# llms.txt editor

> Paste or write an llms.txt and see, as you type, every audit that can be decided from the file alone, the audits that need a scan of your live site, and the token budget.

This is the markdown version of https://llmsrank.dev/tools/llms-txt. The tool is beta. It runs in the browser: the text you paste is parsed and scored there, and nothing is fetched or sent unless you use the optional steps below (compare with your live file, check again, score a draft).

Only the audits a file can decide by itself are scored. The ones that need your live site (the HTTP response, robots.txt, the sitemap, your pages) are listed as "Measured by a scan" with the reason, and are not guessed. The editor does not add the audits into one number, because a number built from part of the audits would not match what a scan reports. The token bar marks 8 000 and 25 000 tokens (bytes divided by 4, an estimate), the thresholds of the token budget audit.

As of September 2026 no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them. The audits check structure and content against a published rubric; they don’t promise any ranking or citation effect.

## Audits it can score from the file

The editor lists every audit of the rubric; each links to its page under /docs/audits.

- [Covering llms.txt](https://llmsrank.dev/docs/audits/covering-llms-txt.md): Publish your llms.txt at https://your-domain/llms.txt on the production host (or at a section’s path, such as /docs/llms.txt), answering 200 to anyone.
- [Single H1](https://llmsrank.dev/docs/audits/single-h1.md): Start the file with exactly one # heading that names your site or organization, and use ## for every other heading.
- [Summary blockquote](https://llmsrank.dev/docs/audits/summary-blockquote.md): Put a one- or two-sentence > blockquote directly under the H1 saying who you are and what the file covers.
- [H2 sections](https://llmsrank.dev/docs/audits/h2-sections.md): Group your links under two or more ## sections other than ## Optional (Docs, API, Pricing), each holding a list of links.
- [Link line format](https://llmsrank.dev/docs/audits/link-line-format.md): Write every link as - [name](url): description, with a description of at least two characters after the colon.
- [Signal-to-noise](https://llmsrank.dev/docs/audits/signal-to-noise.md): Keep the file to the pages that answer questions, 92 links or fewer; cut tag, pagination, cart and login links first.
- [Link descriptions](https://llmsrank.dev/docs/audits/link-descriptions.md): Give every link a description after the colon that says what the page answers, averaging 70 characters or more.
- [Ordering](https://llmsrank.dev/docs/audits/ordering.md): Organize the file into three or more ## sections (not counting ## Optional), with the content people ask about most first.
- [Token budget](https://llmsrank.dev/docs/audits/token-budget.md): Keep the file under about 8,000 tokens (roughly 32,000 bytes): trim low-value links and move long reference text into linked pages.

## Diff, check again and score a draft

- **llms.txt.** The proposed file is shown against your live one as a line diff, with Copy and Download for the proposed file. After you publish, “Check again” fetches only your live llms.txt again (through our server, as LLMSRank-User; it counts as one of your free scans) and says what changed since the last fetch and whether it now matches.

“Score this draft” re-scores your site’s latest scan with the draft in place of the live file and shows the scan’s score with and without it, any cap, and every audit that would change. Nothing is published and the draft is not fetched; it is sent to our server only to be scored and is not kept. Audits that depend on pages the scan fetched for your live file’s links are kept from the scan and named. Limits: llms.txt 1 MB; 100 drafts a day, shared with the robots.txt fetch.

## Section skeletons

Each skeleton is fixed text with bracketed placeholders and guidance comments; replace every placeholder and delete the comments before you publish.

### ## For AI agents

Tells an agent reading the file who it is for, what it covers, what it doesn’t, which section answers which question, and when to offer the user an action page. It helps Role and scope, whose check looks for agent-facing wording such as the heading “For AI agents”, not for the scope statements themselves; it gives Intent routing partial credit when the file has no routing lines; and the action line and the “When a user wants to …” line help Action endpoints. As of September 2026, no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them: this section helps the agents and tools someone points at your file, and nothing documents it as a ranking or citation signal.

```markdown
## For AI agents

<!-- Replace every bracketed placeholder, delete the lines that don’t apply, then delete these comments. -->
<!-- Who the file is for, what it covers and what it doesn’t. -->
This file describes [Your organisation's name] and [what it offers, in a few words]. Use it to answer questions about [the topics this site is the source for].

Out of scope: [topics this site does not cover]. For those, [where to send the user instead].

<!-- Which section answers which kind of question. -->
- Questions about [a topic]: see the [Section name] section.
- Questions about [another topic]: see the [Another section name] section.

<!-- Things a user can do on the site (book, get a quote, apply, buy, get in touch), and when to offer each. Keep only pages that exist. -->
- [Action page title]([Full URL of the action page]): [What the user can do there]

When a user wants to [what the user wants to do], point them to [Action page title]([Full URL of the action page]).

```

### ## Data freshness

Tells an agent how current each kind of fact in the file is, how to caveat it, and which facts it must never confirm as current, so it doesn’t repeat an old price or rate as today’s. It helps Data-freshness tiers, whose check reads the words “Data freshness” (or “volatile”) and a “Never confirm …” rule, not the tiers themselves, and the “Last updated:” line helps Freshness signals. Keep the date current: a stale date tells agents the file is stale. As of September 2026, no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them: this section helps the agents and tools someone points at your file, and nothing documents it as a ranking or citation signal.

```markdown
## Data freshness

<!-- Replace every bracketed placeholder, delete the lines that don’t apply, then delete these comments. -->
<!-- How often each kind of fact in this file changes, most changeable first, and how an agent should caveat it. -->
Last updated: [YYYY-MM-DD]

How current each kind of information in this file is, from most to least volatile:

- [Kind of information that changes often]: changes [how often]. Say "[caveat wording]" and check [URL of the page with the current value].
- [Kind of information that changes sometimes]: changes [how often]. Say "[caveat wording]".
- [Kind of information that rarely changes]: stable; changes only when [what makes it change].

Never confirm [facts that must not be stated as current] as current. Give the date of the source and link [URL of the page with the current value].

```

### ## FAQ

Gives an agent short answers to the questions customers actually ask, in your words, each with the page that has the detail. It helps FAQ blocks, whose check reads the section heading (“FAQ”, “FAQs” or “frequently asked”), not the answers: the answers are for the agent and the user, and only you can check they are right and current. As of September 2026, no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them: this section helps the agents and tools someone points at your file, and nothing documents it as a ranking or citation signal.

```markdown
## FAQ

<!-- Replace every bracketed placeholder, delete the lines that don’t apply, then delete these comments. -->
<!-- The questions customers actually ask, each answered in one or two sentences, with the page that has the detail. Date anything that changes. -->
- **[A question customers ask]** [The answer in one or two sentences.] See [URL of the page with the detail].
- **[Another question customers ask]** [The answer in one or two sentences.] See [URL of the page with the detail].
- **[A third question customers ask]** [The answer in one or two sentences.] See [URL of the page with the detail].

```

### ## Identity

States who you are as labelled facts (legal name, brand, other names, identifiers) and defines the terms you use in a specific way, so an agent doesn’t confuse you with a similarly named organisation or misread your terms. It helps Entity identity: the check reads the labels as written (“Legal name:”, “Also known as:”, “Stock ticker:”, a registration number, a Wikidata ID) and a Glossary heading, so keep the labels and delete the lines you have no value for. As of September 2026, no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them: this section helps the agents and tools someone points at your file, and nothing documents it as a ranking or citation signal.

```markdown
## Identity

<!-- Replace every bracketed placeholder, delete the lines that don’t apply, then delete these comments. -->
<!-- Labelled facts about who you are, so an agent doesn’t mix you up with a similarly named organisation. Keep the labels as written. -->
- Legal name: [Your organisation's legal name]
- Brand: [The name customers know you by]
- Also known as: [Former names, abbreviations and common misspellings]
- Primary domain: [Your main web address]
- Parent company: [Your parent organisation's legal name]
- Stock ticker: [Exchange and ticker]
- Registration number: [Registration number and where it is registered]
- Wikidata: [Wikidata item URL]

### Glossary

<!-- Terms you use in a specific way, each defined in one line. -->
- [A term you use]: [What it means on this site]
- [Another term you use]: [What it means on this site]

```

### ## Intent routing

Maps each thing people come to your site to do or find out to the page that answers it, phrased from the user’s side, so an agent can go straight to the right page. It helps Intent routing, whose check looks for routing wording such as “When a user asks”, not for how many questions you cover; it applies once the file has two or more ## sections. As of September 2026, no major AI assistant vendor documents reading other sites’ llms.txt files, and Google documents that Google Search ignores them: this section helps the agents and tools someone points at your file, and nothing documents it as a ranking or citation signal.

```markdown
## Intent routing

<!-- Replace every bracketed placeholder, delete the lines that don’t apply, then delete these comments. -->
<!-- One line per thing people come to the site to do or find out, in their words, each pointing to the page that answers it. Cover the common questions, not only the ones you want to promote. -->
- When a user asks [a question or task, in the user's words] → [Page title]([Full URL of the page]): [What the page answers]
- When a user asks [another question or task] → [Another page title]([Full URL of another page]): [What the page answers]
- When a user asks [a third question or task] → [A third page title]([Full URL of a third page]): [What the page answers]

```