Methodology

Our Search Methodology

A plain description of what our search does, so you can judge the results for yourself. If something here is wrong or out of date, email us and we'll fix it.

CheaterFinder Editorial·Last reviewed November 3, 2026

Where results come from

Every search draws on up to three sources:

  1. Google web results, retrieved through a search API. This is the same index you could search yourself; we just run many targeted queries at once.
  2. A commercial face-search service, used only if you add a photo. It compares the face in your photo with faces it has indexed from public web pages.
  3. An AI language model, which reads the results from the first two sources together with the details you gave us. It doesn't search anything itself.

We don't have private access to Tinder, Bumble, Hinge or any other app. Nobody legitimate does. Dating apps don't offer search access to outside services, and getting into one without permission would break both the app's terms and, in many places, the law.

How queries are built

From your details we build up to 16 queries and run them in parallel. Depending on what you gave us, they include:

  • The exact name in quotes (plus the city, if you gave one) combined with dating terms, dating app names, Instagram and TikTok, LinkedIn, and Reddit and forum discussions
  • The name in the "Name, age" format many dating profiles display
  • Site-restricted searches on the apps you said you suspect (tinder.com, bumble.com, hinge.co, instagram.com)
  • Each username you gave, exactly, and restricted to Instagram and TikTok
  • The phone number in its exact format, if it has at least seven digits

Duplicate links are merged and the first 40 unique results go to analysis. Each result keeps its original title, link and snippet so you can see exactly what Google returned.

How face matching works

  1. 01

    Upload

    Your photo is stored with your search and sent to the face-search provider.

  2. 02

    Face found

    The provider detects the face and converts it into a numeric description of facial features.

  3. 03

    Compared

    That description is compared against faces from public web pages it has indexed.

  4. 04

    Scored

    Each candidate gets a similarity score from 0 to 100. We keep the top 10.

How a photo becomes a list of visual matches.

A similarity score measures how alike two faces look, not whether they're the same person. High scores between relatives and lookalikes are common, especially with small, blurry or side-on photos. That's why visual matches appear in the report with their raw score and a link, and why the summary treats a face match as one signal rather than a verdict.

The photo quality checker tells you whether a photo is likely to match well before you search.

How results are weighed

The AI model receives your details and every result, and it follows a fixed set of rules:

  • A result needs at least two independent matching details (for example name and city, or name and a matching photo) to be treated as a likely match.
  • Results where the age is more than about four years off, or where the gender clearly doesn't match, are down-ranked.
  • An exact username match is treated as a strong signal.
  • A single weak signal, such as a common first name alone, is never treated as proof.

It returns the likely profiles, a short summary and an overall level (low, medium or high) describing how much public evidence it found. That level is about the strength of the evidence, not a judgement about the person or the relationship.

What confidence means

Each profile in the report carries a percentage. For visual matches it's the raw face-similarity score. For other profiles it's the analysis's estimate of how well the result fits your details. Either way, read it as a position on this ladder, not as a probability that the person is cheating.

Possible

One detail lines up, usually the name. Treat it as a lead to check, nothing more.

Plausible

Two details line up, such as name and city, but nothing ties the profile to the person directly.

Likely

Several independent details agree, or a username matches exactly.

Strong

A strong face match plus matching details, or a profile that shows something only the real person would. Still verify before acting.

How to read the confidence attached to a result.

Where the search fails

We'd rather you know the limits up front:

  • In-app-only profiles. Most dating profiles are never indexed by Google. If a profile only exists inside an app, a web search can't see it.
  • Fake names and no photo. If someone uses a different name and you have no photo, there's very little to match on.
  • Very common names. Without a city, age or photo, results will include many unrelated people.
  • Timing. A result can show that a profile existed, but rarely when it was last active.
  • Search index lag. New profiles take time to appear in Google, if they appear at all, and deleted ones can linger.

How we improve it

We record anonymous counts for each search: which kinds of details were provided, how many results and face matches came back, and the overall evidence level. No names, photos, handles or locations are kept in those counts. They tell us which inputs genuinely improve results, and they're what our advice about photos and details is based on.

When the method changes in a way that affects how you should read results, we update this page and the date at the top.