Safety Claims About Online Dating, Checked Against the Data

Is online dating safe? The honest answer is: safer than its reputation suggests in some ways, and worse than its marketing suggests in others. The claims that circulate — "everyone gets harassed," "verified badges mean verified people," "matchmakers are safer by default" — are each partly true and partly folklore. We pulled five of the most repeated ones and checked them against Pew Research, FTC complaint data, and the academic literature on trust and safety in online introductions.

This matters more than usual right now. Justin McLeod — the founder of Hinge — just launched Overtone, a matchmaking app with Esther Perel on its board, built explicitly around the idea that profiles and open decks create exposure that most people never signed up for. Safety isn't a side conversation anymore. It's part of why the founder of the category's biggest swipe app walked away from swiping.

Claim 1: "Most online daters experience harassment"

True, and not close. Pew Research Center's 2023 survey found that online dating harassment is common but unevenly distributed: roughly one in three users overall report being called an offensive name, and larger shares of women under 35 report unsolicited explicit images, sustained unwanted contact, or physical threats. Women under 35 are the group most affected by a wide margin.

The nuance the headline number hides: this isn't randomly distributed bad luck. It's a structural feature of open, low-friction contact — anyone can message anyone, at any hour, with no accountability beyond a report button that acts after the fact.

Claim 2: "Verified profiles mean you're talking to a real, safe person"

Partly true, mostly misleading. Photo verification (the blue-checkmark-style selfie match most apps use) confirms that the person in the photos is the person behind the account. It says nothing about intent, criminal history, marital status, or whether the account is being run in good faith.

The Federal Trade Commission's consumer sentinel data shows romance scam losses in the billions annually, and a meaningful share of those scams originate on platforms with photo verification in place. Verification proves identity in a narrow, technical sense. It was never designed to prove trustworthiness, and treating it as a safety guarantee is where the myth breaks down.

Claim 3: "Matchmakers and invite-only services are automatically safer"

Directionally true, but the mechanism matters more than the label. An invite-only structure raises the cost of a bad actor getting in — someone has to be admitted, not just download an app — which is a real filter, not a cosmetic one. Stanford sociologist Michael Rosenfeld's long-running couples-formation research shows how thoroughly meeting-online has displaced other paths to partnership; it doesn't, by itself, measure safety outcomes, but it does establish that the mechanism of introduction shapes who you end up talking to at all.

Where this claim overreaches: "invite-only" isn't one thing. A paid membership with a credit card on file is a weaker filter than an actual human agent reviewing a briefing before making an introduction. The safety gain scales with how much friction and judgment sits between "anyone can join" and "you get introduced." We went deeper on this distinction in our invite-only dating research breakdown.

Claim 4: "AI matchmaking means an algorithm is deciding who's safe to meet"

False, and worth correcting precisely because it sounds plausible. No credible agent-mediated service — including the one you're reading this on, including Overtone — claims an algorithm performs background checks or certifies someone as safe. What AI matchmaking actually changes is exposure: instead of a public profile any stranger can screenshot, message, or scrape, your information sits behind an agent that only surfaces it for a deliberate, explained introduction.

That's a privacy design choice, not a safety certification. It's a meaningfully different attack surface than a public swipe deck, but it's not a background-check service, and no honest version of this category claims otherwise.

Claim 5: "This is a bigger problem for women than the data shows"

False — if anything, Pew's numbers understate it rather than overstate it. Self-reported harassment surveys typically undercount because normalization suppresses reporting: behavior that would register as harassment in another context gets waved off as "just how the apps are." Researchers studying online harassment broadly (not dating-specific) have found similar undercounting patterns tied to platform normalization, which means the Pew figures are more likely a floor than a ceiling.

This is one of the clearest cases where the popular belief isn't exaggerating a problem — it's actually being too conservative about one.

Claim 6: "There's nothing you can personally do about any of this"

False, though the useful actions are structural, not just behavioral. The research consistently points to a few things that move the needle:

• Reduce public surface area. Fewer places your photos and details sit exposed to anonymous browsing means fewer entry points for bad actors.
• Prefer introduction over open contact. Any structure requiring a deliberate step — a match, an agent, a mutual connection — before messaging is possible filters out a meaningful share of low-effort harassment.
• Treat verification as a floor, not a ceiling. Use it, but don't let a badge substitute for your own judgment on video calls or first meetings.
• Report even when it feels pointless. Aggregate report data is one of the few inputs that pushes platforms to change defaults.

What Holds Up, What Doesn't

1. Harassment rates are real and higher than casual conversation admits — confirmed.
2. Verification badges are frequently mistaken for safety guarantees — myth.
3. Invite-only structures help, proportional to how much human judgment is actually involved — confirmed, with conditions.
4. AI matchmaking algorithms "deciding who's safe" — myth, and not one any serious player in the category actually makes.
5. Women's experiences being overstated in the data — myth; the real risk is understatement.

Where This Leaves the Question

Is online dating safe? Safer than the discourse implies for people willing to change how they participate in it, and less safe than the badge-and-checkmark design suggests for people who assume the platform has already handled it. The through-line across every claim we checked: friction and judgment before contact consistently outperform verification after the fact.

That's the design principle behind agent-mediated matching generally — an agent reviews your briefing, decides who's worth introducing, and explains why, rather than exposing a profile to open browsing. It's one reason the model is gaining ground beyond just fatigue with swiping. If you're evaluating whether that trade-off suits how you want to date, neverswipe is live today and built around exactly this structure.

Frequently Asked Questions

Is online dating actually dangerous?

The data shows meaningful risk of harassment, particularly for women under 35, but not evidence that online dating is more dangerous overall than other ways of meeting people. The risk profile is different — more exposure to strangers, less accountability — not necessarily higher in aggregate.

Do dating app verification badges actually work?

They confirm the photos match the person, which does reduce catfishing. They do not vet intent, history, or honesty, and shouldn't be treated as a safety guarantee.

Are invite-only apps safer than open dating apps?

Generally yes, proportional to how much real screening happens before you're let in. A credit card paywall is a weak filter; an agent reviewing your briefing before making introductions is a stronger one.

Can AI matchmaking prevent romance scams?

It reduces certain exposure — fewer strangers can find and contact you unsolicited — but no responsible AI matchmaking service claims to eliminate scam risk. Ordinary caution on video calls and early meetings still applies.

Why do harassment numbers seem low compared to personal experience?

Self-reported survey data tends to undercount because normalized bad behavior often goes unreported. Most researchers treat these figures as a floor, not a full picture.