In short: Declare your industry with an Industry: line, then write each blocking and high-severity guardrail of its playbook as an instruction to AI agents.
Category: Compliance & Trust · weight 30 of 100 within the category
Compliance & Trust is worth 20 in regulated, 10 in generic and commerce, 5 in technical (points of 100, by scoring profile)
Audit ID P1 (as shown on reports) · rubric v2.1
What it checks
Whether the file carries the guardrails of the industry playbook you declared, rule by rule.
Why it matters
An agent speaking for a regulated business can create real liability. Guardrails in the file tell it which lines it must not cross.
When it applies
Opt-in (rubric v2.1). Applies only when your llms.txt declares an industry with an Industry: line (before the first section, in ## Identity, or in a section whose heading starts with “Compliance”) and that industry’s playbook is published with reviewer-signed rules. Up to three industries; values we don’t know are reported, not scored. Without a declaration, or while a declared playbook awaits sign-off, the whole Compliance & Trust category is Not applicable and its weight is redistributed: not declaring never costs points, and we never grade you against rules nobody has signed off.
An audit that doesn't apply, or that we couldn't verify, is removed from your score and its weight shared among the Compliance & Trust audits that do apply. It is never scored as a zero.
How it's scored
The severity-weighted share of the declared playbook’s signed-off blocking and high-severity guardrails found in the file (a blocking rule counts twice a high one). Not applicable when the playbook has no such guardrail. The evidence line names the rules missing, by short name and severity.
How to fix
Add Industry: <your playbook> to the file (for example, Industry: housing).
Under ## Compliance rules for AI agents, write each blocking and high-severity guardrail of that playbook as its own list item.
Use your own words if you like: the matcher looks for the rule’s subject and what to do, not our wording.
Example
## Compliance rules for AI agents
- Never say anything that indicates a home or community is preferred for, or not available to, people because of race, color, religion, sex, disability, familial status or national origin.
Frequently asked questions
Why is Regulatory guardrails marked Not applicable for my site?
Compliance & Trust is opt-in since rubric v2.1. It applies when your llms.txt declares an industry with an Industry: line (before the first section, in ## Identity, or in a section whose heading starts with “Compliance”, for example “Industry: housing”) and that industry’s playbook has been published with rules a reviewer has signed off. Without a declaration the whole category is Not applicable and its weight is shared out across the other categories, so not declaring never costs points. Today every playbook is still marked “to verify”, awaiting a reviewer’s sign-off, so a declaration is recorded and shown on your report but no rule is scored yet: the category stays Not applicable until your playbook is published. We never grade a site against rules nobody has signed off.
How do I declare my industry, and which values are accepted?
Add one line such as “Industry: housing” (or “> Industry: Housing, General business” in the summary). Use a playbook’s name or slug, such as housing, lending, health, insurance, securities, banking or general; up to three, separated by commas. Values we don’t know are listed on your report and never scored, and a declaration elsewhere in the file (under an unrelated section, or in a code block) isn’t read.
What exactly does the Regulatory guardrails audit look for in my llms.txt?
The severity-weighted share of your declared playbook’s signed-off blocking and high-severity guardrails that your llms.txt or AGENTS.md states (a blocking rule counts twice a high one). Each rule has its own matcher: an instruction counts when one list item, table row or sentence addressed to AI agents names the rule’s subject and what to do about it. Your own words are fine. Each instruction counts for one rule only, so repeating a line, joining several rules into one sentence or listing keywords earns nothing. The evidence line lists the rules missing, by short name and severity.
What does a good guardrail look like for a homebuilder or real-estate site?
Name the specific line an agent must not cross, for example: “Never say anything that indicates a home or community is preferred for, or not available to, people because of race, color, religion, sex, disability, familial status or national origin.” The Fair Housing Act makes it unlawful to publish any notice, statement or advertisement about selling or renting a dwelling that indicates such a preference; an agent answering questions about your homes should be held to the same line.
In your research, how many sites had regulatory guardrails in their llms.txt?
In our research (187 sites, corpus captured 2026-10-10, scored with rubric v2.1), one file names an industry — att.com writes “Telecommunications” — and no playbook covers it, so this audit still applies to none of them. Under rubric v2.0 it applied to 7 homebuilders and brokerages matched by a domain list, which v2.1 retired. Under v2.0, 1 of those 7 earned full marks on the old phrase check and 6 scored zero.