Romance Scams Are Using AI Now — Here's What That Means
Mark · SUS IT Editorial Team
Mark reports on social media fraud, bot ecosystems, and online scams.
How AI-generated photos, automated chatbots, and language models have supercharged romance fraud — and the detection signals that still work.
Romance scams cost victims more than $1.3 billion in the United States alone in 2023, according to FTC data — and that number has been climbing every year. The introduction of generative AI into the fraud toolkit has made these scams simultaneously cheaper to operate, easier to scale, and harder to detect. Understanding what's changed, and what hasn't, is the first step toward protecting yourself.
What Romance Scammers Used to Do
The original romance scam formula was labor-intensive. A scammer would steal photos from a real person's social media and manually message targets on dating apps or social platforms. Conversations had to be conducted by a human, which limited how many victims could be worked simultaneously. At scale, it was still devastating — but it had a ceiling.
How AI Changed the Economics
Generative AI has removed most of the labor bottlenecks. Profile photos are now synthesized using models like Midjourney or purpose-built face generators — producing unique faces that don't belong to any real person and therefore can't be traced via reverse image search. Conversation management is increasingly handled by large language models fine-tuned on romance scam scripts, capable of maintaining emotionally engaging dialogue across hundreds of simultaneous targets. Human operators now supervise and escalate only the most promising targets, with AI handling the rest.
The Synthetic Persona Stack
A sophisticated AI-assisted romance scam uses multiple components working together. The face is generated fresh for each operation. The persona backstory is crafted by a language model: a widowed engineer on an oil platform, a military surgeon deployed overseas, a successful entrepreneur traveling for business. A thin social proof layer — a sparse LinkedIn profile, a few Instagram posts — is populated with AI-generated content. Conversation is conducted through scripted templates and AI generation, with the model adapting based on the target's responses to maintain emotional engagement.
The Money Ask: How It Evolves
Romance scammers follow a consistent financial playbook. The relationship is established over weeks or months before any financial request appears. This investment period makes the eventual request feel like an anomaly — "I've never asked anyone for help before" — rather than the goal. Common scenarios include medical emergencies, import taxes on shipped packages, business bridge loans, or travel costs blocked by a sudden problem. Amounts often start small to establish a compliance pattern before escalating.
Detection Signals That AI Hasn't Fixed
Despite AI's improvements, several detection vectors remain reliable. Video call resistance is the most consistent: AI cannot yet convincingly conduct a real-time interactive video conversation. Deepfake video exists, but responding naturally to unexpected live requests ("hold up today's newspaper", "wave at me") remains beyond deployed scam technology. Biographical inconsistency is another weakness: language models don't maintain persistent memory, so details claimed early (specific schools, neighborhoods) may shift later. Cross-platform absence is a strong signal: real people almost always have some corroborating web presence. AI personas rarely do.
How to Check a Suspicious Profile
If you're communicating with someone who raises even minor concerns, treat verification as normal. Request a video call. Search their claimed name and employer together across Google, LinkedIn, and Facebook. Run their profile through BOT or NOT at susit.ai if you have a share link — the tool searches for evidence of a real person and flags patterns consistent with synthetic or fraudulent accounts. If they've sent photos, a reverse image search still catches stolen photos (as opposed to AI-generated ones), and it costs you nothing.
The Arms Race
As AI makes scam profiles more convincing, AI-based detection is improving in parallel. Detection tools that analyze cross-platform web presence, behavioral consistency, and known fraud patterns can surface signals that human instinct misses — especially after weeks of emotional investment have made objective assessment harder. The fundamental dynamic of the scam — building trust before making a financial request — has remained constant for decades. Understanding the playbook is the best defense regardless of which tools are being used to run it.