← Back to all posts
Is llms.txt worth it? What the data says

Is llms.txt worth it? What the data says

LA
Jedi Solomon
··5 min read

Skip llms.txt, or give it an hour at most. In May 2026, 97% of llms.txt files received zero AI crawler requests, and Google discourages the file. Nothing I have seen shows it changes which sources ChatGPT, Gemini or Perplexity cite.

What the file is supposed to do

The idea is a plain markdown file called llms.txt at the root of your domain. It lists your most useful pages with a one-line description each. The pitch was a shortcut: hand the models a tidy map so they can skip the navigation, scripts and popups. It resembles robots.txt for AI crawlers and carries no rules. Nobody is obliged to read it, and the crawl numbers suggest nobody does. It is easy to see the appeal: the file is cheap to make and feels like control over how machines read your site.

What the llms.txt data says

The hard number is the 97% above. Those files were never fetched by an AI crawler that month. A file no bot fetches cannot shape an answer. Google discourages it, and I have found no study showing a citation lift from publishing one. That is why I would not spend an afternoon polishing the descriptions.

Schema markup gets similar treatment from the data. Structured data correlates only weakly with AI citations. It still matters for rich results and for feed consistency, so keep it accurate for those reasons.

What AI engines pick instead

To get cited by ChatGPT, Gemini or Perplexity, the evidence points at three things you can control.

First, ranking helps. Ahrefs looked at AI Overview citations in March 2026: 38% came from the top 10 results and 31% from positions 11 to 100. A page buried on page three can still be quoted, so publish the answer before you worry about where it will rank.

Second, other sites talking about you track AI visibility more closely than links to you do. Across 75,000 brands, Ahrefs found a 0.66 correlation between third-party brand mentions and AI Overview visibility, against 0.22 for backlinks. Correlation is all that figure shows, so treat it as a pointer for where to spend effort.

Third, local businesses have a different source list. Gemini grounds local answers on Google Maps. ChatGPT and Perplexity use Bing Places and Yelp. A location with no Bing Places listing gives ChatGPT nothing to read. The SOCi Local Visibility Index 2026 has ChatGPT recommending 1.2% of multi-location businesses, Gemini 11% and Perplexity 7.4%. If Google Business Profile is the only listing you ever set up, two of those three engines are reading sources you may never have touched.

What the work looks like in practice

JCS Fashions is a 4,878-product Shopify apparel store in Milpitas, California, and LynkAI's first customer. Its sessions from AI assistants went from 20 in January to 569 in October 2026. Clicks on its guide pages went from 0 in February to 231 in August. The work behind those numbers was about 290 quality-gated articles published from April to September, each reviewed by a person before it went live. Today 233 guide pages earn impressions on 24,164 queries. One customer is one data point, so treat these figures as a record of what happened there. The JCS Fashions story has the full numbers.

If you want the file anyway

If your CMS generates an llms.txt file in one click, publish it and move on. It costs nothing and risks nothing. Do not put it on a roadmap, and do not pay anyone to build it.

What to do this week

  1. Open your robots.txt in a browser and confirm it does not block GPTBot, OAI-SearchBot, PerplexityBot or Googlebot. A blocked crawler cannot cite you, whatever else you publish.
  2. Write down 10 questions your buyers ask before they purchase. Run each one in ChatGPT, Gemini and Perplexity and record which sites are cited. A spreadsheet with one row per question and one column per engine is enough. Perplexity citations are listed under every answer, so they are the easiest to log.
  3. For each question with no page of yours behind it, publish one page that answers it in the first two sentences. Put the detail after the answer. For a store that is a category guide. For a service firm it is a pricing or process page.
  4. Check that your Google Business Profile, Bing Places and Yelp listings match on name, address, hours and services. Mismatches are cheap to fix and easy to miss.
  5. Take the sites cited most often in step 2 and work out how to get mentioned on them: a review, a roundup, an expert quote or a directory listing.

Steps 2 and 3 are the same exercise our AI visibility audit runs at larger scale, with a panel of 25 buyer prompts. You can do a small version by hand in an afternoon.

If you would rather see where you stand before you start, the free AI visibility scan is the quickest way to get a first look.