out very rare queries to protect privacy. Many unique, long prompts will never appear at all, even when they led to impressions.

Why might LLM query fan-outs inflate Google Search Console impressions?

AI systems expand one question into several searches. Google's help documentation says AI Mode groups a user's question into subtopics and searches for each one at the same time. Other assistants do something similar when they search the web.

Two things follow. First, a single conversation can produce searches on topics the user never typed. If those searches run through Google, and your page appears in the results, an impression can be recorded for a query that looks machine-written. Second, some assistants reportedly draw on Google's results through third-party scrapers. According to The Information in 2025, OpenAI had partly used Google search results scraped by SerpApi to help power ChatGPT answers, and Google later sued SerpApi over scraping. Automated requests like these can add impressions without any human seeing a result.

Google's documentation for AI Mode does not say whether its internal fan-out searches appear as separate queries in Search Console. Treat that as unknown.

Can GEO tools scraping cause unusual query formats in Google Search Console?

Yes. The strongest evidence comes from rank trackers. Between 8 and 10 September 2025, Google stopped supporting the num=100 parameter, which let tools fetch 100 results in one request. In a dataset of 319 properties analysed by Tyler Gargula and reported by Search Engine Land, 87.7% of sites saw impressions fall, and 77.6% lost visible keyword rankings. Many sites also saw average position improve. The widely accepted explanation is that tools had been generating impressions at deep positions, which disappeared when the parameter stopped working.

AI visibility tools that track Google's AI Overviews or AI Mode work the same way. They request Google results for your prompts, often long and conversational ones. If your page appears in those results, the request can register as an impression for a query that no customer typed. Tools running the same prompts daily from the same location can make such queries look like steady demand.

How do you tell tool-generated queries from real ones?

Look for this pattern:

• Long, templated wording, such as "best [category] for [audience] in [year]" repeated with small variations.
• Impressions but almost no clicks, over weeks.
• One device and one country, often desktop and the tracker's default market.
• Regular timing, such as daily spikes that match a tracking schedule.
• Prompts you or an agency track, appearing word for word.

Real long queries from people usually show more varied wording, some clicks, a mix of devices, and a link to real events such as campaigns or seasons.

How should you clean your reporting?

1. Segment by query length. Use a regular expression filter in the Performance report, such as queries of eight or more words, and report them separately.
2. Compare devices and countries for the long-query segment.
3. Check your own tracking lists. Ask whoever runs your trackers for their prompt lists and schedules, and match them to Search Console queries.
4. Use clicks and conversions as the primary measures for long queries, since impressions from automated requests rarely produce clicks.
5. Annotate known events, such as the September 2025 num=100 change, when comparing periods.
6. Keep a record of tracked prompts in a shared sheet, so anyone reading the report can see which queries are yours.
7. Read the generative AI report with the same caution. It is built from the same Web search impressions, so trackers that request AI Overviews or AI Mode add to it, and it has no query column to filter them out. Compare its daily pattern with your trackers' schedules before treating a rise as real visibility.

Should you ask trackers to stop?

Not before you measure. The tracking data is useful, so first compare long-query impressions with your trackers' prompt lists. But if long-query impressions distort reports leadership reads, ask your vendors how they collect Google results. Consider reducing tracking frequency for Google surfaces, or reporting impressions only for queries with at least one click. How to use genuine long-tail queries to find what people ask AI is covered in how to use Search Console long-tail queries to find AI prompts. Why AI assistants offer no Search Console of their own is covered in why there is no Search Console equivalent for AI search visibility.

What might an illustrative investigation look like?

Consider an illustrative software company whose Search Console showed a surge of impressions for queries such as "best project management software for small architecture firms 2026". It had no clicks. The queries came almost entirely from US desktop, every day at similar volumes. Its agency's AI visibility tool tracked 40 prompts on AI Overviews and AI Mode daily from a US location, and 31 of them matched the new queries word for word. The company split long queries out of its monthly report, used clicks as the headline measure for them, and moved the Google-surface tracking to a weekly schedule. Impressions from those queries fell by most of the difference, and the monthly trend line became readable again.

What do people ask about fake queries in Search Console?

Do AI crawlers like GPTBot create Search Console queries?

No. Crawlers fetch your pages directly; they do not search Google. Only searches on Google can create Search Console queries.

Does Google filter bot searches from Search Console?

Google's documentation describes impressions as how often someone saw a link to your site. The num=100 episode showed that automated requests could still affect the numbers.

Are long queries always fake?

No. People now type longer questions, especially in AI Mode. Check clicks, devices and timing before deciding.

Will this affect my rankings?

No. Search Console reporting reflects searches. It does not change how Google ranks your pages.

Can I block tracker queries from appearing in Search Console?

No. Search Console has no setting to exclude another company's searches. You can only filter queries out of your own reports, for example with a regular expression, and save that view for regular use. Keep an unfiltered view too, so you notice if the pattern changes.