If you run a platform where people create profiles, whether it is a dating app, a marketplace or an online community, sooner or later you will discover the same thing: not everyone who signs up is who they claim to be. Some are bots. Some are patient scammers. And some are just people using someone else’s photo because they think it is funny.
Last year I wrote about why dating apps have a trust problem, and last month about how we are building Slide, a dating app with real people. Since then, several founders from very different industries have asked me the same question: concretely, what do you do about fake profiles? So here is, as practically as I can put it, how I think about the problem.
Why fake profiles exist
There is no single type of fake profile, and that is the first mistake many teams make: they look for one fix for a problem with many faces. A few categories you will find on almost any platform with users:
- Romance scammers: they build a relationship over weeks or months, then comes the “emergency” and the request for money.
- Catfishing: someone poses as someone else, using stolen photos, for reasons ranging from insecurity to manipulation.
- Bots and spam: accounts created in bulk that push links, ads or try to move you to another app.
- Fake reviews and reputation: on marketplaces, accounts created only to inflate or bury a seller.
- Opportunists: duplicate accounts used to dodge a ban or a limit.
Each has different motives and different patterns. A bot gives itself away through speed and volume. A romance scammer gives themselves away through the story and the moment money enters the conversation. Treat them all the same and you only catch the obvious ones.
What it really costs a platform
Many founders see fake profiles as a moderation issue, an operating cost. I see them as a product issue. A user who runs into two or three dodgy profiles in their first week does not file a support ticket. They simply leave and tell their friends the app is “full of fakes”. You have lost trust, you have lost retention, and you have gained a reputation that is very hard to shake off.
On top of that comes the support burden: reports to review, accounts to block, upset users to handle. And in the worst cases, people who actually lost money or were put at risk. At that point we are no longer talking about metrics, we are talking about responsibility.
Layers of defence, not a wall
No single measure solves the problem. What works is a sequence of layers, each catching what slipped past the one before:
- Friction at signup. Verifying a phone number and an email address will not stop a determined scammer, but it stops a big share of automatically created accounts. It is cheap and worth it.
- Selfie verification with liveness checks. The user takes a photo or a short clip in a specific pose, and the system compares the face with the profile photos. The point of “liveness” is to confirm there is a real person in front of the camera, not a photo of a photo.
- Optional ID verification. It gives the highest certainty, but it comes at a price I will get to below.
- Behavioural signals. How many messages a new account sends in an hour, whether it sends the same text to everyone, whether it immediately tries to move the conversation elsewhere. These patterns say a lot.
- Human plus automated moderation. Algorithms handle the volume, people judge the nuance. One without the other is either too slow or wrong too often.
Video: much harder to fake than a photo
A photo can be stolen, filtered or picked from a thousand takes. A short video in which a person moves, talks and laughs is much harder to fabricate and tells you far more about who they really are. That is why I think video on a profile is not just a nice feature but a trust tool. When you see someone in motion, you have a better idea of who will turn up on the first date.
That is exactly the philosophy behind Slide: verified profiles and Video Reels, so people can see each other as they really are, not just as they look in their single best photo.
ID documents and the price of privacy
This is where you have to be very careful. Verification with an ID card or passport looks like the ultimate fix, but it means storing extremely sensitive data. Under GDPR (the EU’s data protection regulation), every piece of data you collect has to be justified, protected and kept only as long as needed. I once wrote a practical GDPR guide for online stores, and the core principle applies just the same: data minimisation. Do not ask for more than you need.
In practice, that means making ID verification optional, explaining clearly why you ask for it, not keeping a copy of the document longer than necessary, and working only with providers that play by the same rules. A user who feels interrogated at the door will not stick around to find out how safe your platform is.
Reporting, badges and user education
Very often the best fake-profile detector is your users. But only if you make reporting easy: two steps, not seven. And only if you respond fast. A report that sits unanswered for three days teaches people to stop reporting.
Then there are visible verification badges. If you went to the trouble of verifying someone, show it. Users should know at a glance who they are talking to and, if they want, be able to choose to interact only with verified profiles.
And perhaps the simplest thing of all: educate your users. Short messages, at the right moment, repeating one golden rule: never send money to someone you have not met in person. It sounds obvious. It is not, for someone who has become emotionally attached to a conversation that has lasted a few weeks.
The limits: nothing is perfect
I have to be honest about the uncomfortable part, too. Generative AI makes fake photos far easier to produce. Last year I tested DALL-E 2 and Midjourney, and the tools have only improved since. A profile built from faces that exist nowhere can no longer be caught with a reverse image search alone. Which means verification has to keep evolving: what worked last year may not be enough next year.
Then there is the balance between safety and friction. Every extra verification step loses some people at signup, including perfectly honest ones. Put up too many barriers and you end up with a platform that is safe and empty. Too few, and it is busy and toxic. The art is to add friction where the risk is high and let everything else flow.
In short
Fake profiles are not solved by a feature but by a product culture: trust treated as a core feature, not as a support department. Multiple layers, video wherever possible, respect for people’s data and fast reaction when something slips through.
What have you seen work? Have you ever quit an app because of fake profiles or, the other way round, stayed on one because you felt safe?