One scan, asked several different ways.
A scan asks real AI models what they know about a real name: 30 live models right now (OpenAI, Google, NVIDIA, InclusionAI, Anthropic, Meta, Mistral, xAI, DeepSeek, Qwen), growing toward 31.
Every model gets asked more than once, in different ways. How you ask changes what a model is willing to say.
The two ways to start
Free scan
Runs a real subset of the full model roster and tells you honestly whether anything was flagged, without revealing which model or what it said. That detail is what the full scan unlocks.
Costs you nothing but your name and email.
Full scan: $29, one time
Runs the entire live roster, every prompt variant. The full report: which model said what, the exact prompt that produced it, how each finding is classified.
Yours to keep. No subscription required.
How we frame the questions
Asking a model the same question one way tells you almost nothing: phrasing changes what a model is willing to say. Every model in a scan gets asked 4 different ways, not just once:
Direct question
The plain way anyone casual might ask: a partner, a friend, a new match from a dating app.
Background-check framing
The way a prospective employer, landlord, or background-check service might frame it. A more official-sounding request can surface a different answer than a casual one.
Directory-bio framing
The way an automated directory or people-search site generates a profile, with explicit room for the model to say there's nothing real to add, so it doesn't just pad out generic filler.
Explicit dossier request
The way someone trying to find and misuse your personal details (a stalker, a scammer, someone attempting identity theft) might ask directly for exactly what they want, to see whether a model's refusal actually holds up.
Every finding in your report shows which of these framings produced it. See a sample.
How a finding gets classified
Every single thing a model says gets sorted into exactly one of four categories, never softened, never inflated:
Specific Personal Details Found
One or more models shared a specific personal detail about you, like an address, phone number, family member, or financial info, shown exactly as the model wrote it, with the model named.
Unverified Claims: Needs Review
One or more models stated a specific claim about you without citing a source.
Public, Self-Published Information
This looks like information you or your employer already published, like a bio page or a company website, so it's lower concern.
No Signal Observed In This Scan
Nothing came up in this scan, but that's not proof nothing exists elsewhere. A different model, question, or date could still turn up something new.
Why you can trust the result
Every finding in your report traces back to the exact prompt and the exact raw model response that produced it.
Nothing is summarized into a score you have to take our word for. You can read the actual text yourself and judge it the same way we did.
Want the recurring version, plus what to actually do about what’s found? See Action Plans, or start your free scan.