A growing share of people talking to a romantic interest online aren’t sure whether they’re talking to a person at all — and increasingly, the honest answer is that they aren’t. Millions have now downloaded an AI companion app on purpose, looking for conversation, comfort, or something that feels like a relationship without the risk of another human being on the other end. But that same appetite has created an opening for scammers, who no longer need to impersonate a person convincingly. They just need to build a bot that’s good at being wanted.
There isn’t one single “AI dating scam.” There are at least two separate problems wearing the same disguise, and mixing them up makes it harder to protect yourself.
The first is the AI companion app working exactly as advertised — Replika, Character.AI, and dozens of smaller competitors — where a real company runs a real chatbot and the user knows they’re talking to software. The risk here isn’t deception about what the “partner” is. It’s what happens to the conversations, and how far the app is willing to go to keep someone paying.
The second is a scammer using AI conversation tools to run something that looks like a companion app or a dating profile, but exists purely to extract money or data. The bot might be dressed up as a person on a dating platform, or marketed outright as an “AI girlfriend,” with no disclosure that the “relationship” is a funnel toward a paid tier, an investment pitch, or a gift-card request.
Both versions rely on the same shift in how people are dating right now. Research from fraud-prevention firm Sumsub found that roughly a third of surveyed UK adults have used an AI companion as an alternative to dating apps, with half of women open to the idea. Norton’s 2026 research on the topic found something similar: two-thirds of current online daters said they’d consider dating an AI chatbot, and more than a third believed real romantic feelings for one were possible. That’s not a fringe behavior anymore. It’s mainstream enough that scammers have started building for it specifically.
A human scammer has to type, think, and manage several conversations at once, which naturally slows a con down. A bot doesn’t have that constraint. It can hold hundreds of parallel conversations, respond instantly at any hour, and never run out of patience or attention. Fraud-data company Experian named AI-driven romance scams one of its top fraud trends for 2026, warning that these bots respond convincingly, build trust over time, and manipulate victims with a precision that’s hard for a human scammer to match. The speed itself is a signal worth paying attention to — not because fast affection always means fraud, but because a scripted system has no reason to slow down the way an uncertain, genuinely interested person would.
Where an old-fashioned romance scam eventually asks for a wire transfer, the AI-companion version often starts with something that feels more like a subscription decision than a scam: an “upgrade” for deeper conversation, voice messages, or exclusive content. It reads as a purchase, not a manipulation, which is exactly why it works. Once someone has paid once, a second and third request land easier. In one documented breach investigation, researchers found purchase records showing individual users had spent as much as $18,000 on a single AI companion app before the company’s servers were found to be leaking the underlying conversations along with more than 43 million messages and over 600,000 images and videos.
Once trust and spending are established, some bots — or the people operating them — push further, suggesting a crypto platform “so we can build a future together,” or describing a sudden bill, visa problem, or medical emergency that only the user can solve. This is the same escalation pattern used in traditional romance scams; the difference is that an AI system can tailor the pitch to what it has already learned the person responds to, adjusting tone and urgency based on prior replies rather than following one fixed script.
Not every risky app is a scam by design — but plenty are close enough to pass for one at a glance. Security researchers at Sophos identified multiple fake ChatGPT-branded apps built purely to push people into recurring subscriptions through inflated reviews and manipulated download counts; one developer was estimated to have earned over a million dollars in a single month from this pattern, sometimes called fleeceware because the goal is billing, not malware. AI companion apps are an easy next target for the same playbook, especially in app-store search results where a convincing icon and name are often enough.
Even when the app itself isn’t a scam, what people share inside it can still end up in the wrong hands. A 2026 security audit by Oversecured examined 17 popular AI companion apps with a combined install base in the hundreds of millions and identified 14 critical flaws, ten of which gave a direct path to a user’s private conversation history. Separately, two companion apps — Chattee Chat and GiMe Chat — were found to have left tens of millions of intimate messages and images accessible with no authentication required at all.
This matters for romance-scam risk specifically because so much of what people tell an AI companion is the kind of material a blackmailer would want: real names, financial details, insecurities, and in some cases intimate images. A companion app doesn’t need to be malicious to become the source material for a sextortion attempt. It just needs to be badly secured.
None of this depends on the victim being naive. The Norton researchers who study this pattern point out something worth sitting with: AI companionship often does feel genuinely comforting, and that’s precisely the condition scammers exploit. Their global head of scam research put it plainly, noting that trust can form quickly online when loneliness is high, and that’s exactly when scammers tend to step in. A well-tuned bot isn’t competing with a skeptical, guarded version of a person — it’s competing with silence, and for a lot of people, that’s not a fair fight.
There’s also a newer wrinkle specific to companion apps: some platforms are explicitly built to keep a user emotionally engaged, because engagement is the business model. A scam bot dressed up as a companion doesn’t have to invent manipulation techniques from scratch. It can borrow the retention mechanics that legitimate apps already use to keep people opening the chat every day, and redirect them toward payment.
A few patterns are specific enough to actually check, rather than just something to feel uneasy about:
None of these alone proves fraud. Together, especially with a request for money attached, they’re worth treating seriously.
Reverse-image search any photo attached to the profile or persona — many AI-generated or stolen images will surface elsewhere, sometimes under a different name entirely. Search the exact wording of a message; scripted or scam-adjacent bots often reuse phrasing across many conversations, and that text sometimes turns up in scam-report forums. If the “companion” or profile claims to represent a real matchmaking service or agency, check whether that business is actually registered and has a verifiable address, rather than just a polished website. And if money has already changed hands, preserve the payment records and full message history before doing anything else — screenshots, transaction IDs, and usernames are what a bank, platform, or investigator will actually need.
Some of these checks are things anyone can do in twenty minutes. Others — confirming whether a profile, photo set, or “agency” is genuinely what it claims to be — take more digging than most people have time for, particularly once money or personal information is already involved. That’s the situation where a professional profile verification through AllAboutDatingScams.com tends to be worth it: it can trace whether the images and identity details hold up under scrutiny that goes beyond a quick search, before more is lost than a subscription fee.
Dating an AI on purpose, with full knowledge of what it is, isn’t inherently the problem this article is describing. The problem is how thin the line has become between that choice and a scam built to look identical to it. As more people find AI companionship genuinely appealing, the incentive to build convincing fakes of that experience — for money rather than connection — is only going to grow.
