What this article covers: Official or not: Google’s data versus third-party scores, and what you don’t need.
The question "can AI find our website?" was difficult to measure for a long time. Third-party tools can test answers in individual AI systems and build their own visibility indexes, but an external score is not the same as data from the search engine itself.
In 2026, Google introduced generative search reporting in Search Console and rolled it out worldwide by the end of August. This gives website owners a much better starting point for measurement.
What can we see?
Where a domain has enough data, the report can show:
- impressions,
- pages,
- countries,
- devices,
- changes over time.
An impression does not mean that your website was the only source in an answer. It does show that the URL appeared in the measured generative search environment.
Tip: Official explanation of the report: Google Search Central — generative AI performance reports.
Why does this matter?
GEO/AEO made it easy to sell abstract "AI visibility scores". Search Console supports a more grounded process:
- start with real data,
- find pages that already receive generative impressions,
- compare them with normal organic performance,
- improve content and technical foundations,
- monitor change over time.
What does the report not tell us?
It does not mean:
- an AI system will always recommend you,
- your website was the primary source,
- every impression generated a lead,
- a higher external "AI score" automatically means a better business result.
That is why generative search data should be evaluated together with normal SEO and conversion data.
Do we need llms.txt for this?
Not for Google Search. Google has clarified that llms.txt is not required for Google Search and has no positive or negative effect on Search rankings or visibility.
It may still be used by other systems, but it should not be treated as the reason for expecting more Google AI impressions.
Do we need special AI schema?
No. Google does not require a special "AI structured data" format for generative Search.
Use normal structured data when it accurately reflects visible content, such as Organisation, Article, Product or LocalBusiness.
Tip: Source: AI features and your website (Google Search Central).
What should we check on the website?
- can Google crawl it normally,
- is it indexable,
- is important content available in HTML,
- does each page have a clear purpose,
- is business information consistent,
- does the page contain verifiable information,
- are internal links useful,
- does structured data match the visible content,
- is there a clear next step for the visitor?
Tip: What we check and what we do not promise: AI search optimisation (GEO/AEO). How to test ChatGPT and Perplexity manually: How to check if you appear in AI answers.
Frequently asked questions
Where in Search Console do I find generative search data?
In the Performance report for Google Search, where Google reports impressions in generative features separately. If a domain does not have enough data, the report stays empty — that is not an error by itself.
Does a generative impression mean my page was the main source?
No. It means the page appeared in the measured generative environment. Whether it was the only or main source is not visible from the data.
Does llms.txt increase Google AI impressions?
No. Google has clarified that llms.txt is not required for Google Search and has no effect on visibility. The file may still be useful for other systems.
Do I need special AI schema?
No. Google does not require a special format for generative Search. Use normal structured data that matches the visible content (Organisation, Article, Product, LocalBusiness).