Google says that llms.txt does not help with SEO. Why does Lighthouse check it anyway?

Readers who compare the recent documentation from Google and Lighthouse may have doubts. This is justifiable. Google Search states that you do not need a new machine-readable AI file to appear in regular search results, AI Overviews or AI Mode. At the same time, Lighthouse checks in the experimental category of Agentic Browsing whether there is a llms.txt file on your domain.

This is not contradictory; both systems answer a different question. Google Search wants to determine which pages it can index and display. Lighthouse examines whether a website is technically usable for software that reads and serves pages. A check in Lighthouse is therefore not a ranking factor.

What Google Search officially says

In the documentation on AI features in Google Search, it states that the existing SEO fundamentals also apply to AI Overviews and AI Mode. There are no additional technical requirements, and no special optimization is needed to qualify for them.

Google mentions well-known fundamentals:

1. allow crawling via robots.txt, your CDN, and your hosting;
2. ensure important pages are accessible through internal links;
3. provide important content as readable text;
4. create useful, reliable content for people;
5. ensure structured data matches what visitors see on the page;
6. keep your Google Business Profile and other business information up to date where relevant;
7. ensure a good user experience.

Google also explicitly states that new machine-readable files, AI text files, or special schema formatting are not needed to appear in AI Overviews or AI Mode. A llms.txt file therefore does not provide a direct ranking advantage and does not serve as a separate gateway to these Google features.

What is a llms.txt file?

llms.txt is an emerging convention. The file is located in the root of a domain, for example, https://example.com/llms.txt, and usually contains a concise summary in Markdown with links to important pages.

An AI agent uses it as a starting point to find important information and the structure of a website. This can help certain agents make fewer detours. It is not yet a widely accepted web standard, and support varies by agent.

Why Lighthouse checks llms.txt anyway

The Lighthouse audit for llms.txt is part of Agentic Browsing. This category does not look at your position in Google but at technical signals that may be important when a browser agent tries to understand or serve a website.

According to Chrome, the audit checks whether the file can be technically retrieved. If the server returns an error, Lighthouse reports that. If the file does not exist and the server responds with a 404, then the audit is currently not applicable. The file is therefore optional.

The scoring of Agentic Browsing is experimental. The category is experimental. It does not provide a weighted score from 0 to 100. It shows how many deterministic checks pass and what technical errors or warnings exist. A successful check of llms.txt only means that the file is accessible as expected. It says nothing about Google ranking.

Other checks in this experimental category look at programmatic names and labels, the accessibility tree, semantic relationships, layout stability, and possible WebMCP integration. These are signals for machine interaction, not for the quality or authority of your content in Google Search.

Search engines and browser agents have different goals

A search engine collects and ranks information to compile a relevant results page or an AI answer. A browser agent tries to perform a concrete task on a website, such as finding information, understanding a form, or interacting with an element.

Therefore, a file can be useful for an agent without being a ranking factor. A sitemap helps crawlers discover pages, but it does not automatically make weak content strong. An accessible label helps software understand a button, but it does not prove that the page is the best search result. This separation also applies to llms.txt.

Practical advice for businesses and developers

Do you not have a llms.txt? Then there is no reason to panic. First, work on pages that are crawlable and indexable, clear navigation, unique content, correct business information, semantic HTML, good accessibility, and stable performance.

If you still want to add the file, keep it small and maintainable:

1. clearly describe what the website is about;
2. link only to current core pages;
3. do not include claims that cannot be verified on the website;
4. ensure the URL returns a normal successful response;
5. update the file when services, routes, or documentation change.

An outdated file can direct agents to missing pages. It can also give a false impression of your organization. This is primarily a maintenance issue, not a Google penalty. Therefore, only add it when someone is responsible for the content.

What FlowCore checks about this

FlowCore treats these topics separately. The technical site audit checks whether /llms.txt is accessible and not empty. Additionally, it checks classic signals such as title tags, meta descriptions, one clear H1, canonical tags, robots.txt, sitemap references, structured data, semantic headings, hreflang, and alt texts.

The speed analysis uses PageSpeed Insights for performance, accessibility, best practices, and SEO. The experimental Agentic Browsing category is not yet available via the public PageSpeed API used and is therefore not presented as a full-fledged Lighthouse score. On the page how FlowCore measures, you can read how technical checks and AI visibility measurements remain separate.

This ensures the AI visibility scan does not confuse an experimental technique with a proven SEO factor. You can see separately what search engines technically need, what signals AI systems can find, and where a file like llms.txt truly fits into that whole.

Conclusion

A llms.txt file can be retained or added when you can maintain it correctly. Do not expect better Google ranking or automatic visibility in AI Overviews from it. For now, see it as an experimental layer for certain AI agents, not as a replacement for SEO, qualitative content, or a technically sound website.