Why 94% of AI responses ignore companies' own websites
Companies are facing a new challenge: their websites rank in the top 10 search results, but when a client asks about a contractor via ChatGPT, Perplexity, or Google’s AI summary, the company is absent from the response. Instead of referencing actual providers, neural networks compile selections from external directories and reference sources, writes a Habr author.

The author tested 200 real B2B queries: in 94 out of 100 industry selections where the AI summary appeared, corporate websites were almost never included. The reason is that the model verifies a company through independent sources: it searches for the company name along with the city, then checks the tax identification number (UNP) and separately confirms whether the legal entity exists and whether the services match. A company’s own website content is treated as self-promotion by AI: without external verification via UNP, the site is filtered out.
The llms.txt file offers no solution—Google officially confirmed it does not use it for ranking or AI overviews. Meanwhile, bots actively extract data in Markdown and JSON formats: over a month, a directory containing 30,900 profiles recorded 773,704 requests.
The solution is an external 'digital passport' for the company: an open, machine-readable profile in a neutral registry linked to government databases (UNP/INN), with specific services and geographic coverage, available in HTML, Markdown, JSON, and Schema.org, as well as two-way verification of the connection with the website. Without this, businesses risk being excluded from AI-generated shortlists.




