AI Catfishing: The Trust Crisis Dating Apps Can't Ignore

The gap between perceived and actual ability to spot AI-generated dating profiles has widened into a chasm, according to new research that should alarm trust and safety teams across the industry. A study by fashion brand Pour Moi tested 3,000 people's ability to identify synthetic profiles—and 62% failed, despite 57% expressing confidence they could spot the fakes. The discrepancy isn't just embarrassing—it's a material risk to user retention and platform integrity at a moment when trust in dating apps sits at historic lows.

If users can't distinguish real prospects from AI-generated phantoms, they'll waste time, emotional capital, and potentially money on connections that don't exist. That erodes the core value proposition of every dating platform: facilitating real human connection.

Create a free account

Unlock unlimited access and get the weekly briefing delivered to your inbox.

If 62% can't spot fakes but 57% think they can, members won't report what they don't recognise. That leaves trust and safety teams flying blind whilst AI catfishing scales.

The confidence gap identified here means platforms can't rely on user vigilance as a defence layer.

The overconfidence problem

Pour Moi's methodology deserves scrutiny—this is a fashion brand with commercial interest in dating dynamics, not a peer-reviewed study. But the core finding aligns with what compliance teams already know: members consistently overestimate their fraud detection abilities.

The age breakdown amplifies concern. According to the survey, 32% of respondents aged 55 and older believed all the profiles shown were real—the highest error rate of any demographic. This group also represents the fastest-growing segment of dating app users and, according to Federal Trade Commission data, reports the highest financial losses to romance scams.

Match Group disclosed in its Q3 2024 earnings call that Hinge's fastest-growing user cohort is now 30-plus. The intersection of AI sophistication and digital literacy gaps creates acute vulnerability.

Twelve percent of respondents claimed they'd been catfished previously. The figure lacks context—timeframe, platform, verification of claims—but it suggests baseline trust erosion before generative AI entered the equation. What happens when tools that can create convincing synthetic photos, write contextually appropriate messages, and maintain consistent persona become accessible for £15 per month?

Where the platforms stand

No major dating app has deployed systematic verification for AI-generated profile content. Tinder's photo verification relies on real-time selfie matching, which stops stolen photos but doesn't detect synthetic ones. Bumble introduced selfie verification in 2016 and government ID checks in select markets, but these confirm the account holder matches the photos—not whether the photos depict a real person.

Hinge offers video prompts, which raise the bar for fraudsters but remain spoofable with deepfake tools now widely available.

Traditional catfishing required effort: curating stolen photos, maintaining narrative consistency, managing multiple conversations. AI tools collapse those friction points.

The industry's current approach treats AI catfishing as an extension of existing fraud rather than a categorically different threat requiring new technical controls. That's a miscalculation. A determined bad actor can now operate dozens of convincing synthetic profiles simultaneously, each with unique photos, coherent backstories, and contextual messaging.

Grindr disclosed in its Q4 2024 shareholder letter that it blocked 1.2 million fake accounts in the quarter—a 34% increase year-over-year. The company attributed the rise to 'increasingly sophisticated automation tools' but didn't specify AI involvement. Match Group reported removing 'tens of millions' of bad actors annually across its portfolio but hasn't broken out AI-specific threats in public disclosures.

What actually works

Detection will require layered technical controls, not reliance on user vigilance. Computer vision models can identify synthetic images with reasonable accuracy—academic research published in IEEE Transactions on Information Forensics and Security in late 2024 showed 94% accuracy detecting DALL-E and Midjourney outputs. But these tools require continuous retraining as generative models improve, and they're computationally expensive to run at scale.

Behavioural signals offer another vector. Synthetic profiles controlled by AI typically show temporal patterns inconsistent with human behaviour: perfectly distributed response times, absence of typos or autocorrect errors, conversational patterns that stay oddly on-topic. Bumble's trust and safety team confirmed to investors in November 2024 that it's testing behavioural analysis for bot detection but hasn't extended this to AI-specific threats publicly.

The verification arms race will likely mirror the path email took with SPF, DKIM, and DMARC—technical standards that authenticate sender identity. Dating apps will need equivalent frameworks: cryptographic proof that profile photos originated from device cameras, not generative models; metadata verification; continuous authentication during messaging. This infrastructure doesn't exist yet, and building it will require industry coordination that hasn't materialised.

Regulatory pressure may force the issue before market dynamics do. The UK Online Safety Act requires platforms to prevent users encountering fraudulent profiles. The Act doesn't specifically address AI-generated content, but Ofcom's draft codes of practice suggest that 'reasonably foreseeable risks' include synthetic identity fraud. If regulators determine that AI catfishing constitutes a foreseeable harm, platforms without adequate controls face enforcement action.

The Pour Moi survey, methodological limitations notwithstanding, quantifies what trust and safety teams already suspect: user-side detection has failed. The confidence-reality gap means members won't sound the alarm because they don't know there's a fire. Platforms that continue treating AI-generated profiles as a marginal fraud vector rather than a systemic threat to platform integrity are storing up retention and regulatory problems that will prove expensive to fix retrospectively. The question isn't whether to invest in AI-specific verification infrastructure. It's whether operators move before users—and regulators—force their hand.