The llms.txt Placeholder Problem: Why Half of Adopters Ship an Empty File

Adding an llms.txt file and finishing it are two different things, and most sites only do the first one. In the August 2026 State of GEO edition — GeoReady's audit of 282 live, self-selected domains — 62.8% of sites had some form of llms.txt file. But of the 177 sites that had one, only 85 published a complete, structured file. The other 92 — 52% of everyone who "adopted" llms.txt — shipped a placeholder: a file that exists and returns a 200, but doesn't do the one job llms.txt exists for. Adoption isn't the bottleneck anymore. Follow-through is.

This piece looks at that specific gap: what counts as a placeholder, why it happens, how to tell if your own file is one, and what a real one looks like instead.

Adoption is climbing. Completion isn't keeping up.

The August 2026 State of GEO edition tracked three llms.txt figures worth separating, because they tell different stories:

• 62.8% of the 282 audited domains had an llms.txt file of some kind — up 8.6 percentage points from July's 54.2%. Adoption is genuinely rising.
• 30.1% had a full, structured file — sectioned links, real descriptions, the format actually built to orient an AI system. Also up, but only slightly, from July's 26.9%.
• 52% of the 177 domains with any file — 92 of them — published a placeholder instead: a stub with no meaningful links, not a working file.

Put another way: the gap between "has a file" and "has a useful file" didn't close as adoption rose in August — it's most of the story. For the full August findings — score bands, category efficiency, the other adoption signals — see the August 2026 State of GEO edition ; this piece only follows the one thread.

What counts as a "placeholder" llms.txt

A placeholder file exists at /llms.txt, returns 200, and still fails at its job: giving an AI system a curated, site-specific orientation to what you do and where your important pages live. In practice, a placeholder is missing one or more of the things a real file needs:

• No real sections or links — just a title line, sometimes one generic sentence, no ## sections pointing anywhere.
• Boilerplate copy — text that reads like "This is our llms.txt file" rather than anything specific to the business.
• A couple of links that don't cover what matters — the homepage and one blog post, nothing pointing to product, docs, or pricing.
• An unedited example — a tutorial's or generator's sample output, published verbatim with the placeholder domain still in it.

For the exact format a complete file follows — H1 title, blockquote summary, sectioned Markdown links — see What Is llms.txt? , the definitional guide this piece builds on.

Why this happens

None of this shows up as a measured cause in the State of GEO dataset — the report counts files, not intentions — but the pattern is consistent with how llms.txt tends to get added in practice:

• Treated as a one-time checkbox. Someone reads that llms.txt is "a thing now," generates a file once, and never returns to it.
• Generator output ships unedited. A tool can produce a valid skeleton from a sitemap in seconds — a good start, not a finished product, if nobody fills in real descriptions and prunes the list.
• No clear owner. Sitemap.xml is usually handled by a plugin or build step; robots.txt tends to have someone who owns crawl access. llms.txt often has neither, so it lands once and nobody keeps it current.
• Copied, not written. A generic file is faster to publish than a curated one, and nothing forces a rewrite before it goes live.

How to tell if yours is one

You don't need an audit to get a first read — five minutes will tell you most of what you need:

• Open https://yourdomain.com/llms.txt directly in a browser and read it top to bottom.
• Count the real links. Does it point to actual product, docs, guides, or pricing pages — or just one or two generic ones?
• Read the blockquote summary. Is it specific to your business, or could it belong to any company?
• Check the structure. Does it have ## sections (Docs, Product, Guides) grouping related links, or is it a flat, three-line file?
• Check the date it was last touched against when your site last shipped a major page, product, or rename — a stale file often reads exactly like a placeholder one.

If you'd rather have this checked automatically alongside your other AI-visibility signals, run the free AI SEO audit — it's a single-URL audit with no account required, and llms.txt is one of the eight categories it reports on.

What a real llms.txt contains

A complete file is short but specific: an H1 with your site or brand name, a one- or two-sentence blockquote describing what you do and who you serve, and ## sections of Markdown links to the pages that actually matter — each with a short, factual note about what a reader will find there. That's the whole format; there's no XML and no validator to pass.

The full breakdown, with a real example, lives in What Is llms.txt? . If your file is missing or you're starting from a placeholder, the free llms.txt generator builds a real starting draft from your sitemap — the part that still takes work is going back in and replacing generic descriptions with specific ones, and cutting the pages that don't belong.

A pattern that recurs, not a clean trend

This isn't the first edition to flag something like it. June noted roughly 30% of adopters published "a minimal file: sometimes a single line, sometimes a stub with no URLs"; July put it at "27.3% of llms.txt adopters publish a minimal stub." August's 52% is measured slightly differently — a direct complete-vs-stub split, not the softer "minimal file" language used earlier — so this isn't a clean three-point trend line. What's consistent: a large share of "adopters" haven't shipped a functional file, every month this has been measured.

What this does and doesn't show

This is a description of file completeness, not a claim about rankings or citations. No AI vendor has confirmed that a complete llms.txt file increases how often a site gets cited — this is an implementation-quality finding, not a performance one. The August sample is also self-selected: 282 domains whose owners chose to run a GeoReady audit, not a random sample of the web, so these shares likely don't generalize beyond it. Separately, the dataset does show a correlation between having any llms.txt file and a higher average GEO score — a different question, covered with its own honesty caveats in Does llms.txt Improve Your GEO Score? .

Find out what your own llms.txt looks like

Run the free audit to see whether your file is missing, a placeholder, or complete — alongside your other AI-visibility signals.

Run the free AI SEO audit · Build a real llms.txt