♥MatchAtlas
Editorial methodology

How MatchAtlas evaluates dating services

Our goal is to separate editorial usefulness from affiliate economics. A service can only receive a published score after its factual claims and scoring inputs have supporting evidence.

Evidence firstMaterial claims require a source or direct product test.
Affiliate-neutral scorePayout, EPC and commission never enter the editorial rating.
Market-specificA service may perform differently by country, age group and intent.
Refreshable dataReviews record when important facts were last checked.

Our scoring model

Scores use a 0–100 scale and are weighted across eight criteria. The weights are stored in the same structured data model used by the ranking engine, so a published ranking can be reproduced from reviewed inputs.

Audience fit · 20%How well the service's audience matches the target age group and use case.
Relationship-intent fit · 18%How clearly the product supports serious, casual or exploratory dating goals.
Safety & trust · 16%Verification, reporting, blocking, privacy controls and user-safety information.
Onboarding & discovery · 12%Registration friction, profile setup, search and discovery experience.
Communication · 12%Messaging and interaction tools relevant to starting and continuing conversations.
Account control · 8%Clarity around settings, cancellation, deletion and user control.
Mobile experience · 8%Usability and consistency on mobile devices.
Value transparency · 6%How clearly free and paid functionality is explained.
Commercial firewall: affiliate payout, EPC, conversion rate and commission can be used later for traffic routing only among editorially eligible options. They do not increase an editorial score.

What counts as evidence

We prioritise first-party product pages, help centres, terms, pricing information and direct testing for factual claims. Each research record can store the source URL, access date and the exact claim it supports. We do not publish invented user counts, demographics, prices, safety claims or feature lists.

AI-assisted, not AI-autopublished

AI may help structure research, identify missing evidence, detect possible keyword cannibalisation and draft copy. It does not have permission to turn an unverified research record into a published review. Drafts with unresolved evidence requirements remain blocked.

Local and programmatic pages

Changing a city, country or age label is not enough to create a useful page. Programmatic pages remain noindex until they contain materially distinct, reviewed value. Pages that overlap in search intent should be merged rather than multiplied.

Corrections and updates

Material errors should be corrected promptly. When a service changes pricing, onboarding, features or policies, affected claims and scores should be rechecked. Pages that lose useful value can be updated, consolidated, redirected or removed from indexing.