GeoNow that llms.txt has a name people recognize, a second file is showing up next to it: llms-full.txt. It sounds like an upgrade. It isn't quite that, and the adoption data from this year shows most sites still don't need it.
What is the difference between llms.txt and llms-full.txt?
llms.txt is an index. llms-full.txt is a full-text dump. llms.txt sits at yourdomain.com/llms.txt and lists a short description of your site plus links to your most important pages, the way a table of contents points to chapters. llms-full.txt takes the next step: instead of linking out, it flattens the actual text of your key pages into one long Markdown file that an AI system can read in a single pass without following any links at all.
The size difference is the whole story. A typical llms.txt runs a page or two. A real llms-full.txt for a documentation-heavy site can run to tens of thousands of words, and on a large site it can stretch into the hundreds of thousands, according to reporting from Originality.ai's June 2026 tracking study. That's the tradeoff: llms.txt costs almost nothing to maintain, llms-full.txt costs real, ongoing effort every time your content changes.
Who is actually publishing these files?
Both formats are growing, but from very different starting points. Rankability's monthly crawl of the top 1,000 websites found 8.7% now publish an llms.txt file as of June 2026, up from 0.3% a year earlier. Of those 87 sites, only 15 also published an llms-full.txt, according to Rankability's llms.txt adoption report. llms-full.txt is still the minority format even among sites that have already committed to the index version.
Originality.ai's broader tracker, covering more than 3 million sites, tells a similar story from a different angle: llms.txt grew 8.8x over twelve months to reach 36,120 sites by May 2026, while llms-full.txt grew 107.1x over the same period, from 23 sites to 2,463. That's a fast growth rate off a tiny base, and by count llms-full.txt still made up only 6.3% of the combined file distribution. Fast growth on a small number is still a small number.
Does either file actually get read?
This is the part that should shape your decision more than the growth charts. Ahrefs pulled server-log data from 137,000 domains and found that 97% of llms.txt files received zero requests in May 2026. Where requests did happen, AI retrieval bots such as the ones powering ChatGPT or Perplexity answers made up just 1.1% of them. SEO audit tools accounted for 21.7%, the largest single share, meaning the files are being checked by monitoring software more than they're being read by the AI systems they were built for.
Google has been explicit about its own position. Its May 2026 AI search guidance, updated again in June, states that llms.txt carries no positive or negative effect on Search visibility or AI Overviews. Chrome's Lighthouse did add an agentic browsing audit that checks for the file in May 2026, but that audit is about whether autonomous browser agents can find their way around your site, not about ranking. Two different jobs, easy to conflate.
Who should bother with llms-full.txt specifically?
The use case is narrow. llms-full.txt makes sense for a documentation site, an API reference, or a knowledge base where the entire content set is small enough to fit inside an AI model's context window and dense enough that an agent benefits from having it all in one file instead of crawling page by page. That's a real, if small, category of business.
It does not make sense as a default move for most businesses. A dentist's office, a restaurant, a law firm, or a ten-page local service site doesn't have a documentation problem an AI agent needs solved. For context on how far behind even large companies are here: only 7.4% of Fortune 500 companies had implemented any llms.txt as of March 2026, against 92.8% that had a robots.txt, a standard file that's been around since 1994. This is still an early, mostly unproven layer of the stack, and llms-full.txt sits a step further out on that same curve.
| llms.txt | llms-full.txt | |
|---|---|---|
| What it is | Index with links | Full site text, one file |
| Typical size | A page or two | Thousands to hundreds of thousands of words |
| Top 1,000 site adoption | 8.7% (June 2026) | 1.5% (15 of 1,000) |
| Read by AI search engines for citation | Rarely, per Ahrefs logs | Rarely, same pattern |
| Best fit | Any site, low effort | Large docs or API references only |
So what should you actually do?
If you already generated an llms.txt file, an afternoon project like that is fine to leave in place. It costs nothing to maintain and it may help the agent workflows that do check it. Don't build an llms-full.txt on top of it unless you run a documentation-heavy site and have a real reason to keep a full-text export current as your content changes. For everyone else, that effort is better spent on the things proven to move AI citation: making sure robots.txt actually allows GPTBot, ClaudeBot, and PerplexityBot, adding structured data that names your specific business type, and writing pages that answer real customer questions in plain, direct language.
GeoRoar's free audit checks exactly those signals, the ones with evidence behind them, and gives you a score out of 100 in about 30 seconds. If the fixes need generating, that's the $99 Fix Pack. If you want it watched and re-scored as your site changes, that's Monitor at $19 a month.
Where's your site losing points?
Here's a sample GeoRoar scorecard. Yours takes about 30 seconds.
This is a sample. Your real scorecard is free, no sign-up.
Run your free GEO audit