Skip to main content

Preferences & Tuning

MailPrism's AI works out of the box, but you can shape how cautious or hands-off it is. The controls live in two places in Settings: Rule Defaults (how AI acts on your mail) and Privacy & Activity (what AI is allowed to learn and process).

AI is opt-in

None of this applies until you turn AI on. See AI privacy & consent for how consent works.

How sure the AI has to be

In Settings → Rule Defaults → AI Configuration, the AI confidence threshold sets the minimum confidence an AI signal needs before a rule acts on it.

  • Higher (e.g. 90%+) — fewer false positives, but the AI may skip borderline mail.
  • Lower (e.g. 60–70%) — catches more, at the cost of the odd wrong call.
  • 85% is the recommended starting point for most people.

There's a second, related dial in Settings → Privacy & Activity → Advanced Settings: Minimum confidence, which controls how confident a detected pattern must be before MailPrism suggests it as a rule.

Let the AI act — or just watch

By default MailPrism is cautious. Two switches in Rule Defaults → Safety Settings decide how much it does on its own:

SettingWhat it does
Manual AI label approvalWhen on, AI-suggested labels are logged but not applied automatically — you stay in the loop. Turn it off to let AI apply labels without confirmation.
Dry run modeRules evaluate and log what they would do, but take no action. Perfect for testing a setup before it touches real mail.
Try dry-run first

Turning on Dry run mode is the safest way to trial new automations. Watch the Rule Logs to see exactly what would have happened, then switch it off once you trust the results.

Stay informed

Also in Rule Defaults → Safety Settings:

  • Notify on rule execution — get a notification whenever a rule fires. Handy while you're building trust in your automations; easy to switch off once they're humming along.

What the AI is allowed to learn

In Settings → Privacy & Activity, you control how much MailPrism personalizes to you:

SettingWhat it does
Pattern LearningThe master switch for learning from your habits. With it on, you can choose what to watch — archives, labels, stars, and (optionally) deletes.
AI PersonalizationLets AI use those learned patterns to make personalized suggestions.
Auto-Suggest RulesSurfaces new rules based on patterns it spots.

Turning Pattern Learning off stops MailPrism from learning new patterns entirely. For the full picture — including the activity log, data export, and deletion — see Privacy & Security.

Keep results consistent

A few habits keep AI behaving predictably as your mail changes:

  • Raise the confidence threshold if a smart rule is acting on mail it shouldn't.
  • Pair AI signals with plain conditions — see Best practices.
  • Check the Rule Logs after a change; they capture the AI's reasoning so you can see why a rule fired.

→ Next: Writing profiles · Smart rules