Do Schema and llms.txt Really Help AI Visibility?

Nejc Dobrila

Nejc Dobrila, co-founder of Website Gorillas

Last updated: 4 min read

Illustration: schema code next to an llms.txt file

There is a lot of technical advice around AI visibility. Some of it holds up, some of it is overrated. Here is an honest overview.

Structured data (schema)

What it is: markup in your page code (most often JSON-LD) that tells machines what's on the page: a company, a service, an article, an FAQ.

What the data says: between August 2025 and March 2026, Ahrefs tracked 1,885 pages that added JSON-LD schema and found no statistically significant increase in AI citations on any platform.

When it still helps:

  • Organization schema ties your site, LinkedIn and other profiles into one entity,
  • FAQPage and Article schema help machines understand the structure,
  • it helps in Google results, which remain an important source.

Conclusion: use schema as a foundation, but don't expect it to move mentions on its own.

llms.txt

What it is: a proposed file in the root of your site that gives language models a simple summary of the site's most important content.

What we know: it is a proposed standard, not an official rule. As of October 2026, none of the major AI search providers has publicly confirmed using llms.txt to choose sources. Google says no special files or markup are needed for its AI features.

Conclusion: the file is easy to add and does no harm, but don't rely on it.

robots.txt: what really counts

While schema and llms.txt are extra signals, robots.txt can completely block AI search engines from your site. It is the most common technical mistake.

AI search engineBot namePurpose
ChatGPTOAI-SearchBotAppearing in ChatGPT search (if you block it, ChatGPT won't show you in search answers)
ChatGPTChatGPT-UserVisits when a user asks ChatGPT for a specific page
ChatGPTGPTBotModel training
ClaudeClaude-SearchBotSearch in Claude
ClaudeClaude-UserVisits at a user's request
ClaudeClaudeBotModel training
PerplexityPerplexityBot, Perplexity-UserSearch and visits at a user's request
GoogleGooglebotSearch, including AI Overviews

If you want to appear in AI answers, don't block the search bots. You can block training bots at your own discretion. Also check your firewall or CDN, which sometimes blocks bots without you knowing.

What has a bigger effect

  1. Access. AI must be able to read the site: an open robots.txt, content as text, fast loading.
  2. Clear answers. The answer in the first sentence. More in What is AEO.
  3. Third-party sources. According to Bain (2026), 89% of unbranded AI prompts, those that don't name a specific brand, are answered from third-party sources rather than the brand's own website.
  4. Consistent facts on your website, LinkedIn and in directories.
  5. Working links. 2.38% of URLs cited by ChatGPT return a 404 error (Ahrefs). Proper redirects preserve your visibility.

Frequently asked questions

Should I block GPTBot?

GPTBot is used for model training. According to OpenAI, blocking it doesn't affect appearing in ChatGPT search, which is controlled by OAI-SearchBot. The choice is yours.

Does Google-Extended affect AI Overviews?

No. Google-Extended controls the use of content for Gemini models, while AI Overviews use the regular Googlebot.

Does a developer need to add schema?

Yes, schema is part of the page code. We add it to every new site and in our AI plans.

AI visibility

Let's check your technical foundations

Send an inquiry and we'll check whether AI search engines can read your site at all.

Sources

Preberite ta članek v slovenščini