How to Tell If You Are Talking to an AI on a Dating App
AI chatbots on dating platforms write better English than most people, remember your conversations perfectly, and are available around the clock. These same qualities are the giveaway. Here is how to test for them.
How can you tell if someone on a dating app is an AI?
No single test is conclusive, but a combination of behavioural and conversational checks produces a reliable picture. The strongest tests exploit the fundamental limitations of language models: inability to be genuinely spontaneous, difficulty with real-time physical verification, and complete absence from real-world public records.
Run through the checklist below before investing significant emotional or financial energy in any online relationship where you have not met the person in real life.
The Checklist: 10 Tests to Run
- Response timing consistency. Note what time you send messages and how quickly replies arrive over several days. Human responses vary with sleep, work, and social schedules. AI responses arrive within a consistent window regardless of the hour.
- Spontaneous topic initiation. Does the persona ever bring up a genuinely new subject unprompted, or do their messages always respond to what you said? AI grooming scripts are reactive — they rarely generate spontaneous conversational pivots.
- Hyper-specific environmental questions. Ask what they can see from their current position right now. Follow up immediately with a detail question. Real people describe messy, specific, slightly mundane reality. AI produces a plausible scene.
- Long-gap memory test. Reference something minor they mentioned three or four weeks ago. AI context windows vary — some models lose early conversation detail. A human romantic interest remembers minor things because they mattered emotionally.
- Genuine absurdity test. Say something mildly absurd or make an unexpected joke. Assess whether the response feels naturally surprised or carefully managed and redirected.
- Slang and regional language. If they claim to be from a specific city or country, use local slang or a region-specific cultural reference and ask if they know it. AI personas are trained on general internet text, not local cultural specifics.
- Unscripted video call. Request a video call with one hour or less notice. Ask them to perform a specific physical task you name during the call — hold up three fingers, write a word on paper, pick up something blue. Real-time deepfake generation cannot reliably handle compound spontaneous physical tasks.
- Reverse image search all photos. Run every profile image through Google Reverse Image Search, TinEye, and Yandex. AI-generated faces that have not been published elsewhere will not return results — but that absence combined with other signs is itself a flag.
- Consistency cross-check. Keep a simple note of key biographical claims: employer, city, age, family details. Test the same facts in different conversations weeks apart. AI models can produce consistent but not always identical versions of a stored backstory.
- Identity verification. Commission a professional check on the claimed name, employer, and location against real public records. This is the only test with a definitive binary outcome.
When Tests Are Inconclusive
A skilled human scammer can also fail spontaneity tests. And a well-configured AI can pass some of the behavioural checks above, particularly in the early stages of a conversation before its scripted patterns become visible. If you are uncertain after running these tests, the right next step is professional identity verification — not continued testing. AllRussian can determine within 48 to 72 hours whether a claimed identity corresponds to any verifiable real person in CIS and Eastern European public records.
Order an Identity Check