Do I need an llms.txt file?

What Google has actually said

Google's own guide to optimising for generative AI features on Google Search states that you do not need to create new machine readable files, AI text files, markup or Markdown to appear in Google Search, including its generative AI capabilities, because Google Search itself does not use them. It adds that keeping such files for other services is completely fine, and that doing so will neither harm nor help your visibility or rankings in Google Search, because Search ignores them.

That is not a hint from a spokesperson on a podcast. It is the published documentation on Google Search Central, and it is the clearest statement anyone has on the subject.

The second useful public number comes from Ahrefs, who looked at 137,210 domains using their web analytics product and reported that 28% published an llms.txt file, and that 97% of those files received zero traffic in May 2026. Their write-up is public, with the sample size and the month stated. Files that nothing requests cannot be doing much.

What the correlation in our data really means

It means the file marks a certain kind of site, not that the file causes the score. Of the 300 businesses we scanned, 87 had an llms.txt and 213 did not. The 87 averaged 46.8 out of 100 for AI visibility and the 213 averaged 34.8. Among the 137 with complete records from both engines, 11 of the 41 with a file were never named, which is 27%, against 45 of the 96 without, which is 47%.

Now look at what else is true of those 87 sites. Seventy nine of them also had JSON-LD. Eighty of them allowed all six AI crawlers. Their mean site readiness was 93.6 out of 100 against 67.3 for the 213 without. A site with an llms.txt is overwhelmingly a site where somebody is paying attention to everything else as well.

Recomputed from data.csv, all 300 businesses. The 213 without a file are the 71% published in the study. One further confound worth naming: llms.txt is 15 of the 100 points in our own readiness score, and readiness feeds the AI visibility score, so part of that 12 point gap is our own arithmetic rather than the world's.

Why Answerable publishes one anyway

Because agentic readers and coding tools do fetch it, and because it costs us almost nothing to keep current. Our llms.txt is a plain index of this site for a machine that wants the map rather than the pages.

What we do not do is sell it. It is 15 of 100 points in our own readiness check, we say on The Receipts Engine that Google has stated it does not use the file, and we would rather lose a line item than let a client believe a text file bought them anything with Google.

If you like the idea, publish one. Keep it short, keep it accurate, and treat it exactly as seriously as you treat your sitemap, which is to say necessary plumbing that nobody is going to thank you for.

What llms.txt costs to add, honestly

About twenty minutes, once, if your site is small. That is our own estimate from doing it, not a measured figure.

It is a plain text file at the root of your domain listing your key pages with a line of description each. There is no build step, no library and no ongoing fee. The maintenance cost is remembering to update it when your pages change, which is the part that actually fails.

What to spend the money on instead

On the things that decide whether an engine is willing to name you, which are corroboration first and legibility second.

1. Stop blocking the retrieval crawlers. Free, and the 39 blockers in our sample averaged 21.3 against 40.9 for the 261 that allowed them.
2. Get the facts right everywhere, not just on your site. Listings, profiles, association pages, review platforms.
3. Earn mentions on pages you do not own. This is the slow, expensive, unavoidable one, and it is the one that separated the anchor brands from everyone else in our data.
4. Add the structured data. Cheap, correct, and modest in effect. See does schema markup help with AI search.

Notice that llms.txt is not on that list, and that the first item costs nothing at all.

Where to start

You do not need us to write a text file. If you want to know which of the four items above is actually your problem, the free scan checks the site and asks both engines, and you keep the report either way.