AI Disclosure

Nothing is sent to an AI model unless you ask for it by using one of these features. There is no background AI processing of your projects, messages or profile.

• Creator wizard suggestions — drafting or tightening a question, hypothesis, variables, materials, tools or steps for the project you are editing.
• Project structure extraction — turning a description you paste in into a draft project outline.
• Improving or restructuring text you wrote, when you click an improve/rewrite control.
• Project check and readiness summaries — a review of your own draft that flags gaps, including possible missing safety notes.
• Experiment Assistant — a question-and-answer helper for members on the Scientist plan.
• Ask AI — inline answers about the project you are working on.
• Administrator-only tools used by the platform team: scoring publicly available pages during outreach research, and polishing outreach email drafts. These do not read member projects.

The platform is built so the AI provider can be changed in configuration. At the time this page was last updated the application is configured to call the Google Gemini API, and the application also supports the OpenAI API as an alternative configuration. Historically some AI features were routed through Lovable's AI gateway; that path is not used by the AI features today.

One request goes to one provider — the one currently configured for that feature. We cannot promise a provider will never change, so treat this section as a description of the current configuration rather than a permanent commitment. Separately, our payment and email integrations are routed through Lovable's connector infrastructure, which is described in the Privacy Policy.

Only the content needed to answer your request: the relevant parts of the project draft or text you are working on, and the question or instruction you gave. Your password is never sent. We do not send your payment details, payout details or support ticket history to an AI provider.

We keep a record of each AI call for cost and abuse control — the feature used, the provider and model, your account id, whether it succeeded, token counts, estimated cost and how long it took. We do not store the prompt text or the model's answer in that record. Anything you then save into your project is stored as normal project content.

These models predict likely text. They can be confidently wrong, invent sources, miss a hazard, give a measurement or calculation that does not hold, or suggest something that is unsuitable for your age group, classroom, lab, budget or country.

• Read and edit AI output before you publish it. Draft quality is not evidence of accuracy.
• Independently verify procedures, safety requirements, quantities, measurements, calculations, materials and results.
• AI output is not scientific, educational, medical, engineering, legal, safety or other professional advice, and does not replace a qualified person.
• We do not guarantee that an AI-assisted experiment is safe, scientifically valid, original, reproducible or suitable for any particular setting.
• An AI readiness check that raises no warnings does not mean a project is safe or complete.

Do not enter confidential, proprietary, personally sensitive or otherwise restricted information into AI features unless you are comfortable with it being processed by the applicable AI provider under that provider's own terms and privacy policy. That includes unpublished research you are not free to disclose, other people's personal details, and anything covered by a school, employer or institutional confidentiality rule.

We do not control and do not guarantee how a third-party AI provider handles data it receives. Review the provider's policies before submitting anything sensitive.

Do not present AI-generated content as human-authored research, original data or your own experimental results when it is not. Science fairs, schools, journals and competitions often have their own rules about AI assistance — those rules are set by them, and following them is your responsibility.

AI features can be switched off, rate-limited or capped by the platform at any time, and Experiment Assistant usage on the Scientist plan is limited to 60 assists per month. AI features may also be unavailable because of provider outages or configuration. The rest of the platform works without them.

• Privacy Policy — what data the platform handles and who processes it.
• Science & Safety Disclaimer — before you build anything from this site.
• Terms of Service — your responsibilities for AI-assisted content you publish.

For anything about this policy, contact support@bestexperimentideas.com or open a request from the support page.

Operating entity: [legal entity name — not yet configured]. Registered address: [business address — not yet configured].