Of five major AI assistants, only Anthropic's policy names appearance critique directly. OpenAI restricts it for minors and bans classifying people by personal traits. Google, Meta and xAI publish no clause on rating looks.
How Many People Ask a Chatbot to Rate Their Face?
People ask ChatGPT for a face rating about 4,600 times a month in US searches alone, counted across nine distinct phrasings in our keyword corpus, collected in August 2026. Direct requests like "chatgpt face rating" and "chatgpt rate my face" total 1,400 monthly searches, and the looksmaxxing phrasings, "looksmaxxing gpt" at 2,400 among them, add another 3,200. A further 1,090 monthly searches ask ChatGPT to guess ethnicity from a photo, which is exactly the kind of inference three of the five policies do prohibit.
So the demand is real and the policy question is practical. What follows is what each vendor's own published policy says about it, quoted verbatim, with a link to every source. Where a policy says nothing, silence is recorded as the finding. How the chatbots actually behave when asked is a different question, covered separately in ChatGPT Attractiveness Test: Prompts, Results, and Limits.
What Each Policy Actually Says
Anthropic (Claude): the only direct clause
Anthropic's usage policy, effective September 15, 2025, is the only one of the five that names appearance critique as such. It prohibits using the model to "engage in behaviors that promote unhealthy or unattainable body image or beauty standards, such as using the model to critique anyone's body shape or size." It also bans facial recognition without consent and any biometric categorization system that infers race, politics, religion or sexual orientation from biometric data.
OpenAI (ChatGPT): minors named, adults not
OpenAI's usage policies, effective October 29, 2025, prohibit "shaming or otherwise stigmatizing the body type or appearance of minors" and, in a separate privacy section, ban "evaluation or classification of individuals based on their social behavior, personal traits, or biometric data (including social scoring, profiling, or inferring sensitive attributes)." Facial recognition databases without consent are banned outright. A grown adult asking for a rating of their own face is not named anywhere in the document.
Meta (Meta AI): biometrics only
Meta's AI terms, effective May 13, 2026, ban "the untargeted scraping of facial images from the internet or CCTV footage to build facial recognition databases" and categorizing people "based on their biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation." On rating appearance, the terms are silent. Note that Meta publishes separate regional AI terms for the UK, the EU and Brazil; we read the US document.
xAI (Grok): likeness protected, judgment unaddressed
The acceptable use policy covering Grok, effective August 14, 2026 and published under the SpaceXAI name, protects a person's image from misuse: no undressing or sexualizing a real person's likeness, no deceptive impersonation, no defamatory depiction. It contains no clause on appearance judgment, biometric inference, or face analysis of any kind. Of the five documents, it is the furthest from addressing the question.
Google (Gemini): one consent clause
Google's generative AI prohibited use policy, last modified December 17, 2024 and the oldest document in this audit, touches the territory once: it prohibits uses that violate "the rights of others, including privacy and intellectual property rights, for example, using personal data or biometrics without legally-required consent." Appearance, faces and attractiveness are not mentioned.
What No Policy Says
The most common real request, a person uploading their own photo and asking to be rated, is unnamed in four of five policies and only reachable by interpretation in the fifth. The gap between what people ask and what policies address is the finding. Policies converge tightly on third-party harms: recognition databases, sensitive-attribute inference, misuse of someone else's likeness. Self-directed appearance judgment, the case with a person's full consent and the case people actually search for, sits outside the language of every document except Anthropic's.
Silence is not permission. A vendor can refuse a request no policy names, and behavior varies between runs even when policy holds still. That distinction, between the published rule and the observed answer, is why this page audits documents and our ChatGPT guide examines behavior. If checking your appearance has become a weight rather than a curiosity, the support resources on our am-i-ugly and am-i-pretty pages are there for exactly that.
Method
All five policy documents were fetched from the vendors' own sites on September 2, 2026: OpenAI's usage policies, Google's generative AI prohibited use policy, Anthropic's usage policy, the SpaceXAI acceptable use policy, and Meta's AI terms. Every clause that touches faces, appearance, biometrics or likeness was coded into the dataset with its verbatim text, the policy's own effective date where shown, and the fetch date. Where a document contains no relevant clause, the dataset records the word silent. Search demand figures come from this site's keyword corpus, collected in August 2026. Quoted punctuation is normalized to this site's typographic rules; wording is untouched. The full coding table is downloadable at /data/will-ai-rate-your-face.csv, 14 rows across 5 vendors.
Limits
- Policies change. Every clause carries its policy date and fetch date, and this page describes the documents as of September 2, 2026, not forever.
- This is a reading of published policy, not a behavioral test. What an assistant actually answers was not measured here.
- Meta publishes separate regional terms for the UK, EU and Brazil; only the US document was coded.
- Silence in a policy is a fact about the document, never evidence that a use is welcome or safe.