Check whether someone you met on a dating app or social media is who they claim. You can test their profile photos for AI on the home page.
Free with no sign-up, and the analysis runs inside your browser — files never leave your device. On our public gate (30 real + 83 AI images) it measured 94.7% at original quality and 87.6% on recompressed copies, with zero real photos confidently mislabelled as AI. Treat it as an indicator, not proof: the newest commercial generators evade every detector.
A romance scam manufactures a fake person — stolen or AI-made photos, devoted daily chat, then money talk a few weeks in. Check both the photos (home checker, reverse search) and whether the relationship pattern matches the formula (checklist above).
Week 1: an attractive profile approaches you first. Weeks 2–3: daily affection builds trust, and the chat moves off the app. Week 4 onward: an accident, illness or investment opportunity arrives with the first money request. Few scams deviate from this schedule — place your situation on it.
Upload a few profile photos to the home checker. If AI signals appear, or reverse search finds the same face under another name, then whoever you are talking to, it is not the person in the photos.
It used to be an accident, a hospital bill, a customs charge: a hard-luck reason attached to a direct transfer. The commoner form now is investment. They mention a method they use to make returns, introduce a particular exchange or app, and invite you to join. Romance fraud fused with investment fraud.
It works because money really does come out at first. Put in a small amount and returns appear on screen, and withdrawals go through. That experience builds trust and the amounts grow. Only after a large deposit does withdrawal stop working, followed by demands for tax or a deposit before funds can be released.
Judge the structure rather than their sincerity. An exchange installed from a link rather than an app store, an operator not registered in your jurisdiction, withdrawals that are conditional — any one of those means stop, whatever the relationship. The person who introduced it may be entirely sincere and also being defrauded themselves.
The occupations repeat. Soldier deployed overseas, engineer on a rig, doctor with an international organisation — roles with the difficulty of meeting built into the job. They make identity hard to verify and sudden financial emergencies easy to stage.
Time zones are a clue. When the hours they are online contradict the country they claim, or replies arrive instantly at every hour, you may be talking to a team sharing one account. Activity concentrated into narrow windows says something about working patterns too.
Watch the pace of feeling. Talk of a future arrives early and forms of address grow intimate quickly, while video calls and spontaneous photos keep being deferred for natural-sounding reasons. When those two appear together, the relationship and the verification are moving in opposite directions.
Stop sending. Almost without exception, a demand framed as recovery follows — pay the tax and the withdrawal clears, cover the fee and the account unlocks. The largest losses happen after the first transfer, not during it.
Then preserve records: the full conversation, their profile, transfer records, and the address of any app or site you were pointed to. For a bank transfer, ask your bank and the police for a payment stop; for crypto, report the receiving address as well. Recovery chances fall sharply with time.
And this is not something to be ashamed of. The other side is often an organisation running a scripted approach across many people, and that script targets attachment rather than intelligence. Choosing to cover it up rather than report is what produces the next victim.
The profile photo scored high for AI generation, video calls were deferred every time as travel or company policy, and about three weeks in the conversation turned to an investment opportunity. Photo signal, avoidance and a money ask arrive in that order. The order itself is the romance-scam formula, and the first two are enough to stop before the third.
They accepted a spontaneous video call, and the account carried years of posts including photos with other people. Almost no risk signals fire. It is not a guarantee, but live verification and a long social trail are the two things a scam operation finds hardest to manufacture.
Aggressive beauty filters erase pores and fine lines and flatten skin texture, which is statistically what generated faces look like. A real person’s selfie can score high for that reason alone. So never judge a person on one photo score. If you need certainty, stop analysing images and ask for a live video call. It is cheaper and far more conclusive.
Real-time face swaps exist but fail on side profiles and occlusion. Casually ask them to turn sideways or wave a hand across their face. Sudden quality drops or dodging are signals.
Refuse any further payment regardless of the story ("pay the tax to get it back" is the classic follow-up), keep all transfer records, and report to your bank and police immediately. No shame. You were run through a professional script.
A good sign, but not conclusive. Real-time face replacement exists, and pre-recorded video gets played back. Ask for something spontaneous during the call — turn to show a profile, wave a hand across the face, hold up a specific number of fingers. Live synthesis tends to break on exactly that.
The money request usually arrives weeks in. Until then there is nothing to find, which is normal. So judge on whether verification is possible rather than on whether money has come up — how they respond to a spontaneous video call, a photo request or a suggestion to meet answers the question at any stage.