Compare
What each level of autonomy actually does.
These are not pricing tiers. They are control settings, applied per site and per action type — a site can be on Autopilot for metadata and Watch for everything else, which is the expected starting position.
The autonomy model is designed and stored in the schema. No level executes anything today, because nothing writes to a site yet.
| Capability | Watch | Recommend | Assist | Autopilot | Full |
|---|---|---|---|---|---|
| Monitors your search data | ● | ● | ● | ● | ● |
| Detects findings and ranks them | ● | ● | ● | ● | ● |
| Explains the reasoning and evidence | ● | ● | ● | ● | ● |
| Alerts you to damaging changes | ● | ● | ● | ● | ● |
| Writes recommendations | — | ● | ● | ● | ● |
| Prepares the exact change as a diff | — | — | ● | ● | ● |
| Captures a snapshot before acting | — | — | ● | ● | ● |
| Applies low-risk changes unattended | — | — | — | ● | ● |
| Applies medium-risk changes unattended | — | — | — | Policy | ● |
| Applies high-risk changes unattended | — | — | — | — | Policy |
| Requires approval for irreversible changes | n/a | n/a | ● | ● | ● |
| Respects protected pages | ● | ● | ● | ● | ● |
| Verifies the change landed | — | — | ● | ● | ● |
| One-click rollback | — | — | ● | ● | ● |
Policy means the behaviour depends on the risk bands you have chosen to allow for that action type. Changes that cannot be cleanly reversed always require a human, at every level.
The point
Levels are not a paywall.
Nothing here is withheld to sell you an upgrade. Autonomy is a safety control, and the right setting is whatever you are comfortable with on a given site for a given kind of change — which for most people starts at the cautious end and moves as the system earns it.
Find out what's holding your site back.
Free during early access. Connect Google Search Console and see your own list in minutes.