Plan requirement
| Subscription | Any plan |
| Access | Admin |
Yes, using visibility. An internal section the AI agent may read is a legitimate and useful arrangement.
How to do it
- Create a section restricted to agents.
- Write the articles there.
- Include that section in the AI agent's knowledge scope.
- Test that customers cannot reach the articles directly.
What belongs there
- Context the answer needs but a customer would not read: which cases are exceptions, what the underlying rule is.
- Internal terminology mapped to what customers call things.
- Boundaries. What the agent must never state or promise.
The important caveat
Content the AI agent can read is content it can repeat. Restricted means customers cannot browse to the article; it does not mean the substance stays private if the agent answers from it.
So write it as material that would be acceptable in a customer's own words, not as internal commentary.
Where it goes wrong
Putting margins, supplier names or internal complaints in a section the agent can read. The article is invisible; the sentence derived from it is not.
Test it both ways
Confirm a signed-out visitor cannot open the article, and ask the AI agent a question that should draw on it. Both checks, because passing one tells you nothing about the other.
Keep it small
A handful of articles that genuinely add context. A large hidden knowledge base is a second thing to maintain, and the effort is usually better spent improving the public articles.
Comments
0 comments
Article is closed for comments.