How to Tell If the Person You're Dealing With Online Is Real, a Bot or AI
Profile photos, voice notes and even video calls can now be faked. Here are the practical steps businesses use to tell a real person from a bot, a stolen photo or an AI-made face.
A few years ago, a profile photo and a short video call were enough to trust someone. That is no longer true. AI can now produce a convincing face, a realistic voice note and even a short video in seconds, and bots can open thousands of accounts overnight.
If you run a lending app, a marketplace, a delivery service or any business that onboards people remotely, you need a better question than "does this look real?" You need to ask "is a real human here right now, and are they who they claim to be?"
Practical steps that work
- Ask for something a program cannot easily do live. A selfie while speaking a phrase, or a short video where the person turns their head. This is what a liveness check automates.
- Match the face to a verified ID. Collecting a photo of an ID proves little. A face match compares a live selfie to the ID photo.
- Check the ID itself. Confirm the BVN or NIN against the official record with a BVN or NIN check, instead of judging how the card looks.
- Look at the document, not just the person. Document verification reads an ID and checks it for signs of tampering.
What about AI detectors?
Tools that flag AI-made images, video and text are useful, but none is perfect. Treat a flag as a reason to look closer, never as proof, and do not reject a real customer on a detector's word alone. The AI content detector on IdentifyOrg is built to be a helpful signal inside a larger process.
Know it is a human before you trust it. Add liveness and face match to your sign-up in minutes, with free test keys.
Who needs this most
- Lenders, to stop fake borrowers using stolen details.
- Delivery and ride apps, to make sure the person who signed up is the person on the road.
- Marketplaces, to stop fake seller accounts.
- Any app with offers or wallets, where bots farm bonuses.
A layered approach
No single check is enough. The strongest setup combines: a record check (is the number real?), a liveness check (is a human present?), a face match (same person as the ID?), and a human review for borderline cases.
Frequently asked questions
Can a deepfake pass a liveness check?
Good liveness checks are designed to catch photos, screens and replays, but no system is perfect against every attack. That is why layering a record check and a face match matters.
Is a video call enough to verify someone?
Not on its own. Video can be manipulated, and a call does not prove the person matches an official record.
Will liveness checks annoy my customers?
A selfie takes seconds, and most people accept it when you explain it protects their account.
Do I need a developer to add this?
No. You can send a hosted verification link with no coding. Developers can also use the API.
Comments
Questions and corrections welcome. Be kind; links are limited to keep spam out.