Look for a pattern of requests
One delayed reply or an awkward video call has many ordinary explanations. A repeated pattern of urgency, secrecy and requests that benefit the other person deserves a different response. Focus on observable behaviour: what are they asking you to send, install, reveal or ignore?
A useful note might say “asked me to send a payment before meeting” rather than “looks like a scammer”. The first describes an event you can review. The second turns an impression into an identity claim.
Separate three kinds of uncertainty
- Who is pictured? A reused photograph may not belong to the account controller.
- Who is communicating? A genuine photograph does not establish who is sending messages.
- What is being requested? Even a correctly identified person can make a harmful or dishonest request.
Photo search addresses only part of the first question. It cannot resolve the other two on its own. An empty search is not a clean bill of health.
Choose a proportionate response
Slow the interaction down. Decline requests for money, account access or one-time codes. Talk to someone you trust outside the conversation. If the account becomes threatening or coercive, use the platform’s block and report controls. You do not owe someone continued access while you try to prove every detail of their story.
If you document a concern, preserve the messages that show it and keep them private for the appropriate reporting channel. Avoid contacting people found through lookalike results; the person in a stolen image may have no connection to the account.
Use image clues carefully
A photograph found on an older, unrelated page can be relevant, but the source needs checking. A model portfolio, fan account and personal profile have different meanings. If you cannot access the page, label the result unverified.
The FTC explains common romance-scam patterns. Use that as practical safety context, not as a template for accusing someone who shares a broad characteristic. Facehunch’s test mode is for checking the integration and must not be used to label people or accounts.