SilentShield

AI agent control

See which AI agents visit your site, and decide what they may do

AI search, AI agents and AI training are three different decisions. SilentShield shows who comes, whether they are who they claim to be, and lets you set a rule per category.

Start free All features
The agent report in the dashboard with counts of requests, verified, spoofed and unsigned visits, agent traffic over time and a pie chart of verified versus unsigned visits
The agent report: who visited, and how many were verified. Screenshot of the English dashboard.

SilentShield's AI agent control records which AI crawlers and agents visit your pages, sorts every visit into AI search, AI agents or AI training, and shows whether the visitor is verified, unsigned or only uses the name of a known bot. You set a rule per category — allow, watch or block — with a recommendation for each. Recording works without blocking anything; enforcing a rule on your server is a second, separate step.

From the first visit to your rule

  1. Observe

    A small observer on your server reports which AI services read which pages. It blocks nothing and changes nothing on your site.

  2. Sort

    Every visit is sorted into AI search, AI agents or AI training.

  3. Verify

    Each visit is marked verified, unsigned or spoofed. Someone merely using a bot's name shows up as its own number.

  4. Decide

    Allow, watch or block each category — with quick presets from maximum visibility to maximum protection.

Blocking AI search also removes you from AI answers: that is a visibility decision, not only a security one. Enforcing rules on your server is a separate step; the components are currently provided on request during early access.

What it gets you

See before you decide

Most site owners never see which AI services read their content. The report shows it, with request counts per agent.

Spoofing is its own number

Verified bots, unsigned bots and visits that only claim a bot's name are counted separately, so a fake never makes the real bot look bad.

A rule per category

AI search, AI agents and AI training are three decisions. Each can be allowed, watched or blocked, with a recommendation so you do not cut yourself out of AI answers by accident.

Observe first, enforce later

Run observation for a week or two, read the report, and only then block — the same staged rollout as with DMARC.

Also part of it

Signed requests

Verification uses signed requests (Web Bot Auth, RFC 9421) as well as detection by name.

Not tied to WordPress

Observers are available for WordPress, Next.js, Node.js, PHP and Go.

Hashes instead of addresses

IP address and user agent are hashed on your server before they are reported.

Data protection

The observer sets no cookies and reports only hashes of the IP address and the user agent, not the values themselves. It runs on your server and never changes what your human visitors see.

Questions and answers

Does observing block anything?

No. Observation only records which AI services visit; it never changes the response.

What does “spoofed” mean?

A visit that uses the name of a known bot without being that bot. It is counted separately from verified visits.

Can I block AI training and stay visible in AI search?

Yes. The three categories are separate decisions, and “allow search, watch agents, block training” is one of the quick presets.

How do I enforce a block?

Enforcement runs on your server or at the edge as a second step. The components are currently provided on request during early access; contact support.

Find out who is reading your site

Start free and set up observation in a few minutes.

Start free

Technical details

Setup and reference are in the documentation: