How ScamSam works
How ScamSam works
- 1 Paste the suspicious message into the box — a mail, a text, a chat, an offer. Or forward the mail to scamsam@grgmyr.com (registered accounts).
- 2 A language model reads it and looks for the familiar signs of a scam: spoofed senders, manufactured urgency, payment demands, credential harvesting, offers too good to be true.
- 3 The verdict comes back at once: scam, suspicious, legitimate or unclear — with the reasons and one piece of advice on what to do now.
- No account 3 checks a day on the free model.
- Account More checks, your choice of model (Claude models included), checking by mail.
- Extended Higher limits still and the strongest model — enabled by the operator.
What you submit is never stored. What ScamSam does keep, and for how short a time, is in the privacy policy.
The mail route
Registered accounts with a verified address can simply forward a suspicious mail to scamsam@grgmyr.com — best as an attachment, which keeps the original mail's headers so the check sees the real sender too. The answer comes back by mail, and the forwarded mail is then deleted from the service's mailbox.
The four verdicts
- Scam
- The message carries clear signs of fraud. Do not reply, click or pay.
- Suspicious
- Several warning signs, but no proof. Verify through a second, independent channel before anything else happens.
- Legitimate
- No known scam sign found. That is an assessment, not a guarantee — stay careful whenever money or passwords are involved.
- Unclear
- Too little content or context for a verdict. Submitting more of the message usually helps.
The models
Without an account a free model on a machine the operator owns does the reading. An account additionally opens Anthropic's Claude models; the submitted content then goes to Anthropic for the check — details in the privacy policy. Every account has a monthly cost budget, counted on each check.