Catfishing used to mean stealing someone else’s photos. Now a scammer can generate a face that belongs to nobody, write fluent messages in any language, and produce new “selfies” on request. The FBI warned in December 2024 that criminals use generative AI to create believable profile photos, write messages without the spelling mistakes that used to give them away, and even make images to “prove” they’re a real person.
Most AI-generated profiles are still catchable. It just takes a few deliberate checks instead of a gut feeling.
Not every AI photo is a catfish
A real person might use an AI headshot made from their own selfies. That can break app rules if it’s misleading, but it’s a different problem from a fake identity. The question that matters for your safety: is there a real, reachable person behind this profile who looks like these photos? The ethics of real people using AI photos are covered in using AI photos on dating apps.
Photo red flags
The FBI’s list of things to look for in images and videos includes distorted hands or feet, unrealistic teeth or eyes, indistinct or irregular faces, odd accessories like glasses or jewelry, inaccurate shadows and watermarks. In practice:
| Look at | What gives AI away |
|---|---|
| Hands | Extra, fused or bent fingers; hands hidden in every photo |
| Ears and jewelry | Earrings that don’t match; glasses that melt into skin |
| Teeth and eyes | Too-perfect teeth, mismatched pupils or reflections |
| Hair | Strands that blur into the background |
| Background | Warped text, signs that are gibberish, impossible architecture |
| Lighting | Light on the face from a different direction than the scene |
| The set as a whole | Every photo looks like a magazine shoot; no casual, blurry, group or old photos |
That last row matters most. Real people’s photo sets are uneven. A profile of six flawless, similarly lit portraits is suspicious even if no single image has a visible glitch.
AI images keep improving, so a photo with none of these flaws isn’t proof of anything.
Message red flags
Language models write well, which removes an old warning sign. Watch for patterns instead:
- Replies arrive at any hour, instantly, at the same length and in the same upbeat tone.
- Answers are warm but generic, and specific questions get vague responses.
- They don’t remember details from earlier in the chat, or contradict them.
- They push to move to WhatsApp, Telegram or another app quickly.
- They can’t or won’t do a live video call, or the call is “glitchy” every time.
The FTC’s guidance on romance scams adds the ones that matter most: someone you haven’t met in person who asks for money, gift cards, wire transfers or cryptocurrency, or wants to “teach” you to invest. Its advice is blunt: never send money or gifts to someone you haven’t met in person.
Five checks before you trust a profile
1. Look for the verification badge, and know its limits. Bumble requires photo verification in the U.S. Hinge’s Selfie Verification and Tinder’s Photo Verification are optional. Hinge says its badge shows someone looks like their profile photos but doesn’t guarantee their safety, and Tinder says verification isn’t a guarantee that the information a user provides is authentic (help centers checked 2026-11-06). A badge helps; it isn’t proof.
2. Run a reverse image search. Save a profile photo and search it with a reverse image search tool such as Google Lens or TinEye. Matches on a stock site, a model’s Instagram or an old news story mean stolen photos. No matches at all can mean a generated face, or just a private person. AI image detectors can add a signal; they are compared, with their limits, in AI image detectors for spotting fake dating photos.
3. Ask for something specific and live. A video call is the best single test. Real-time deepfakes exist, so ask for something natural that’s hard to fake: wave a hand across the face, turn sideways, say your name. The FBI suggests watching for lag, voice mismatches and unnatural movement.
4. Check the story against itself. Job, location, timezone, photos. Scams often claim jobs that explain why they can’t meet: offshore, military, overseas contract.
5. Keep it on the app until you’ve met. In-app chat keeps the dating app’s safety tools and reporting in play. A rush to move off the platform is one of the FTC’s listed warning signs.
What not to do
- Don’t publicly accuse someone based on a detector score. Real people get false positives.
- Don’t send money, gift cards or crypto to test whether someone is real.
- Don’t use face-recognition search sites to identify a stranger. They raise serious privacy problems and aren’t needed to spot a fake.
- Don’t keep chatting to “catch them out.” Report and move on.
How to report a suspected AI catfish
Report in the app first. It protects the next person and gives the platform evidence.
- Bumble: report the profile as a fake profile. Bumble added an option in 2024 for “Using AI-generated photos or videos.”
- Hinge: report from the profile or chat. Its prohibited-content rules ban fake identities and AI-generated content used to mislead. Hinge also has a reporting path for non-consensual intimate images, including deepfakes.
- Tinder: report from the profile or chat; its community guidelines prohibit fake accounts and pretending to be someone else.
If money or personal information is involved, also report to the FBI at ic3.gov and to the FTC at ReportFraud.ftc.gov. Be wary of anyone who contacts you offering to recover lost money for a fee; the FTC says real agencies don’t charge to help.
The scale of the problem
The FBI’s 2025 Internet Crime Report logged 23,159 confidence and romance scam complaints with about $929 million in reported losses. It also flagged more than $19 million lost to romance scams with a likely AI connection, such as AI-generated profiles or scripts. Those are only the cases people reported.
When AI moves beyond photos into live video and voice, the pattern becomes a deepfake romance scam. How those work, and what to do if you’re caught in one, is in deepfake romance scams: how they work.