Adding a file to the root of a website takes minutes. Untangling a consultancy homepage that calls the same service three different things, hides key copy behind JavaScript and omits its organisation details from structured data does not.
That difference explains why llms.txt has become such an attractive GEO task. It is neat, visible and easy to report. For Dubai, UAE and UK businesses trying to improve AI visibility, though, it can become a distraction from the work that determines whether a machine can properly identify the business and extract a useful answer.
The file is simple. The assumption behind it is not.
llms.txt is an emerging convention for publishing a concise, machine-oriented guide to a site’s important information. It can point a system towards useful pages, explain the site’s structure and provide a cleaner route to material that might otherwise be scattered across navigation, PDFs and marketing copy. Related files such as llms-full.txt and agents.md can serve similar orientation or instruction purposes.
The leap happens when that practical publishing aid is treated as a direct ranking, citation or recommendation mechanism. Publishing llms.txt does not give a site a confirmed advantage in Google AI features, ChatGPT, Gemini, Perplexity or Copilot. Independent platforms decide what they crawl, retrieve, cite and recommend using their own systems and changing policies.
Does llms.txt help AI visibility? It may help when a system chooses to consume it and the file accurately points to clear, accessible and useful published information. Current evidence does not support treating it as a reliable shortcut to rankings, citations or recommendations. Crawlability, stable page access, consistent entity information, answer-ready content and credible corroborating sources remain more dependable foundations.
Requests matter more than enthusiasm
A file cannot guide a system that never requests it. Ahrefs analysed server logs across more than 137,000 domains and reported that 97% of published llms.txt files received no requests during its study period. Its analysis of llms.txt server-log requests is a useful corrective to confident claims about widespread consumption.
That does not make the convention pointless. It means teams should apply a normal technical standard: publish it where it is tidy and maintainable, then watch for actual requests in server logs rather than assuming adoption. A file which receives no requests is not harming a site by existing, but it is not doing much work either.
Where the real failure usually sits
Consider a consultancy that publishes llms.txt immediately. Its homepage still describes the offer as advisory, strategy and transformation support depending on the paragraph. Service pages use different names again. The organisation schema is absent, the contact page has a different trading name, and the primary service copy is loaded only after a script runs.
An orientation file can link to those pages. It cannot make their claims consistent, establish what the company actually does, repair an ambiguous entity relationship or guarantee that a crawler sees rendered content correctly.
That is the more expensive failure. A team can keep updating a file that few systems request while leaving buyers and machines with no dependable answer to basic questions: who is this business, what does it provide, where does it operate and what evidence supports those claims?
Put llms.txt in the right place in the work queue
Before publishing or revising a machine-readable file, check whether the underlying pages deserve to be pointed at. The order below is less glamorous than deploying a new text file, but it is usually more useful.
| Check | What to inspect | Priority |
|---|---|---|
| Access | Important pages return a valid status, are not blocked by robots.txt and appear in the XML sitemap where appropriate. | High |
| Service definitions | The homepage, service pages, page titles and structured data use the same clear service names. | High |
| Entity clarity | Organisation details, trading name, location, contact information and relevant relationships are consistent. | High |
| Answer-ready pages | Pages answer the questions a buyer would ask without requiring a sales call to decode the offer. | High |
| Machine orientation files | llms.txt, llms-full.txt or agents.md accurately point to maintained, canonical published material. | Medium |
There is one boring detail worth checking: do not list URLs in llms.txt that redirect, return errors, are blocked from crawling or duplicate a stronger canonical page. That turns an orientation document into another source of ambiguity. Keep the file short, factual and aligned with the public site rather than filling it with promotional claims.
A sensible use case for publishing it anyway
llms.txt is most defensible when a site has a substantial set of maintained resources and a clear reason to help machine-oriented consumers navigate them. A consultancy with defined services, named sectors, useful explainers, case material and policy pages may benefit from a concise index of canonical routes.
It is also a low-risk addition when it sits inside a broader technical programme. In the FlareFalcon AI-readiness case study, llms.txt, llms-full.txt and agents.md were part of wider technical, structural and business-clarity improvements. The files were not presented as the reason a readiness score improved. That distinction matters. FlareFalcon’s automated score measures selected website signals, not whether an independent AI platform will cite or recommend an organisation.
FAQ
Does Google use llms.txt?
Google has not published guidance identifying llms.txt as a Google ranking factor or a control for citation in AI features. Google can discover and assess pages through established crawling and indexing systems. Treat llms.txt as an optional publishing convention, not a substitute for indexable pages, sensible internal linking and technically accessible content.
Do AI assistants read llms.txt?
Some tools and agents may support or experiment with the convention, but consumption is not universal and platform behaviour changes. The Ahrefs server-log study found limited observed requests across published files. Check your own logs if possible, but do not infer broad AI visibility gains from a single request or a favourable answer.
When is llms.txt worth publishing?
Publish it when you can maintain a concise file that directs machines towards clear, canonical and useful public material. It is a reasonable finishing task after access, page quality, entity consistency and structured relationships are in order. Skip the theatre of treating it as the first or only GEO deliverable.