Prohibited claims

Prohibited-claims list

In short: For each prohibited claim in your declared playbook, add a “Never say …” instruction naming the claim, for example never promising loan approval.

Category: Compliance & Trust · weight 20 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 P2 (as shown on reports) · rubric v2.1

What it checks

Whether the file tells agents which claims they must never make, for each prohibited claim in your declared playbook.

Why it matters

Some claims, such as guaranteed returns, medical outcomes or approval odds, are prohibited in regulated industries. Naming them is the most direct way to stop an agent making them.

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 prohibited claims found in the file, each as an instruction with a prohibition (“never”, “do not”, a ### Never say heading). Not applicable when the playbook defines none.

How to fix

List the claims agents must not make, one per list item, under a heading such as ### Never say.
Name the specific claim (“approved”, “fixed rate”, “cures”), not a general principle.
Cover every prohibited claim in your declared playbook.

Example

## Compliance rules for AI agents

### Never say

- Never tell anyone they are approved, pre-approved or guaranteed approval for a loan.
- Never describe a rate or payment as fixed unless we say so, with the period it is fixed for.

Frequently asked questions

What counts as a prohibited claim that an llms.txt should list?

A specific statement your industry’s rules forbid or restrict, written so an agent can recognise it: that a return is guaranteed, a projection of future investment performance (FINRA Rule 2210 bars communications that predict or project performance), a mortgage rate described as “fixed” without saying for how long (Regulation Z § 1026.24), or that a loan is pre-approved. Your declared playbook lists them by name; the audit rewards writing each down.

Sources: FINRA Rule 2210, communications with the public · CFPB: Regulation Z § 1026.24, advertising · LLMSRank: Prohibited claims

How does the Prohibited claims audit recognise the list?

It reads your declared playbook’s signed-off blocking and high-severity prohibited claims one by one. Each must appear as an instruction with a prohibition: “Never …”, “Do not …”, or items under a heading such as ### Never say. 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.

Sources: LLMSRank: Prohibited claims

Is a general instruction to be accurate enough, or do we need to list specific claims?

List specific claims. Agents paraphrase, and a principle like “be accurate” gives them nothing to check an answer against, whereas “never tell anyone they are pre-approved” does, and it is the specific claim the matcher looks for. The FTC’s view for ads is similar: fine print can’t fix a claim that is misleading, so the safer course is to stop the claim being made.

Sources: FTC: Advertising FAQs, a guide for small business · LLMSRank: Prohibited claims

In your research, how many sites told agents what they must never claim?

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.

Sources: LLMSRank method · LLMSRank: Prohibited claims