Solutions/Dating Platform

Dating Platform Fraud Detection

Romance scams start with a face that looks real. Today that face takes seconds to generate. Sightova checks each profile photo for invisible AI watermarks, content credentials and pixel-level signs of generation, so you can flag a fake before a single message is sent.

AI PROFILE PHOTOS
GOOGLE + OPENAI WATERMARKS
C2PA + EXIF
example.com/admin/new-profiles
AI-generated dating-profile photo of "Daniel", a man in a hi-vis jacket and hard hat on an offshore platform

Daniel, 41

Offshore engineer

3 km away

“Six more weeks on the platform, then I'm all yours. Looking for something real.”

  1. SynthID watermark

    Google Gemini · OpenAI GPT-Image

    Checking
  2. Content credentials

    C2PA manifest · EXIF metadata

    Pending
  3. Synthetic media model

    Pixel-level forensic analysis

    Pending
Checking new profile photo…
// HOW IT WORKS

Check profile photos before the first match

A scammer needs a believable face before anything else. Check photos at upload and route flagged profiles to your reviewers before anyone can message them.

Uploading profile photo

Step 1

A photo is added

At signup or whenever a member changes their photos, send each image to the API before the profile goes live in search and matching.

SynthID watermarkFound
Content credentialsAI declared
Synthetic media model95%

Step 2

Three checks per photo

Hidden AI watermarks from Google and OpenAI, signed content credentials, and our own model reading the pixels. Any one of them is enough to flag a generated face.

Genuine member

No AI signals

Approved

Ryan, 35

AI-generated photo

Held

Step 3

Your team decides

You set the threshold and decide what happens next. Photos with no signals go through, and flagged ones go to a reviewer, who can ask the member for another photo.

// REAL SCANS OF FAKE PROFILES

Five romance-scam personas. None of them exist.

We generated these with one of today's best image models, using the stories scammers actually tell. A quick glance won't catch them, and a reverse image search finds nothing because the faces are new. Our checks flagged all five. See how we test on 250k+ benchmark images and 50+ generative models in our published testing methodology.

Test it on your own profiles
AI-generated profile photo of "Daniel"
AI-generated · 96%Away for work

Daniel, 41

Offshore engineer

“Six more weeks on the platform, then I'm all yours. Looking for something real.”

OpenAI watermarkC2PA: AI declared
AI-generated profile photo of "Sofia"
AI-generated · 95%Too good to be true

Sofia, 29

Nurse, loves travel

“Coffee first, adventures second. Here for the long run.”

OpenAI watermarkC2PA: AI declared
AI-generated profile photo of "Mark"
AI-generated · 94%Mission abroad

Mark, 48

Doctor on a medical mission

“Widower, working abroad until spring. Kind hearts only.”

OpenAI watermarkC2PA: AI declared
AI-generated profile photo of "Elena"
AI-generated · 94%Investment pitch

Elena, 33

Investor, lives by the sea

“Happy to show you how I grew my savings. Ask me anything.”

OpenAI watermarkC2PA: AI declared
AI-generated profile photo of "Ryan"
AI-generated · 95%The everyday guy

Ryan, 35

Hiker, dog dad

“Me and Max, looking for a third. Weekends are for mountains.”

OpenAI watermarkC2PA: AI declared
Real photo

Genuine members

Most real photos raise no signals, so genuine members get on with meeting people. We don't publish real members' photos, so they appear here as a placeholder.

Every profile photo above was generated by Sightova with OpenAI's GPT-Image 2.5; no real person is behind any of them. Names, ages and bios are invented. Scores and labels are Sightova's production checks on the original files.

Profile Photo Authenticity Modules

SELECT * FROM dating_trust_matrix
DAT-SYN-01
TRUST MODULE

AI-Generated Portrait Detection

Flag profile photos generated by GAN and diffusion models before they reach your members. Each photo gets an AI probability and a verdict (ai-generated, real or uncertain). The model reads the pixels, so it works even when metadata has been stripped.

AI PROBABILITYVERDICTWORKS ON STRIPPED FILES
DAT-HMP-02
TRUST MODULE

Heatmap for Reviewers

See exactly where a photo looks AI-edited. The heatmap shows your reviewers where to look, so a flagged photo arrives with more than a number. In the dashboard on the Starter plan and up.

HEATMAPREVIEWER AIDSTARTER PLAN +
DAT-WMK-03
TRUST MODULE

Watermark & Credential Evidence

Know when a photo was made by Gemini or GPT-Image, because the watermark says so. Each check looks for invisible Google SynthID and OpenAI watermarks and reads C2PA content credentials and EXIF metadata, which can name the exact tool behind the image.

SYNTHIDOPENAI WATERMARKC2PAEXIF
DAT-GEN-04
TRUST MODULE

Closest Generator Style

For photos that look AI-generated, the result reports the closest generator style, giving reviewers a lead to follow. When a watermark or C2PA credentials are present, they name the tool.

CLOSEST GENERATOR STYLEREVIEWER CONTEXT
DAT-AGE-05
TRUST MODULE

Optional Age Estimation

Optional age estimation: an estimated age and likely range for each person, switched on in Settings. It appears right alongside the AI detection result for the photo.

OPTIONAL EXTENSIONESTIMATED AGELIKELY RANGE
DAT-API-06
TRUST MODULE

REST API for Your Workflow

Connect through our REST API from your own sign-up or moderation flow. It comes with every plan, runs on an API key, and you can start free with no credit card. You set the threshold and decide what happens next.

REST APIAPI KEYSTART FREE
PLATFORM INTEGRITY

Flag Fake Profile Photos at Upload

A single AI-generated profile can run a romance scam for months before being reported. By then, the damage—financial and emotional—is done. Sightova flags AI-generated profile photos at upload, so your team can step in before the profile reaches other members.

  • An AI probability and a verdict for each photo, tested on 250k+ benchmark images and 50+ generative models. The results are published in our testing methodology.
  • Evidence, not just a score: invisible Google SynthID and OpenAI watermarks and C2PA credentials can name the exact tool behind a photo.
  • Results in seconds. Run checks in the background after upload and send flagged photos straight to your reviewers.
  • Hosted in Germany. Member photos are deleted after analysis and not used for training.
RESPONSE — POST /api/v1/detect/ai
{
  "request_id": "req_5oQjGqU4mB2nQp9Hg7vTz",
  "endpoint": "/api/v1/detect/ai",
  "media": {
    "filename": "profile_photo.jpg",
    "mime": "image/jpeg"
  },
  "result": {
    "ai_probability": 0.94,
    "real_probability": 0.06,
    "confidence": 0.88,
    "verdict": "ai-generated"
  }
}

How AI Generated Profiles Turned Romance Scams Into a Billion Dollar Industry

Romance fraud has evolved from a niche internet scam into one of the most financially devastating forms of online crime. At the center of this transformation sits generative AI. Modern image generators produce photorealistic human faces in seconds: faces that have never existed, complete with natural skin textures, realistic lighting, and diverse demographics. These synthetic portraits let scammers create convincing profiles at industrial scale.

Why Traditional Moderation Cannot Keep Up

Unlike stolen photos, which can be flagged through reverse image search, GAN generated faces have no original source to match against. This makes them invisible to the moderation tools most dating platforms depend on. Manual review teams, already stretched thin by volume, cannot reliably distinguish a high quality synthetic face from a genuine photograph.

The same technology fueling deepfake detection challenges in media and politics is now the primary weapon in romance fraud. Scammers pair synthetic faces with AI written messages to maintain dozens of emotionally manipulative relationships simultaneously, all at near zero marginal cost.

$1.3 Billion in Losses, and the Curve Points Sharply Up

The FTC reported $1.3 billion in U.S. romance scam losses in recent years, a figure projected to more than double by 2030. Each new generation of image models produces faces with fewer detectable artifacts, narrowing the gap between synthetic and authentic imagery. Large language models further automate the conversational side of scams, creating fully automated fraud pipelines that scale without human effort.

For dating platforms, the fallout extends well beyond direct victim losses. Platforms known for fake profiles suffer user attrition, negative press coverage, and growing regulatory scrutiny. Governments are exploring liability frameworks for AI enabled fraud, and image content moderation is increasingly viewed as a baseline trust requirement rather than an optional feature.

The Organized Fraud Rings Behind the Fake Profiles

Individual scammers are alarming enough, but organized rings amplify the damage exponentially. A single operation can deploy hundreds of AI generated profiles across multiple platforms, reusing the same generators and prompts to create networks of fictitious personas. Platforms that invest in forensic grade ai image detector technology check new photos as they arrive and send the suspicious ones to reviewers, instead of relying on member reports alone.

How Sightova Checks a Profile Photo

When you send Sightova a profile photo, you get an AI probability and a verdict, a check for invisible Google SynthID and OpenAI watermarks, and a read of C2PA content credentials and EXIF metadata. The model reads the pixels, so it works even when metadata has been stripped. A watermark or C2PA credentials can name the exact tool that made the photo, and photos that look AI-generated also come back with the closest generator style. We test on 250k+ benchmark images and 50+ generative models and publish the results in our testing methodology.

Connect through our REST API with an API key and put the check wherever it fits in your sign-up or moderation flow. Results come back in seconds, and you set the threshold and decide what happens next. Optional age estimation adds an estimated age and likely range for each person, switched on in Settings. Photos are processed in Germany, deleted after analysis and not used for training. The same image detection supports cybersecurity AI detection use cases.

Build Trust on Your Platform

Flag AI-generated profile photos before other members see them. Connect Sightova's REST API to your own upload flow, send flagged photos to human review, and start free with no credit card.