What the file is
llms.txt is a proposed convention: a markdown file at the root of your site that summarises what the site is and points to its most important pages, written for language models rather than browsers.
The reasoning is that a model arriving at a site has to work out what matters by crawling everything, whereas a curated summary could hand it the answer. It is a sensible idea in the same family as robots.txt and sitemap.xml.
The difference is adoption. robots.txt is honoured by every major operator. sitemap.xml is consumed by search engines. llms.txt is a proposal that some sites have implemented and no major engine requires.
Which engines consume it today
Honestly: no major AI search product documents llms.txt as a requirement or a ranking input, and none has committed to it as a standard. Some developer tools and smaller products read it. The large assistants your customers use retrieve your pages the normal way, as described in Lesson 5: build a local site AI can find.
Because status here changes, this page is reviewed monthly and the updated date at the top tells you when it was last checked. If that date is old, verify before acting on it.
A minimal example
If you decide to add one, keep it short and accurate:
# Example Refrigeration Services
> Commercial refrigeration repair and maintenance for cafes, restaurants and food
> retailers across Dandenong, Springvale and Noble Park. Trading since 2011.
## Services
- [Emergency repair](https://example.com/emergency-repair/): same-day attendance, callout $180 including first hour.
- [Scheduled maintenance](https://example.com/maintenance/): quarterly and six-monthly plans for commercial kitchens.
- [Coolroom installation](https://example.com/coolroom-installation/): supply and install for small food businesses.
## About
- [About us](https://example.com/about/): team, licences and service history.
- [Contact](https://example.com/contact/): phone, email and hours.
Two rules. Every URL must resolve, and every fact must match your profile, your pages and your structured data. An llms.txt that disagrees with the rest of your site is a new inconsistency, which is the exact failure mode Lesson 2 warns about.
The honest verdict
Adding it is harmless and cheap. Expecting it to change your AI visibility today is not supported by the evidence.
For a local business with limited hours, it ranks below every item in the 90-day checklist. It sits in the same category as a small optional refinement: fine to do once the foundation is solid, not a priority while your hours are wrong in two directories.
If you do add it, treat it as a maintenance item. A stale llms.txt listing a service you dropped is worse than no file.
What to do instead with that hour
If you have an hour and want it to count:
- Run the listing audit from the NAP consistency guide and fix the worst mismatch.
- Rewrite the pricing section of your main service page so it states a range instead of “contact us”.
- Reply to every unanswered review.
- Check your server logs to confirm AI crawlers are arriving and getting 200 responses.
- Set your crawler permissions deliberately using the robots.txt guide.
Each of those has an observable effect. llms.txt, for now, does not.
How it compares to what already works
It is worth placing llms.txt beside the files that do have adoption, because the comparison explains the verdict.
| File | Status | Consumed by | Effect if missing |
|---|---|---|---|
| robots.txt | Long-standing convention | All major crawlers | Crawl permissions default to open, and you lose control |
| sitemap.xml | Established standard | Search engines | Discovery is slower, especially for deep pages |
| Structured data (JSON-LD) | Established standard | Search engines and assistants | Entity ambiguity persists |
| llms.txt | Community proposal | Limited tooling today | Nothing observable |
The bottom row is not a criticism of the idea. Conventions start somewhere, and robots.txt itself was a proposal once. It is a statement about where your limited hours should go this quarter.
If you add one, maintain it
A stale llms.txt is worse than none, for the same reason a stale directory listing is: it is another source stating facts about your business, and if it disagrees with your site you have created exactly the contradiction Lesson 2 warns about.
So treat it as a maintenance item rather than a one-off. At each quarterly check, confirm every URL still resolves, every service listed is still offered, and the summary still describes the business you actually run today.
And keep expectations proportionate. If a tool or a consultant presents llms.txt as the reason your visibility will change, ask which engine consumes it and how they verified that. The honest answer, as of the updated date on this page, is that the foundation work in the 90-day checklist matters far more.
Who is genuinely served by it today
Three cases where writing one is reasonable rather than optional.
Documentation-heavy sites. If your site is largely reference material and developer tools are a real part of your audience, some of those tools do read it.
Sites with a complex structure. A curated map of what matters can help any system, human or machine, that arrives without context.
Businesses who simply want the tidiness. Writing an llms.txt forces you to state, in a few lines, what the business does and which pages matter. That exercise has value even if nothing reads the file.
For a single-location service business with eight pages, none of those apply strongly. Your structure is already obvious, your important pages are one click from the homepage, and the 90-day checklist has a dozen items with observable effects waiting ahead of it.