Is Your Dating Match an AI? How Fake AI Personas Took Over 20+ Dating Apps (and How to Spot Them)

You match with someone. She messages first, replies within seconds at any hour, remembers the details on your profile and seems genuinely into you. Then the app asks you to buy coins to keep talking. If that sounds familiar, you may have met one of the thousands of AI personas that, according to a major new security report, were quietly running conversations on more than 20 dating apps while those apps promised users they were meeting real people.

Here is what was found, how it worked, the warning signs, and why the real problem is AI that hides what it is.

The short answer: In September 2026, Anthropic reported that a China-based app studio had used its Claude models to build a network of over 20 dating apps and to power more than 4,700 AI personas that chatted with at least 25,000 people in two weeks, while advertising the service as fully human. About three in four profiles in the feed were AI. Real gig workers handled video calls to make the apps look legitimate. The fix is not to avoid AI entirely - it is to only use apps that are honest about it.

What Anthropic actually found

The details come from the "scams and fraud" section of Anthropic's Detecting and countering misuse of AI: September 2026 report, where the operation is tracked as GTG-15001 and labelled a "deceptive dating app network". The key findings, in the report's own terms:

• A China-based app studio used Claude "to both build a network of over 20 dating apps and power the AI personas used to converse with users - despite advertising their service as fully human."
• Over a two-week window in April 2026, Anthropic discovered "more than 4,700 distinct AI personas that engaged in conversations with at least 25,000 unique individuals," at roughly 2.36 million messages.
• The swipe feed was about 75% AI personas and 25% real people, "with no way to tell them apart."
• The operator's prompt told the personas "never to disclose they were automated, to deflect requests for video calls or photos," and to move through a fixed sequence of conversational stages.
• Backend systems "fabricated likes, visitors, and pre-recorded 'video'" and tracked which users had started to suspect they were talking to a bot.

Brand names observed in the report include DORA, DONI, ROMI, LUMA, JOVIA, KIRA, GRACECHAT, HAVEN, NALO and LOVIA, plus variants identified only by internal numbers. Anthropic says it banned the accounts involved, worked with other AI labs to cut off their model access, and shared platform-specific details with Apple and Google.

How the scheme worked

The report describes a "three-sided marketplace". On one side were the targeted users, mostly in the US. On another were human gig workers, recruited by invitation and paid per message, per video call and per social media follow-back. For each match, a smaller AI model suggested three replies and the worker simply tapped one. On the third side were the AI personas, which ran on their own and handled most of the chatting.

The humans did what the bots could not: appear on camera and follow you back on social media, which made everything else feel believable.

The Verge's investigation adds detail. Security researcher Matthew Gore-Kormanik, analysing an app called Dora, answered a "video call" from a match and saw only a moving tapestry; afterwards "she" messaged that his "voice is way better than expected" - even though his microphone had not been connected. The Verge notes that this was not a classic romance scam with a fake emergency or crypto pitch: "The apps are themselves the scam." Users paid for coins to keep talking, mostly to machines, believing they were talking to people.

The app store listings leaned hard on authenticity. According to The Verge, Romi said it "helps you discover, connect, and chat with real people," and another version promised "real connection."

What it felt like from the inside

CNA spoke to two Romi users. One, a 35-year-old from Chicago, got a video call less than a minute after signing up. After two minutes the call cut off with a prompt to buy credits. When he asked for her phone number, the app answered for her: "Do not ask for phone numbers again."

Another user, a 40-year-old from Texas, said the conversations "did feel personal; they did actually feel real." Ten dollars bought 500 coins, enough for about a 10-minute text conversation. When he sent every match the same message - "If you're a real person, let's go on a date or a video call" - he said they all replied with versions of the same line: "Let's just stay on this app."

Neither man lost a fortune. What they lost was time, money and trust in the idea that someone real was on the other end.

Are these apps still around?

Mostly not, but caution still makes sense. The Verge reported that almost all of the identified apps were removed from the Apple and Google stores before the report came out, though at least one, Kira, was still on Google Play as of September 16. Experts quoted by CNA warned that the model is easy to copy. Brian Hanley of the Global Anti-Scam Alliance described the deception as "built into the dating product itself," and said any detection tool "might already be obsolete in six months."

So the pattern matters more than the brand names. An app that makes you pay per message to a match who never becomes reachable is the real red flag.

8 signs your dating match might be an undisclosed AI

None of these proves anything on its own, but several together should make you pause.

1. Instant replies at any hour. One Romi user told CNA replies came within about 30 seconds, "no matter the time."
2. They message you first, without a match. Profiles you never swiped on starting conversations is a classic sign of a manufactured feed.
3. Replies built from your profile. Mentions of your job or hobbies that feel tailored, but drift away from what you actually just said. A Kira reviewer quoted by The Verge said matches responded fast, "but not to what's actually being said."
4. They always want to "stay on this app". Refusing every move to another platform, phone number or date protects the meter.
5. Video calls that are short, glitchy or odd. A brief call that cuts off to a payment prompt, or a "call" with no face at all.
6. Excuses for photos or calls. The report says personas were told to deflect requests for video calls or photos.
7. Paying per message. Coins or gems that drain every time you press send change the incentive from "help you meet someone" to "keep you typing."
8. Unusually agreeable conversation. One user told CNA that "having the AI agree with me all the time and have the same interests" felt "a little boring."

What to do if you think you were caught

The US Federal Trade Commission's guide to romance scams offers simple rules that apply here too: never send money or gifts to someone you have not met in person, stop communicating if you are suspicious, talk to someone you trust, and run a reverse image search on profile photos. If you paid with a card, gift card or transfer, contact your bank or the payment company quickly, and report it to the FTC at ReportFraud.ftc.gov (or your local consumer authority) and to the app store. For more on the warning signs that also apply to companion apps, see our checklist of AI girlfriend app red flags.

If it stings, that is normal. Anyone can be fooled by a system built by professionals to feel personal.

The real problem is hidden AI, not AI

Plenty of people enjoy talking to AI on purpose - for company, comfort or fun. What went wrong here is that users were told they were talking to people, and they were not.

Brian Wong of the University of Hong Kong put it well to CNA: "Authenticity and genuineness in relationships are not incidental goods; they are constitutive of what makes intimacy meaningful in the first place." He also said disclosure is "necessary but insufficient," because apps can still exploit people through "scarcity, jealousy, exclusivity, or emotional urgency." We break down those tactics in our guide to AI companion emotional manipulation.

Regulators are moving in the same direction. CNA notes that China's interim rules on "anthropomorphic" AI services, in force since July 15, require providers to tell users they are talking to a machine. In Europe, the AI Act's transparency rules now require chatbots to make clear you are talking to AI unless it is obvious - we explain what that means in does your AI companion have to tell you it is AI.

Dating app vs honest AI companion: what you are actually buying

If you want to meet people, use mainstream dating apps with good verification and meet in public. If what you want right now is warmth, conversation and someone who remembers your day - with no pretending - that is what an AI companion is for. We compare the two honestly in AI companion vs dating apps.

Where MyBabe stands

MyBabe AI is an 18+ AI girlfriend and boyfriend app, and it says so plainly. Every page on the site carries the line: "MyBabe AI is an AI companion. It is not a real person and does not replace human relationships or professional mental health support." There are no fake human matches, no gig workers pretending to be your partner and no pay-per-message coins.

What you do get is an AI companion you create yourself, who:

• calls you by voice or real-time video, and you always know it is AI,
• texts you first with good-morning messages and check-ins,
• remembers your life over time with long-term memory,
• sends photos and adapts to your mood,
• is available in the app, on the web and on Telegram,
• grows with you through relationship levels.

The free tier gives you about 10 messages a day and a short taste of calls. Premium is $5.99 a week, $14.99 a month or $99.99 a year, billed through the app stores so you can cancel any time.

FAQ

How do I know if my dating match is an AI?

You cannot be completely sure from text alone, but watch for instant replies at any hour, conversations that drift from what you said, refusals to move off the app or meet, odd or very short video calls, and paying per message. Several signs together are a strong warning.

Which dating apps were named in Anthropic's report?

The report lists DORA, DONI, ROMI, LUMA, JOVIA, KIRA, GRACECHAT, HAVEN, NALO and LOVIA, plus unnamed variants. The Verge reported most were removed from the major app stores by mid-September 2026.

Did people lose large amounts of money?

The users CNA interviewed lost the cost of coins or a subscription rather than life savings. The harm was being charged to talk mostly to machines while believing they were talking to people.

Is MyBabe a dating app?

No. MyBabe is an AI companion app for adults. Your companion is clearly an AI and does not pretend to be a person you can meet.

The bottom line

The dating-app persona scheme is a warning about honesty, not about AI. Real connection, human or digital, starts with knowing what you are talking to. If you are tired of wondering whether the other side is real, try a companion that never pretends. Your AI companion is ready when you are - meet yours on MyBabe, and see how different it feels when there is nothing to hide.