Explain the result a reader actually receives

Tool pages describe their own inputs and outputs. Headings should say what the tool does, and examples must be identified as examples. We do not add unsupported accuracy percentages, testimonials, usage counts or expert credentials.

Our face-shape profiles, score weights and similarity percentages are product choices. They must not be described as validated population norms or a probability that a person's classification is correct.

Distinguish data from demonstrations

  • Computed experiment: a result produced by specified code from published inputs. Current Camera Lab studies fall into this category.
  • Simulation: a simplified mathematical model with explicit assumptions. Its findings apply to that model.
  • Illustration: a diagram, synthetic portrait or mock result used to explain a concept. It is not evidence of model accuracy.
  • Photo study: would require an image set, permission, documented capture conditions, model version and a reproducible measurement procedure. The current computational studies are not photo studies.

Any future participant study must use consented adult images or clearly identified synthetic material. Synthetic image variations cannot establish the optical effect of a real focal length or camera distance without a physically specified rendering setup.

Publish enough to reproduce a numerical claim

Research pages include the question, inputs, method, calculated output and limits. Tables and downloadable data come from the same calculation code. A failed or inconclusive result should not be replaced with a more appealing invented number.

Our current experiments test the calculation layer. They do not measure how accurately MediaPipe locates landmarks across people, cameras or demographic groups.

Use sources for the claim they support

We prefer model documentation and original papers. Each reference explains how it is used and what it does not validate. A paper about a particular ratio or study sample is not evidence for a universal beauty standard.

Authorship, AI assistance and updates

AI Rate Face is the named publisher. AI-assisted drafting and coding may support the work. The publisher is responsible for verifying claims against the implemented method and cited sources. We do not invent a human reviewer or imply medical review.

Calculation dates identify when the published experiment was run. Content dates change for substantive updates, not merely to make pages appear fresh. The changelog records meaningful product and publication changes.

Corrections and commercial separation

Send the page URL, the disputed statement and supporting details to contact@airateface.com. Please do not include private photographs in the initial message. Substantive corrections should identify what changed and update affected data or examples.

The optional paid plan has its own pricing and consent flow. Free-tool explanations and research remain readable without purchasing it. Sponsored material, if introduced, must be clearly labelled.