Something doesn't add up. The photos are too good. The account feels hollow. You can check whether the images and videos on any public social media profile are real or AI-generated — in seconds, without them knowing.
Fake social media profiles fall into three main categories: romance scam accounts using AI-generated faces, bot accounts with stolen or synthetic photos, and impersonation accounts mimicking real people. In all three cases, AI-generated images are increasingly the content of choice because they are unique, photorealistic, and can't be reverse-image-searched to an original source.
These are the structural signals — things you can observe on the profile itself before running any detection:
Instagram is one of the most common platforms for fake profiles because it is visually focused and a public following provides social credibility. Here's how to verify any public Instagram post with KweliAI:
Free for X and Reddit · Pro for Instagram, TikTok, YouTube, Facebook
Start scanning →The process is the same on every platform — find a public post with their photo or video, copy the post URL, paste it into KweliAI. Here's what to look for on each platform:
AI face generators — including Midjourney, DALL·E, StyleGAN, and ThisPersonDoesNotExist — can produce photorealistic human portraits that most people cannot distinguish from photographs of real people. The faces are unique (they belong to no real person), high-resolution, and can be generated in seconds with no photography skills required.
The visual tells that used to help identify AI-generated faces — odd hands, strange ears, asymmetrical features, blurred backgrounds — have been largely fixed by newer models. Midjourney v6 and DALL·E 3 produce faces that consistently pass casual visual inspection by most people.
What AI generators cannot fully eliminate are statistical artifacts at the pixel level — frequency domain anomalies, patterns in how textures are distributed, subtle inconsistencies in lighting physics that real cameras don't produce. These are invisible to human visual inspection but detectable by models trained specifically to find them.
This is why KweliAI's detection works at a model level rather than a visual one. The three detection models don't look at the image the way a person does — they analyse statistical properties of the pixel data that AI generation processes leave behind, regardless of how convincing the image looks to a human viewer.
AI-generated — High confidence means the image was almost certainly created by an AI tool. The face in the photo does not belong to a real person. The profile is almost certainly fake.
AI-generated — Medium confidence means significant AI generation signals were detected, but not enough for a definitive verdict. This may indicate a heavily edited real photo, or an AI image that has been processed to reduce detection signals. Treat this with serious caution and scan additional posts.
Real — High confidence means no AI generation signals were detected in this specific image. The photo is consistent with a camera-captured image. Note that a real photo can still belong to a different person than the one claiming it — combine this with a reverse image search to verify identity.
Deepfake means AI was used to manipulate a real person's face or voice in the content. This is different from a fully AI-generated face — it means a real video was altered with AI. This is used in more sophisticated scams and impersonation attacks.
Scan at least three different posts from the same account. A single result is informative. A pattern across multiple posts is much more reliable.
Is there a way to tell if an Instagram profile is fake without following them?
Yes. If they have public posts, paste any post link into KweliAI. You don't need to follow them, interact with them, or have an Instagram account yourself. KweliAI works from the public URL alone. Instagram scanning requires a Pro plan at $12/month.
Can someone make a fake profile with real photos of someone else?
Yes — this is called identity theft for catfishing. The scammer steals real photos from another person's real social media account and uses them under a different name. KweliAI will return a "real" verdict for these because the photos are genuinely camera-captured — they just belong to someone else. To catch this, run a Google reverse image search on the photos to see if they appear elsewhere online under a different name.
What if only some of their photos come back as AI-generated?
This is common. Scammers sometimes mix real stock photos or photos taken from real people's accounts with AI-generated ones. Even one AI-generated verdict on a photo that is supposed to be them should be treated as a serious red flag.
How accurate is KweliAI's detection?
KweliAI achieved 99.01% accuracy on the 2025 Braintrust model validation benchmark. Every result includes a confidence rating — high, medium, or low — so you always know how certain the verdict is.
Can I check a profile that blocked me?
If an account that blocked you is set to public (visible to people not logged in), KweliAI can still scan their public posts. If the account is private, KweliAI cannot access the content.