Keeping a marriage-first dating app free of scammers

A new match looks perfect: a charming serviceman with a clear plan for the future. The photos are stolen or AI-generated, the name is invented, and within a few messages he pushes the chat over to WhatsApp. Early on, many matches at True for Two involved an account the team would later ban.

These accounts reach real people who signed up to find a husband or wife, and that trust is the whole product. "We had to be sure that nothing trash or illegal gets through," says Chief Product Officer Natalia Sergeeva. Explicit photos and fake profiles break the promise the moment one slips past, and the team could not keep up by checking every profile by hand.

Lasso analyzes every message and photo as it arrives, and accounts are automatically flagged or pulled before a real user swipes past. The platform facilitates custom rules, photo matching, a scammer score, selfie verification, and face detection, with user blocks and statuses handled in the same dashboard. One moderator works through whatever the automation cannot decide on its own.

Matches with a banned account fell by more than 90%. Scam profiles now get banned within an hour instead of a day, and a single moderator covers the queue in four to five hours a day.

A dating app with one job

Most dating apps let you tick a box for "marriage" or "long-term," then drop you into a mixed crowd looking for everything from friendship to something casual. Natalia Sergeeva, Chief Product Officer at True for Two, says that the product was built around that mismatch.

"When you want something committed and you go to any of these apps, you carry extra work," she says. "Does this person actually want what I want? We designed the app for the people who already know the answer."

This is why True for Two opens with a questionnaire. It works as a filter and as a source of truth. Someone who is not serious rarely finishes it. The serious ones give the app the details that decide a match: whether they have kids, whether they want kids, where they live, whether they would like to move.

Buying instead of building

The team reached the moderation question fast. Hand moderation was never considered. "It is 2026. Doing everything by hand is slow and unreliable." Writing their own detection software was not the plan either, because the team did not have the expertise to build and maintain it.

Then came the part most teams underestimate. Catching bad content is only half the job; someone still needs a place to review it, see a user's history, and block or clear an account. True for Two expected to build that admin layer themselves, on top of the detection. Lasso already had both, in one platform.

"I used to work as an admin, so I know how bad these interfaces get," she says. "When we saw Lasso, we realized we didn't need to build one. We use theirs, and we just send API requests. It saved us a lot of effort, time, and money."

Catching scammers, cats, dogs, and fish

As the platform grew, scammers arrived in volume, and the operations were organized: many devices, a stream of invented identities, photos lifted from real profiles or generated by AI.

Scammers betray themselves through repetition. They run the same scripts to pull people off the platform, recycle the same faces across fresh accounts. True for Two turned this information into a signal that Lasso could act on.

A reused line like "let's move to Signal" became a rule that bans on sight. "In 99% of cases, that is a scammer," Natalia says. A recycled face became a match against an already-banned account. A cluster of suspicious behaviors became a single risk score, passed to Lasso as a custom field, that decides whether an account is flagged or removed.

Selfie verification came next, as a separate flow. The team runs it as an experiment on half the users and forces it on anyone who looks suspicious, using a "verification required" tag that their own backend reads.

Not every rejected photo comes from a scammer. Plenty of people upload a pet or a trophy catch instead of their face. "There is a whole group of proud fishers," Natalia says. "Me and my fish. Different people, same photo." When a photo has no human face, the app rejects it automatically. It is one of their most common rejections.

Custom setup for each challenge

The team can shape Lasso to fit, without touching their own code.

When they wanted to catch more scammers, they dropped the risk score threshold in Lasso custom rules from 100 to 80 with one click. "Nothing changed on our backend," Natalia says.

When a full-body photo had a face too small for the system to catch, like someone in front of the Eiffel Tower, she asked, and the Lasso team tuned it to detect the smaller face. "It really helped, and I am happy it is that customizable."

Requests from her moderator sometimes ship within a day. When she gets stuck, she emails, and the answers come back. For a small team that would rather build its product than its tooling, that responsiveness is the point.

What changed

Natalia watches two numbers. The first is the time between a sign-up and a ban. Before moderation, a scam account could live more than a day. Now it goes within the hour.

The second measures harm: how often a banned account had already matched with a real user. That share has dropped by more than 90% since the team put moderation in place. "It is not only Lasso," she says. "It is the mix of solutions we built. But without passing our data into the system and acting on it automatically, none of it would be reachable."