Where should the Content-Signal line go?

Inside the robots.txt group it applies to — usually User-agent: *. A line before the first User-agent line applies to no group, and the checker warns about it.

What does ai-input mean?

Using content as input to an AI answer — retrieval-augmented generation, grounding, or AI search summaries. ai-train covers training or fine-tuning models; search covers ordinary search results.

Do I need Content Signals to appear in Google's AI features?

No. Google says no special files or markup are needed for AI Overviews or AI Mode; Google-Extended and snippet controls are its documented controls.

Is a Content-Signal HTTP header enough?

Cloudflare documents Content Signals in robots.txt, not as a header. The checker still reports a header if you send one.

Free tools

Content Signals checker

Enter a domain to read the Content Signals its robots.txt declares — Cloudflare's search, ai-input and ai-train preferences — for each User-agent group, with the line each value comes from.

Runs the same real audit as the full report and opens it at this check — free, with no account or installation.

What this checks

Reads every Content-Signal line in the live robots.txt and assigns it to its User-agent group.
Reports search, ai-input and ai-train as yes, no, unspecified, or ambiguous when one group declares both yes and no.
Warns about a line outside any group, values other than yes or no, malformed pairs and repeated keys; notes unknown keys without interpreting them.
Reports Cloudflare's testing-stage use extension (immediate, reference, full) without treating it as a core signal.
Shows a homepage Content-Signal response header if one is sent, and compares the AI-training signal with any RSL licence.

Worked example

Cloudflare's managed robots.txt (documented example)

robots.txt

User-Agent: *
Content-signal: search=yes, ai-train=no, use=reference
Allow: /

What the report shows

Example results
search	yesline 2
ai-input	unspecifiedneither granted nor restricted
ai-train	noline 2
use	referencetesting-stage extension, reported only

How to read the result

yes / no: The publisher grants or restricts that use through Content Signals.
Unspecified: No value for that signal: the publisher neither grants nor restricts it here. It is not a block and not consent.
Ambiguous: The same group declares both yes and no; a reader cannot tell which applies.

What this does not tell you

Content Signals state preferences about use. They do not stop crawling, and not every crawler reads them.
A robots.txt Allow is not consent to training, and a Disallow is not a licence term — CrawlPact never infers one from the other.
Content Signals is a Cloudflare proposal (September 2025), not a ratified standard.

See limitations andmethodology.

Questions

Where should the Content-Signal line go?: Inside the robots.txt group it applies to — usually User-agent: *. A line before the first User-agent line applies to no group, and the checker warns about it.
What does ai-input mean?: Using content as input to an AI answer — retrieval-augmented generation, grounding, or AI search summaries. ai-train covers training or fine-tuning models; search covers ordinary search results.
Do I need Content Signals to appear in Google's AI features?: No. Google says no special files or markup are needed for AI Overviews or AI Mode; Google-Extended and snippet controls are its documented controls.
Is a Content-Signal HTTP header enough?: Cloudflare documents Content Signals in robots.txt, not as a header. The checker still reports a header if you send one.

Sources

Want the complete picture — and to know when it changes?

This check is one section of the full audit, which also covers every other policy signal and a Policy Health Score. Save the domain from the report to monitor it.

Run the full audit How monitoring works View a sample report