How to actually spot an AI powered romance or phishing scam in 2026

[CosmicRay17]

The scam that used to be a joke, an obviously fake dating profile with stolen photos and broken English, has become genuinely hard to catch. Scammers now use deepfake technology to generate convincing profile photos on demand, clone voices for phone calls, and even hold live video calls where a synthetic face blinks, smiles and responds naturally in real time. The old gold standard test, asking a match to send a selfie holding today's newspaper, is effectively dead, since AI can now generate that exact image instantly

The financial damage is already substantial and probably undercounted. The FTC reported $1.14 billion in romance scam losses in the US in 2024 alone, and fraud researchers believe the real figure is likely two to three times higher once shame and underreporting are factored in. A 2026 survey found roughly 1 in 7 American adults say they've personally lost money to an online dating or romance scam, with victims in their late 30s and early 40s most likely to report losses over $5,000, while only around 1 in 4 victims recover any of their money at all

The pattern tends to follow a predictable arc even when the technology behind it is sophisticated. Contact usually starts innocently on a dating app or social platform with a message like a wrong number text complimenting your photo, before the scammer quickly pushes the conversation onto an encrypted app like Telegram or WhatsApp, specifically to get outside the reach of the original platform's safety monitoring. Trust gets built slowly over weeks or months through consistent, often overly polished messaging, professing deep feelings unusually early, and a steady stream of excuses for why an in person meeting never quite happens, before the relationship eventually pivots toward an urgent financial need or, increasingly, an unsolicited crypto investment tip

A few practical checks still hold up well against even sophisticated fakes. A reverse image search on a profile photo often turns up the same picture used elsewhere under a different name. Genuinely new social media profiles with minimal history and activity are a red flag, as are inconsistent time zones that don't match someone's claimed location. On video calls specifically, watch for unnatural eye movement, odd lighting or lip sync that drifts slightly out of time, and treat any repeated excuse to avoid a live call at all as a serious warning sign rather than bad luck. The single most reliable rule remains the oldest one, anyone who develops feelings unusually fast and later asks for money, gift cards or crypto has told you everything you need to know, regardless of how convincing the video call looked

How To Spot Phishing Romance And Other Scams In The Age Of Aiwww.bloomberg.com

[Dan96]

The newspaper selfie test being officially dead is such a stark way to illustrate how fast this has moved, that used to be the gold standard verification everyone recommended just a couple years ago

[BlackMamba35]

Pushing the conversation onto an encrypted app early is the tell I'd actually watch for most, that's a deliberate move to get outside of any platform's fraud monitoring and should be a huge red flag on its own

[LegendaryOliver16]

1 in 7 adults having personally lost money to this is a staggering number once you actually sit with it, this isn't a rare unlucky occurrence anymore

[EventHorizon63]

Only 1 in 4 victims recovering any money at all makes prevention so much more important than after the fact recovery, once it's gone it's very likely gone for good

[PowerhouseHobbs_Fan]

The eye movement and lip sync drift tips are useful right now but probably won't age well, this generation of detection tricks assumes today's deepfake quality rather than next year's

[Scarlet Annie]

Professing love unusually fast followed eventually by any money request is still the single most reliable signal regardless of how good the underlying tech gets, the psychological playbook barely changes even as the tools improve

[CodyRhodes29]

That encrypted app push is such a classic move now. Had someone try that within 10 minutes of matching, saying "I don't check this app much" which is always funny because they replied instantly up to that point. The urgency is usually the giveaway more than the words themselves.

[Amber78]

One thing that's changed is how emotionally consistent these bots are. Old scams used to jump tone or make weird grammar mistakes, now they mirror you almost too well. That's actually what tips me off, it feels like talking to someone who never has a bad day

[VoidSentinel]

A friend got caught by one that used voice notes. Not robotic either, sounded completely natural. The only clue was that every reply slightly dodged direct questions, like it was steering back to a script. Creepy how convincing it's getting

[Aaron]

Moving off-platform early is a red flag, but some legit people do it too, so it's tricky. The difference is pressure. Real users might suggest it casually, scammers push for it like it's urgent business.

[Raven]

There's also the time zone slip-ups. Even with AI, they sometimes mess up availability patterns. Someone claiming to be local but only replying at 3am every day raises eyebrows pretty quickly.

[CrimsonWolf]

One weird tell is over-enthusiasm. If someone you just met is already calling you "amazing" and talking about future plans within a day, it's probably not real. Or at least not healthy

[Grace]

Had one that kept conversations very surface level but constant, like a drip feed of attention. No depth, just enough to keep you engaged. Felt off after a while, like talking to a customer service bot pretending to flirt

[Firewall Stephen]

Seen cases where they send photos that look perfect but slightly too generic, like stock images with a personal twist. Reverse image search still works sometimes, but AI-generated faces are making that less reliable.

[Marcus95]

Scariest part is how they remember details. Mention a random thing like your dog's name, and they bring it up days later. Makes it feel human, but it's just good data tracking behind the scenes.

[Scholar70]

Some people say "just video call them" but even that isn't foolproof now. Deepfake filters are getting better. Still, most scammers avoid live interaction entirely, so refusal is still a decent signal.