How we test LooksCompare
Our manual process, what counts as a useful check and why we do not publish unsupported accuracy percentages.
Read testing process →Practical LooksCompare guides about visual similarity, family resemblance, photo quality, score interpretation, privacy and responsible testing — plus original development notes that explain how we challenge the tool with known examples.
These pages describe tests we can perform only because we operate LooksCompare: known same-person age-gap checks, look-alike negative tests and controlled photo-condition changes. They are not a claimed biometric benchmark; they document our real testing process and its limits.
Our manual process, what counts as a useful check and why we do not publish unsupported accuracy percentages.
Read testing process →A known same-person stress test across large age gaps, glasses, headwear and camera differences.
Read test note →Why similar-looking but different people are essential negative examples for same-person comparison.
Read test note →How we isolate photo conditions to understand why an image-based similarity score can move.
Read test note →How facial information becomes a numerical representation used for visual comparison.
Read guide →Lighting, angle, sharpness and face position can significantly affect an estimate.
Read guide →What a percentage can tell you — and what it cannot prove.
Read guide →Why resemblance scoring and identity recognition are different tasks.
Read guide →How lighting, pose, lens perspective and expression can move a result.
Read guide →Photo and interpretation mistakes that most often confuse results.
Read guide →Why a child can resemble both parents in different traits and at different ages.
Read guide →How an AI mom-vs-dad comparison works and how to interpret close scores.
Read guide →Privacy, permission and responsible interpretation when children's photos are involved.
Read guide →Why growth and aging can change a visual comparison over time.
Read guide →A practical experiment showing how photo choice alone can move a similarity result.
Read guide →Why a precise-looking percentage is not genetic relatedness or identity probability.
Read guide →Ten variables that can change what a family resemblance comparison sees.
Read guide →Why resemblance to mom or dad can appear to change as a child grows.
Read guide →A repeatable method for checking stability without chasing a preferred result.
Read guide →Verified Android release information and how future app updates will be documented.
View Android app history →