AI content detection: why this rule is retired

What AI content detection is

AI content detection estimates whether a passage was written by a language model. Detectors work on statistical fingerprints of the text and they are not reliable at the level of a single page: human writing on technical subjects scores as machine-written often enough that acting on the result costs more than it saves.

Google’s guidance on the subject is about the content, not its production method. Its documentation on AI-generated content says the reward goes to content that is helpful and original, however it was made, and that using automation to game rankings is what its spam policies address.

What squirrel checks

Nothing. The rule is retired and hard-guarded so that it can never call a model. Its previous implementation made one model call per page with no credit charge wired to it, which was an uncatalogued cost, and it was replaced with a no-op rather than left running.

• ai-content, skipped with the message AI content detection is retired and the skip reason Retired pending a metered opt-in endpoint. This is the only result the rule can produce.

The rule is also marked disabled in its metadata, and the rule loader drops it before any enable or disable setting is read, so in this release it does not run at all and no configuration turns it on. Re-enabling detection for real requires a metered server endpoint and an opt-in before any model call.

How to fix it

There is nothing to fix. The check is a no-op and the skip carries no weight in the score.

If you want the checks that do read the content with a model, ai/llm-parsability scores how well a model can extract a page’s facts and ai/page-type-match compares what a page reads as against what its markup declares. Both are included in the audit base charge on a logged-in run.

Enable / disable

There is nothing to enable or disable: the rule never loads, so enable and disable entries for ai/ai-content do nothing and can be removed from your config.

Enable all AI rules

[rules]
enable = ["ai/*"]

Related rules

• ai/llm-parsability: how well a model can extract the page’s facts, which does run.
• ai/page-type-match: the classified page type against the declared structured data.
• content/quality: the LLM content quality score, also retired in this release.
• content/placeholder-text: unedited template text, which is a defect a rule can establish from evidence.

Agent experience findings ship in every audit next to the SEO, performance and security rules. See Generative engine optimization for how an agent works through a report.

References

• AI-generated content and Google Search
• Spam policies for Google web search

Check your site

Run squirrel audit https://example.com and open the Agent Experience section of the report. This rule reports a skip on every page. Signed-out local audits need no account and cost nothing. Signed-in audits spend credits.