Plan requirement
| Subscription | Suite Growth or higher |
| Also required | Copilot |
| Access | Admin |
Generative search answers the question rather than listing articles. Powerful, and only as good as what it reads.
What changes for the reader
Instead of ten results to work through, they get an answer, with the articles it came from. For a specific question that is a considerable improvement over reading three articles to assemble it yourself.
What it depends on
Your articles. It does not know anything else, and it cannot be better than what it summarises.
So the preparation is the same as for an AI agent: one subject per article, the answer first, no contradictions, nothing out of date.
Where it goes wrong
- Contradictory articles. It may blend them, producing an answer that is in no article.
- Conditional information whose conditions are unstated. It applies them to everyone.
- Old content, quoted with the same confidence as current content.
Watch what it says
Ask it your most common questions and read the answers as a customer would. That is the only real test, and it points straight at the articles that need work.
Usage is metered
Generative search consumes part of your AI allowance, so it has a cost that grows with traffic. Worth watching alongside the quality rather than only switching it on.
Comments
0 comments
Article is closed for comments.