How Cayve learns
Your corrections, signals shared across sites, and the scrubbed copies that improve detection.
Cayve gets better in three ways, each with its own safeguards.
1. Your corrections, on your site, straight away
When you mark a message, Cayve remembers the sender and the message on that site:
- It’s spam (dashboard) or Spam (WordPress): that sender and that message fingerprint go on the site’s own block list.
- It’s a real message (dashboard), Release and send, Approve or Not spam (WordPress): they go on the site’s allow list, so the sender is welcome from then on.
The next message from the same sender is decided with that in mind. Reasons say so: “You marked a message with the same email address as spam before.”
2. What the network has seen
When the same message or sender shows up on many Cayve sites in a day, or other Cayve accounts have marked it as spam, it counts against that message on your site too. Reasons name it: “The same message was sent to 31 other websites in the last day.”
These signals are built from one-way fingerprints, not message text. Labels only count across the network when at least two or three separate accounts agree, so one account’s clicks can’t change what happens on everyone else’s site.
3. Better detection over time
Unless an owner or admin switches off Help Cayve learn (Account → Settings), Cayve keeps a scrubbed copy of each message for up to 12 months to improve detection. Before it’s stored:
- the sender’s name, email addresses, phone numbers, links and address fields are removed, and
- the copy is encrypted.
Switching Help Cayve learn off deletes your account’s copies. Copies are never shown to other customers.
Something unclear or missing? Email hello@cayve.co.