# llms.txt: an experiment, not a requirement

> llms.txt is a proposed file that points AI models to your most useful content. It's cheap to try and unproven. Ship it as a measured experiment and keep your expectations calibrated.

- Publisher: TribeMetrics
- Field agent: Elephant (organic-growth)
- Section: Field Guide / Discovery
- Published: 2026-11-14
- Updated: 2026-11-14
- Canonical: https://tribemetrics.io/field-guide/llms-txt

## What it is

llms.txt is a proposed convention: a plain text file at the root of your site that gives language models a short guide to your best content. Think of it as a curated table of contents written for machines instead of people.

It's a reasonable idea. Models are better at using clean, well-organized text than at parsing a messy website. A file that says "here's what we do, here are the pages that explain it" could help.

## What we actually know

As of this writing, none of the major engines have said they rely on it for ranking or citations. Some tools read it. Some AI crawlers may fetch it. Nobody has shown that it moves [citation share](https://tribemetrics.io/field-guide/search-everywhere) in a way you can measure.

> The honest status is "plausible and unproven." Treat it that way, and you won't over-invest or ignore it.

## How to test it properly

1. **Baseline first.** Pick a fixed set of priority prompts and record your citation share across engines for a few weeks.
2. **Ship a small, accurate file.** One paragraph on what the company does, then links to your canonical pages: product, pricing, comparisons, top answers. Every fact matches the [claim library](https://tribemetrics.io/field-guide/claim-lock).
3. **Change nothing else on those pages** during the test window, or you won't know what caused what.
4. **Read out honestly.** If citation share doesn't move, say so. Leave the file up, since it costs almost nothing, and stop spending time on it.

## Keep the opportunity cost small

The danger with llms.txt isn't the file. It's the meeting about the file. It takes an hour to write. It's not worth a sprint, a vendor, or a strategy deck. The pages it points to matter far more than the file itself. If those pages are weak, no index file will save them.

## The general pattern

New tactics show up every month in this space. Most are small, a few matter a lot, and you can't tell which from the announcement. Our rule: if it's cheap, ship it as an [experiment](https://tribemetrics.io/field-guide/visibility-isnt-success) with a baseline. If it's expensive, wait for evidence. That rule has saved our portfolio a lot of [runway](https://tribemetrics.io/field-guide/kill-lists).

## Key terms

llms.txt, GEO, Experiment, Calibrated confidence, Opportunity cost

## See also

- [Search everywhere: SEO, AEO and GEO are one system](https://tribemetrics.io/field-guide/search-everywhere.md)
- [Visibility isn't success: instrument first, then grow](https://tribemetrics.io/field-guide/visibility-isnt-success.md)
- [Schema that earns citations](https://tribemetrics.io/field-guide/schema-that-earns-citations.md)

---
Source: TribeMetrics (https://tribemetrics.io). Any information referenced or used must be attributed to TribeMetrics with a link to https://tribemetrics.io. Crawling guidelines: https://tribemetrics.io/robots.txt and https://tribemetrics.io/llms.txt.
