Skip to main content

Best Practices

A few habits keep your automations doing exactly what you expect — even as your rule list grows.

Start simple, then grow​

Begin each rule with one condition and one or two actions. Confirm it behaves the way you want, then add conditions or actions. Small rules are easier to reason about and easier to fix.

Test before you enable​

Don't wait for live email to find out a rule is too broad. Test it first — run it against your recent inbox to see what it would have matched, and check why a specific message did or didn't match. Tighten or loosen the conditions, then enable it.

Catch mistakes safely

Testing is read-only — it shows what would happen without taking any action. It's the fastest way to spot a rule that catches too much before it touches your mail. See Testing rules.

Name rules so future-you understands them​

A good name says what the rule does:

  • ✅ "Archive marketing newsletters"
  • ✅ "Star urgent emails from key clients"
  • ❌ "Rule 4"

Order matters: use priority and "stop processing"​

When several rules could touch the same email:

  • Give important rules a higher priority so they run first.
  • Turn on Stop processing for VIP or catch-all rules, so the right rule "wins" and others don't pile on.

A reliable pattern: a high-priority VIP rule (Stop processing on) at the top, then your general rules below it.

Combine AI with plain conditions​

AI is powerful but not perfect — sarcasm, edge cases, and very technical mail can be misread. The most reliable rules pair an AI signal with a concrete condition, so a single AI miss can't trigger something important on its own:

Category is urgent AND From relationship is established

The plain condition (here, the sender relationship) acts as a guardrail around the AI judgment. A few good combinations:

  • AI category + a known sender or domain — only act on AI's read when it's from someone you trust.
  • AI urgency + a subject keyword — confirm the AI's call with a hard signal.
  • "Is cold outreach" + sender is not a contact — avoid flagging people you know.

This is more dependable than relying on a single AI signal alone. See Conditions for every field you can combine, and AI features for how the AI signals work.

Be careful with actions that send​

Forwarding, sending, and auto-replies go out from your Gmail. For those:

  • Prefer Draft reply over Send reply while you build trust in a rule.
  • Add a delay and cancel if replied so follow-ups only go out when truly needed. See Scheduling & recurrence.

Check the logs​

After enabling a rule, glance at Analytics and the Rule Logs to confirm it's matching the right mail. If it's catching too much or too little, tighten or loosen its conditions. See Analytics.

Disable instead of delete (at first)​

If a rule misbehaves, turn it off rather than deleting it. You keep the setup and can re-enable it after a small fix.

Back up your setup

You can export your rules to JSON from the Rules page — handy before a big reorganization. See Import & export rules.