The wrong reflex
The temptation is strong: see llms.txt appear in Lighthouse and conclude that the file is becoming an AI SEO factor.
That would be a mistake.
Google Search documentation states that site owners do not need to add an llms.txt file, Markdown or special files to be eligible for AI features in Search. That does not mean llms.txt is useless. It means it should not be sold as a ranking shortcut.
The distinction matters. A useful file is not necessarily a ranking factor. A context surface is not proof of recommendability. An emerging convention is not a citation guarantee.
What llms.txt actually provides
A well-designed llms.txt file can help a system understand where to look.
It can present:
- canonical pages for the organization;
- important resources for LLMs and agents;
- AI policies and use boundaries;
- governance files;
- proof pages;
- long-form content or reference documents;
- areas to avoid or interpret cautiously.
In that role, llms.txt acts as a machine orientation table. It does not replace the site. It only reduces the amount of guessing required to locate important resources.
What llms.txt does not fix
The trap is treating llms.txt as a magic layer placed on a fragile site.
An llms.txt file does not fix:
- a poorly structured offer;
- an ambiguous service page;
- weak or scattered proof;
- a brand confused with other entities;
- a form that is unclear in the accessibility tree;
- generic CTAs;
- contradictory pages;
- an interface that becomes unstable after hydration;
- an incomplete sitemap;
- missing editorial governance.
If the site is incoherent, llms.txt may simply point more efficiently toward that incoherence.
The right role in a Pagup strategy
At Pagup, llms.txt is treated as part of AI governance and machine readability, not as an end in itself.
It must align with:
- human-facing pages;
- sitemap;
- structured data;
- machine files;
- AI policies;
- inspectable proof;
- internal links;
- method and service pages.
That coherence matters more than the file’s isolated presence.
The connection with agentic readiness
In an agentic readiness strategy, llms.txt helps orient discovery, but it does not guarantee action.
An agent may find a page through llms.txt and then fail because the form has no labels, the main button is a fake interactive element, the confirmation is invisible or the page does not connect the action to the right context.
Agentic readiness therefore requires two layers:
- context surface: help the agent find the right resources;
- action surface: help the agent use the interface correctly.
llms.txt mainly supports the first. The Agentic Readiness Audit verifies the second.
Conclusion
Stop asking whether llms.txt “ranks”.
The better question is: does the site expose an architecture clear enough for a non-human system to know what to read, what to ignore, what to connect and what to do next?
llms.txt can help. But it will not save a site that has not already made itself readable.