What llms.txt actually is
llms.txt is a plain text file you put at the root of your site — yoursite.com/llms.txt — that lists your key pages and content in a clean, machine-readable form. The idea is loosely borrowed from robots.txt and your sitemap: a simple signpost for crawlers, except this one is aimed at large language models instead of search bots. It’s cheap. It’s harmless. And it takes about ten minutes to add.
Nameworthy publishes an llms.txt and treats it as hygiene rather than a lever: in the largest study to date, across 137,210 domains, 97% of llms.txt files received zero requests in a month.
Why you’re suddenly hearing about it
The format went from obscure proposal to talking point over 2026. Free generators appeared — paste your URL, get a file back — so the cost of adding one dropped to nothing, and adoption climbed. An Ahrefs study of 137,210 domains found 28% of them publish llms.txt — and that 97% of those files received no traffic at all in the month measured. Adoption is real; the reading is not.
Growing adoption is not the same as proven effect, though. That gap is where most of the noise lives.
The honest verdict: hygiene, not a lever
Here’s the part the pitches skip. There is no strong evidence that the major AI engines currently read llms.txt to decide who they recommend — and Google has now said its half on the record: its AI-features documentation states you don’t need machine-readable files, AI text files, or markup to appear in Search or its AI features. None of ChatGPT, Claude, or Perplexity has confirmed using the file either, and independent testing hasn’t turned up a signal. So publish it for what it is: a near-zero-cost bet on the standalone assistants and agents beyond Google — never a Google deliverable, and never ahead of the schema and crawlability every engine actually reads.
So treat it like hygiene, not strategy. Adding llms.txt is fine — it’s cheap, low-risk, and if the engines do start leaning on it, you’re already set up. But it is not a lever that moves whether AI recommends you, and nobody should sell it to you as one.
Don’t buy an “llms.txt strategy”
If someone pitches you a paid llms.txt strategy — an audit, ongoing optimization, a monthly retainer to maintain a text file — that’s the upsell to walk away from. The file is a ten-minute job you can do yourself with a free generator. Paying for it every month is paying for the appearance of work. It sits in the same bucket as schema markup: technically fine to add, easy to oversell, and not where the recommendation is won.
Where your effort actually goes
The things that move AI recommendations are the unglamorous ones, and they’re roughly the same across engines. None of them is a text file.
- Directory profiles — claimed, complete, and on the directories your engines actually read.
- Genuine reviews — score, volume, and the specifics reviewers mention, because several engines rank on them directly.
- Quotable pages — each answering one real customer question with a concrete, checkable fact.
- Consistent entity information — the same name, address, and details everywhere, so engines trust who you are.
When we run a report, llms.txt isn’t on the fix list, because it isn’t what’s holding you back. The fixes that are on it are the ones above.
Add it, then move on
By all means add llms.txt this afternoon. Ten minutes, done, forget it. Then put your real effort where the recommendations are actually decided — that’s what the full playbook walks through, engine by engine. If you’d rather see where you stand before you touch anything, start with a free report.