Human review
AI outputs are assistive. Customers should review generated recommendations, content, claims, metadata and internal-link suggestions before publication or implementation.
Grounding and sources
AI workflows may use available project context, crawl evidence, search data, brand guidance and customer-provided instructions. The product is designed to keep relevant source context available for review.
Google data in AI generations
When AI assistance and the relevant Google sharing permission are enabled, selected GSC search-query and page-path summaries with performance metrics and GA4 landing-page and traffic summaries may be sent to OpenAI with relevant project context to produce the requested content opportunities or recommendations. Each Google source is included only if you have enabled it for that project. These summaries remain Google-derived data even when aggregated. Google connection credentials are not sent as AI prompt evidence.
Manage your choices in Settings > Privacy & AI. AI assistance and Google sharing start off and are not accepted automatically when you register or connect Google. AI assistance allows OpenAI processing of your submitted messages, business answers and selected SEO evidence. Search Console and GA4 sharing are separate, optional choices for each project and your own requests. Saved choices apply to future requests until you change them; we do not ask again for every generation. Turning a permission off blocks future transmissions that have not already been sent, including pending work.
Google sign-in and viewing reports from connected Google services do not require AI sharing. Leaving the choice off does not recall earlier transmissions or delete saved results.
We do not opt in to sharing Google data with OpenAI for model improvement. We do not use Google data to train general-purpose AI models.
OpenAI may retain data for security and abuse prevention under its applicable terms. This is not a zero-retention promise. See OpenAI API data controls.
See the Google data privacy disclosures for Google sign-in, read-only connections, data use restrictions, disconnection and deletion details.
Training and model use
Customer content is handled according to applicable provider terms, product configuration and contractual commitments. Provider roles and processing purposes are described through our subprocessor disclosures.
Restricted use
- Do not use AI outputs for deceptive, spam-oriented or unlawful activity.
- Do not generate content that impersonates people or misrepresents expertise.
- Do not publish high-stakes claims without qualified review.
Uncertainty and accountability
Generated output may be inaccurate or incomplete. Human users remain responsible for evaluating recommendations and deciding whether to act on them.
Learn more about the product approach on AI Recommendations.