Intent routing

Intent-routing table

In short: Add an intent-routing list that sends each common question to the page that answers it, phrased “When a user asks … → [Page](url)”.

Category: Agent Readiness · weight 20 of 100 within the category
Agent Readiness is worth 30 in technical, 25 in generic and commerce, 20 in regulated (points of 100, by scoring profile)
Audit ID A3 (as shown on reports) · rubric v2.1

What it checks

Whether the file routes each kind of user question to the right page.

Why it matters

A routing table turns the file from a list into a map: the agent matches the user’s intent and goes straight to the page that answers it.

When it applies

Applies when the file has two or more ## sections, which the scanner reads as a site serving at least two distinct intents. A file with one section or none is treated as single-purpose and is Not applicable.

An audit that doesn't apply, or that we couldn't verify, is removed from your score and its weight shared among the Agent Readiness audits that do apply. It is never scored as a zero.

How it's scored

1.0 when the file routes by intent in words: “intent” and “route” or “routing” on the same line, or phrases such as “when a user asks”, “user says” or “buyer says” · 0.4 without that but with agent-facing framing (see Role and scope) · 0 otherwise.

How to fix

List the main things people come to your site to do or learn.
For each, name the page (and section) that answers it, phrased from the user’s side: “When a user asks …”.
Cover the common intents for your industry, not only the ones you want to promote.

Example

## Intent routing

- When a user asks to compare plans → [Pricing](https://acme.com/pricing.md)
- When a user asks for a refund → [Refunds](https://acme.com/refunds.md)

Frequently asked questions

What is an intent-routing table in llms.txt?

A section that maps the things people come to your site to do to the page that answers each one, so an agent can match the user’s intent and go straight there. Write it as a list: “When a user asks how to get a refund, use Refunds”, one line per intent, covering the common intents for your industry and not only the ones you want to promote.

Sources: LLMSRank — Intent routing audit · llmstxt.org — The /llms.txt file (v2, 2026-08-10)

What wording does the intent-routing check look for?

Today it scores full marks when the file routes by intent in words: “intent” and “route/routing” on the same line, or phrases such as “user says”, “buyer says” or “when a user asks”. Without that, a file that has agent-facing framing (see Role and scope) scores 0.4 and any other file scores 0. Where a judged review of the file exists, its score replaces this phrase check.

Sources: LLMSRank — Intent routing audit · LLMSRank — How scoring works

Why is intent routing Not applicable to my file?

The rubric applies it only to sites serving at least two distinct user intents. Today the scanner infers that from the file having two or more H2 sections, so a file with one or no H2 sections is treated as single-purpose and the check is removed, with its weight redistributed within the category.

Sources: LLMSRank — Intent routing audit · LLMSRank — How scoring works

Isn’t a list of H2 sections already a routing table?

Not quite. H2 sections group links by topic, which llms.txt v2 expects; a routing table goes further and names the user’s question or goal, then the page that answers it. Agents are told to “view or search llms.txt to find the information they need”, and phrasing lines as intents makes that match direct.

Sources: llmstxt.org — The /llms.txt file (v2, 2026-08-10) · LLMSRank — Intent routing audit

In our research, how many llms.txt files route user intents?

Almost none. Across our 187-site corpus (captured 2026-10-10, scored with rubric v2.1), intent routing applied to 167 files and was Not applicable to 20; just 2 files earned full marks, and 75.4% scored zero.

Sources: LLMSRank — How scoring works · LLMSRank — Intent routing audit