How We Review Dating Apps and AI Companions

A dating app review is only as useful as the method behind it. A score of 9.3 means little unless you know what was measured, how much each element counted and where the information came from. This page sets out exactly that.

Our reviews are editorial assessments of verified information. We do not run laboratory tests, recruit panels of testers or invent anecdotes about weeks spent swiping. Instead, we gather what each company publishes about its own product, check it against independent sources, and weigh it against a consistent set of criteria. Where we rely on a company’s own description, we say so.

The four criteria behind every dating app score

Each mainstream dating app is scored out of 10 on four criteria. Those scores are then weighted and combined into the overall figure you see at the top of a review.

Match quality (35%)

The heaviest weighting, because this is the reason anyone downloads a dating app. We look at how the service builds a profile, whether it encourages substance (prompts, questionnaires, values) or rewards quick judgements, what the company discloses about how its matching works, the filters available to free and paying members, and the size and focus of its user base where reliable figures exist, such as those in investor reports.

Features (25%)

What you can actually do in the app: messaging tools, voice and video calling, ways to signal interest before a match, profile prompts, events or group features, and the quality of the design on iOS and Android. We also consider how much of this is locked behind a subscription, since a feature you cannot use without paying is worth less to most readers.

Safety and verification (25%)

How the service protects people, particularly women and LGBTQ+ users, who face disproportionate harassment online. We examine photo or video verification, reporting and blocking tools, moderation policies, in-app safety guidance, privacy controls such as hiding a profile or concealing distance, and the company’s record with regulators and data protection authorities.

Value for money (15%)

What the paid tiers cost, what they add and whether the free version is genuinely usable. Prices vary by country, age and promotion, so we quote verified prices with the market and source attached. Where we cannot verify a price, we describe what the tier includes without one. We also note auto-renewal terms and how easy a subscription is to cancel.

What our 10-point scale means

The overall score is the weighted average of the four criteria, rounded to one decimal place. In practice the bands read like this:

• 9.0 to 10: Outstanding. Among the best services available for the people it targets, with few meaningful weaknesses.
• 8.0 to 8.9: Very good. A strong choice with clear strengths and some limitations worth weighing.
• 7.0 to 7.9: Good, with caveats. Worth considering for particular needs, though better options exist for most readers.
• 6.0 to 6.9: Mixed. Significant drawbacks in at least one criterion.
• Below 6.0: Not recommended. Serious concerns about safety, value or honesty.

Scores compare apps within the same category. A niche community app is judged on how well it serves its audience, not on whether it can match the sheer scale of the largest services.

How we judge AI companion apps

AI companions are a different product with different risks, so they get a separate framework. Every app in our AI companions coverage is scored on five criteria:

• Memory and conversation (30%): how well the companion holds context, recalls earlier exchanges and sustains a coherent personality.
• Customisation (20%): control over the character’s personality, appearance, voice and boundaries.
• Privacy and data (20%): what the company says it collects, whether conversations are used to train models, how data can be deleted and where it is stored.
• Price (15%): what is free, what costs extra and whether subscriptions or in-app currencies are clearly explained.
• Age safeguards and wellbeing (15%): how the service keeps under-18s out, how it handles sensitive topics and whether it signposts support.

The same 10-point bands apply. Our AI companion coverage is for adults and is kept non-explicit. An app with weak age checks or opaque data practices cannot score highly, however engaging its conversation.

The sources we rely on

Every fact in a review must come from a source we have opened and checked. In order of preference, those are:

• Official help centres, pricing pages and terms of service, which tell us what a company commits to in writing.
• App Store and Google Play listings, for features, in-app purchase prices and published user ratings. When we cite a rating, we give the store and the date.
• Investor reports and annual filings from listed companies such as Match Group and Bumble Inc., for user numbers, revenue and strategy.
• Regulators and government bodies, including data protection authorities, consumer protection agencies and online safety regulators, for enforcement action and official guidance.
• Published research from universities, peer-reviewed journals and established survey organisations. Surveys paid for by dating companies are labelled as such.

Reputable news reporting provides context, particularly for breaking stories. Our editorial policy explains these sourcing rules in more detail.

How often reviews are updated

Dating apps change constantly. We review every scored page at least every six months, and sooner when something material shifts: a price rise, a new subscription tier, a change of ownership, a significant safety feature or a regulatory finding. The date at the top of each review shows when it was last checked.

If an update changes a score, the new figure applies straight away and our best dating apps ranking is adjusted to match.

Money and scores

Some links on On Dating Journal earn us a commission, as our affiliate disclosure explains. Commercial relationships have no influence on scores or rankings. An app that pays us nothing can top a list, and an app that does pay us can be marked down.