Module new 6 min
llms.txt
A proposed /llms.txt manifest that hands LLMs a curated, markdown map of your key content: promising, but unproven.
Why it mattersCheap to add, but do not expect citations from it yet; no major engine has confirmed using it.
Definition & Foundation
A low-cost bet, honestly labelled
llms.txt is a proposed root-level file (at /llms.txt) that hands an LLM a curated, markdown map of your important content: links to clean markdown versions of your key pages, stripped of nav, ads, and scripts, so a model can grab your content without wading through page furniture. The idea is elegant: a machine-readable table of contents, the way robots.txt is a machine-readable access policy.
The honest caveat is the whole story here: no major engine has confirmed it reads llms.txt. Over 844,000 sites had added one by late 2025, yet zero major LLM providers have committed to using it, Google has publicly dismissed it (comparing it to the old keywords meta tag), and no controlled study has shown a citation lift. This node exists so you make an informed decision: it is cheap to add and does no harm, but it is a speculative bet, not a strategy, which is exactly why it carries no impact score.
What It Is, and What It Isn't
The mechanism, and the adoption reality beside it
The manifest
A markdown file listing your key pages with short descriptions and links, optionally pointing to .md versions of each page. Think "sitemap for meaning": a curated shortlist of what matters, in the format a model reads most cleanly.
The clean-markdown promise
Its real value proposition is delivering your content without the noise: no nav, no cookie banner, no JavaScript. If an agent ever does read it, it gets your substance directly. That "if" is doing a lot of work.
The adoption reality
844k+ sites, 0 confirmed major-engine consumers, one public rejection from Google, and no measured citation lift. High adoption driven by low cost and FOMO, not by evidence that it works. Treat popularity as noise, not signal.
Where it fits the standards race
llms.txt is the informal end of a spectrum that includes Google's Open Knowledge Format and Microsoft's NLWeb. It gestured at "markdown for machines" first; newer, vendor-backed specs are formalizing the idea; see the standards race.
Myths vs Reality
Common misreadings, corrected
Putting It to Work
Add it cheaply, expect nothing, keep it honest
If you add llms.txt, do it as a low-effort hedge with correct expectations, and never let it substitute for the access, rendering, and content work that actually earns citations.
The low-cost-hedge playbook
Do the proven work first
Confirm crawler access, server-side rendering, and schema are solid. llms.txt is only worth minutes once the things engines demonstrably use are already in place.
Author a curated, honest manifest
List your genuinely important pages (docs, key guides, canonical facts) with a one-line description each, most important first. Curate ruthlessly; a manifest listing all 4,000 URLs is noise. Keep it to what you'd want an agent to read if it read anything.
Serve it at the root and keep it current
Place the file at /llms.txt (and optional .md page versions). Wire it into your build so it doesn't drift out of date; a stale manifest pointing at moved pages is worse than none.
Set expectations in writing
Note in your GEO docs that llms.txt is an unproven, no-confirmed-consumer bet. This stops a future teammate from over-investing, and stops leadership from expecting citations it can't yet produce.
Watch the standards race, not the hype
The formats worth real investment will be the ones engines commit to. Track OKF, NLWeb, and any official engine guidance; graduate effort toward whichever standard a major engine actually adopts.
Verification Checks
How to know it's really done
You've handled llms.txt correctly when it's cheap, current, and honestly framed. Click a check to mark it verified:
Works Together With
The nodes this one leans on
Tools for this
Go deeper from Technical GEO
Voices to follow
- Michael King · iPullRank The deepest technical explanations of how AI retrieval and ranking actually work.
- Crystal Carter · Wix Structured data and AI-search education with concrete implementation examples.
- Aleyda Solis · Orainti Practical frameworks, checklists and free learning roadmaps for search & AI search.
Canonical reads
Always current
These links resolve live: what you get is generated or filtered the moment you click.