Romance Scam Checker | Free Verification | LegalAudit

Romance scams cost victims more than $1.3 billion in 2024 in the US alone (FTC Consumer Sentinel) and the median victim is no longer the stereotype — it's a working professional aged 35-65 who met someone on Bumble, Hinge, LinkedIn, Instagram, or a language-exchange app. The pattern is so consistent that you can pre-empt the scam before any money changes hands, if you know what to look for. The Romance Scam Checker codifies that pattern recognition into a 90-second analysis.

We do three things in parallel: reverse image search against TinEye, Google Lens equivalents and our own stolen-photo corpus (military officers, oil-rig engineers, doctors-with-Daughters-International stock are the top abused identities); language-model fingerprinting on the chat to detect script-following and LLM-bot operation; and cross-reference against 388 international scam pattern entries including SW-ROMANCE-MILITARY, SW-PIG-BUTCHERING, SW-CARETAKER-EMERGENCY. You get the verdict in plain English: GREEN if statistically plausible, AMBER if some red flags surfaced, RED if the chat or photo matches a known scam fingerprint.

How it works

Upload a profile photo or paste 10-20 messages

Drag a screenshot of the profile or save the photo and upload it. For chats, paste a handful of recent messages — even better if you can export the full conversation from WhatsApp / Telegram.

Add one-line context

Tell Mythos how you met (which app, how long ago), where they say they live, and what they say they do for work. That context lets the model check internal consistency.

Read the cross-checked verdict

You get a verdict with: image search hits (if any), chat-pattern matches against the corpus, and a plain-English summary of the specific red flags found. If RED, you also get the recommended next action — usually 'do not send money, preserve evidence, block when safe to do so'.

Generate the dossier if money is involved

If money already changed hands or is about to, the forensic dossier captures the analysis with timestamps, hash-locked screenshots, and Swiss case-number format suitable for a complaint to Polizia Postale, IC3, Action Fraud, or your national equivalent.

What we detect

Stolen photos (military officer, doctor, engineer abroad personas)

Profile pictures with reverse image hits on known scam photo sets

Scripted opening lines and pacing (LLM bot fingerprint)

Persona inconsistencies (military deployed somewhere they aren't deployed)

Urgency escalation toward gifts, crypto, or emergency wires

Investment pivot patterns (pig butchering precursors)

Inability to video-call or always-fails-at-video patterns

Premature 'I love you' / 'I want to marry you' timing

Pivot from dating app to WhatsApp / Telegram early

Cross-reference with SW-ROMANCE-* corpus entries

Frequently asked questions

Can you confirm a profile is definitely real?

No — and beware any service that claims to. We can confirm a profile is definitely suspicious (RED): the photo is stolen, the chat matches a scam script, the persona is internally inconsistent. We cannot certify the inverse. GREEN means 'no red flags detected', which raises confidence but is not proof. The fundamental asymmetry of scam detection: it's easier to catch fakes than to certify originals.

I have already sent money. What can the dossier do for me?

Two things. First, it gives you a forensic narrative with hash-locked exhibits that you can attach to a complaint with Polizia Postale (Italy), IC3 (US), Action Fraud (UK), Kantonspolizei (Switzerland) — credible complaints get more priority. Second, your bank may treat the case as a payment-fraud incident under PSD2 / DSP3 if you can demonstrate the recipient was a known fraudulent identity at the time of transfer; the dossier provides that evidence.

Do you actually do reverse image search, or just a chat analysis?

Both. We run perceptual hashing against our internal corpus of ~12,000 known-stolen scam photos (military officers, oil-rig engineers, etc.), and we use third-party reverse image APIs (when consented) to broaden the search to TinEye + Google Lens equivalents. The chat analysis is separate: it uses LLM-based intent classification and cross-references our 388-entry scam pattern corpus.