Does llms.txt Actually Work?
Short answer: not in the way it's usually sold. llms.txt is a real, growing proposal, but as of 2026 there is no evidence that publishing one increases AI citations, traffic, or visibility in Google AI Overviews. That does not make the file worthless. It means the file is being oversold.
The realistic view: llms.txt is a lightweight, optional format that may help some AI agents and developer tools read a site more efficiently. It is not a proven ranking factor, not a proven citation booster, and not a substitute for crawlable pages, clear structure, or genuinely useful content.
What llms.txt actually is
llms.txt is a proposed standard for publishing a plain-text Markdown file at the root of a domain, usually at /llms.txt, giving language models a curated map of a site instead of forcing them to parse the full page the way a browser would.
Most websites are built for browsers, not language models — navigation, JavaScript, cookie banners, repeated templates. llms.txt tries to strip that away and offer a short reading list instead. It is not a crawler-control file like robots.txt, and it is not a ranking signal like a sitemap or schema. It is closer to an optional orientation layer some AI systems may choose to read, if they read it at all.
Why it spread so fast
The idea fits a believable narrative: AI search is growing, brands want to be cited, and a small text file at the root of a site sounds exactly like the low-friction fix people are looking for. It is also easy to explain and cheap to implement, and the name deliberately echoes robots.txt, which makes it simple for agencies and plugins to package as a must-have tactic.
That is why llms.txt became popular — it is plausible, not because it is proven. That distinction matters if you are deciding where to spend an afternoon.
What Google actually says
According to Google Search Central, Google updated its AI optimisation guidance in June 2026 under the heading "Clarifying guidance on llms.txt files," stating plainly that these files are not needed to appear in Google Search or its generative AI features, and that having one creates neither a positive nor a negative effect on visibility or rankings.
That guidance did not come out of nowhere. At Google's Search Central Deep Dive event in the Asia-Pacific region in mid-2025, Gary Illyes said ranking in AI Overviews requires ordinary SEO and that Google does not support llms.txt and has no plans to. John Mueller went further and compared it to the long-abandoned keywords meta tag, noting that no AI service had asked Google to adopt it.
So if llms.txt is being sold as a shortcut into Google AI Overviews specifically, that claim is ahead of the evidence. The honest version is "some non-Google systems might find this useful in some contexts" — not "Google needs this."
What the independent data shows
Two large studies from 2026 tell a consistent story.
According to an SE Ranking analysis of roughly 300,000 domains, there is no statistically significant correlation between having an llms.txt file and being cited more often in AI answers. When the file was removed as a variable from their citation-prediction model, the model's accuracy improved — a sign it was adding noise, not signal. Of the 50 most-cited domains in the study, only one actually had an llms.txt file.
According to an Ahrefs study covering 137,000 domains, 97% of published llms.txt files received zero requests during the monitored period, and fewer than 20% of the requests that were logged came from identifiable AI bots or tools. In practice, the overwhelming majority of these files are read by no one — not even the systems they are meant for.
The evidence, in one place
Where it might genuinely help
The most defensible case is a site with dense documentation, API references, or structured knowledge bases. There, an llms.txt file can point an agent to the right sections quickly and reduce ambiguity about what matters most — especially if the site also offers Markdown versions of key pages.
It may also help in workflows built around navigation rather than ranking — a coding assistant or research agent working through a curated list of canonical pages and short summaries is a very different use case from consumer-facing AI search visibility.
There is a smaller, less measurable benefit too: writing the file forces a team to decide which pages are actually central and how to describe the business cleanly. Even if no system reads it tomorrow, that editorial exercise is not wasted.
Where it probably will not help
For most marketing sites, local businesses, e-commerce stores, and brochure-style company sites, the likely benefit is small. These sites usually are not held back by a lack of machine-readable orientation — they are held back by thin content, inconsistent entity signals, weak internal linking, or unclear positioning.
That is also why plenty of sites with no llms.txt at all still get cited constantly. AI answers are built from the same underlying signals that already matter on the web: clear content, structured meaning, and strong external corroboration. A new text file does not rescue pages that are weak on those fundamentals.
Is there a downside?
Not much, as long as it is treated as optional and low-stakes. A file that is accurate and maintained is unlikely to hurt anything.
The real risk is not technical. It is opportunity cost and false confidence — a team that ships an llms.txt and quietly stops working on the fundamentals that actually move citations, or a generic, auto-generated file that goes stale and defeats the entire point of offering a curated map in the first place.
Should you publish one anyway?
In many cases, yes — for the right reason. If it costs little, is easy to maintain, and the site has a real body of reference content, publishing one is a cheap forward-compatibility bet. It is optionality, not a performance tactic.
That case is strongest for documentation sites, SaaS companies with substantial help centres, developer platforms, and API products. For a local services business or a small brochure site, the upside is minor — fine to ship, but it should sit near the bottom of the priority list, well below technical SEO, content depth, and clear business information.
What to do instead of overthinking llms.txt
If the real goal is visibility in AI-generated answers, the boring strategy still wins: publish pages that answer real questions clearly, keep them crawlable and indexable, organise the site logically, and reinforce entity signals across the web — reviews, third-party mentions, consistent business information.
In practice that means building better comparison pages, sharper FAQs, deeper product and service detail, and clearer authorship and topical expertise. None of it is as trendy as a text file at the root of a domain, but all of it is more likely to move the needle. Publishing llms.txt on top of that work costs nothing. Publishing it instead of that work is the actual mistake.
Verdict
If "work" means producing a measurable boost in AI Overviews, citations, or organic visibility, the 2026 evidence says no. If "work" means offering a lightweight, structured guide that some agents or future systems might find useful, then yes — in that narrower sense.
That is the honest answer right now. llms.txt is real, cheap, and reasonable to ship. It is not an adopted universal standard and it should be treated as optional infrastructure, not a core growth tactic.
