AI is an assistive tool, not an authority. A model output is not evidence, not a credential and not clinical review.
Where AI may be used
- organizing research questions and possible content structures;
- drafting or editing language that is later checked against real sources;
- software development, QA and repetitive technical work;
- search-intent clustering and internal-link planning;
- specific assessment features that explicitly disclose AI-assisted evaluation or generation.
AI-assisted content
AI assistance may be part of drafting, restructuring or checking content. The named author or editorial owner remains accountable for what is published. AI is not listed as a human author and must not invent first-person clinical experience, quotations, credentials, patient stories or references.
Search-engine optimization is not a valid reason to publish AI-generated commodity content at scale. New pages still have to pass the same intent, evidence, usefulness and cannibalization checks as manually drafted content.
Source verification
AI output cannot be cited as factual support. Material claims must be checked against the underlying source, and references must be opened and verified before publication. If a source cannot be verified, the claim is removed, qualified or held.
Clinical limits
AI does not perform independent clinical review and does not turn educational content into medical advice. Clinical pages remain subject to the qualified clinical review gate. AI is not permitted to fabricate diagnostic certainty, emergency instructions or treatment recommendations that are not supported by authoritative guidance.
AI in assessment features
Some development or paid cognitive products may use AI-assisted evaluation of open-text responses. When AI materially affects a user-facing result, the product should explain that role, the limits of the interpretation and the fact that an automated evaluation is not equivalent to a supervised professional assessment.
Production AI scoring, Worker calls and entitlement flows remain separate product-release concerns and must pass their own technical, privacy and reliability gates before the service is offered.
Privacy and sensitive information
AI features should receive only the data necessary for the specific function. Raw clinical answers or inferred conditions are not sent to providers, advertising systems or unrelated partners by default. The current Privacy Policy governs user data handling.
Disclosure and accountability
We disclose AI use when a reasonable reader would want to know how content or a result was produced, especially when automation materially shapes an assessment result. Google Search Central also recommends clarity about the “how” of content production when automation is substantially involved. See Google's people-first content guidance.