Jeremy Howard
Co-founder & CEO, Answer.AI
"Site owners should decide what an LLM reads, no more random scraping."View LinkedIn profile
Does Llms.txt Really Affect GEO?
Not enough evidence yet. llms.txt is a real proposal with a reasonable use case, but the public record still does not show strong proof that it directly improves AI citations today.
The Case For Yes
The strongest argument that llms.txt does affect GEO.
- It gives you control over what LLMs read. The proposal lets site owners hand models a clean, flattened map of their best content instead of leaving them to scrape at random.1
- It is a maturing, recognized convention. The spec is formalized at a standard root-level path and is being picked up by docs platforms and tools,2 so early movers are positioned if engines adopt it.
- It lowers the cost of understanding your site. A concise, link-rich markdown index reduces the work an LLM does to locate your key pages, which can help inference-time use cases.
- It is low-cost and reversible. A single static file is cheap to publish and trivial to remove,3 so the downside is minimal against the optional upside.
- It is a forward bet on an agent-first web. As autonomous agents proliferate, an explicit machine-readable guide aligns with how sites are increasingly consumed.
Practitioners making this case
Andrea Volpini
Co-founder & CEO, WordLift
"A curated, machine-readable map of your site is a reasonable bet on an agent-driven web."View LinkedIn profile
Gianluca Fiorelli
Strategic & International SEO Consultant
"llms.txt costs almost nothing to publish, a cheap hedge if adoption ever comes."View LinkedIn profile
The Case For No
The strongest argument that llms.txt does not move GEO.
- No AI system currently uses it. Google has stated plainly that no AI system reads llms.txt today,1 so there is no live mechanism for it to affect citations.
- Google won't support it and likens it to meta keywords. Search leadership says normal SEO is what works for AI Overviews, and llms.txt won't be used.2
- A spec is not adoption. A formalized file format means nothing for GEO until crawlers actually request and act on it,3 which they don't.
- It is a spam and drift magnet. A separate hand-maintained file easily falls out of sync with the real site or gets gamed, adding risk without reward.
- It duplicates HTML the models already read. LLMs have parsed normal web pages from the start, so a parallel file is redundant effort.
Practitioners making this case
John Mueller
Search Advocate, Google
"FWIW, no AI system currently uses llms.txt."View LinkedIn profile
Kevin Indig
Growth Advisor & author, Growth Memo
"llms.txt is a solution looking for a problem. No major LLM is reading it today."View LinkedIn profile
Will Critchlow
Founder & CEO, SearchPilot
"There's no test or platform showing llms.txt does anything. Treat it as a bet, not a tactic."View LinkedIn profile
Why The Claim Stays Uncertain
- A public spec is not the same thing as confirmed platform adoption.
- Major AI platforms have not clearly documented llms.txt as a standard citation, crawl, or ranking input.
- The best current argument for llms.txt is future-readiness and cleaner LLM guidance, not proven present-day GEO impact.
My Expert Opinion
I see llms.txt the same way I see many emerging protocol ideas: strategically interesting, operationally cheap, and easy to overstate. If you already maintain clean documentation, markdown-friendly pages, and machine-readable site structure, adding llms.txt is reasonable. But I would not sell it as a proven GEO win yet.
The right framing is optional infrastructure. It may become useful if AI systems standardize around it more explicitly. Right now, the evidence is too thin to call it TRUE and too open-ended to call it DEBUNKED.
Verdict
PLAUSIBLE
llms.txt may affect GEO in the future or in narrow inference workflows, but there is not enough public evidence to say it currently and reliably improves AI citations or generative visibility across platforms.