Legal

AI Usage & Transparency

Last updated: July 2026.

Where AI is used

Every analysis — the score, the issues, the recommendations and the optimised rewrite — is generated by a large language model. Nothing about that is hidden or hand-written.

Which models

We route by optimisation mode: Quick mode uses a fast, efficient model; Professional and Expert modes use a more capable model. Requests are routed through OpenRouter to providers such as Google and Anthropic. The current routing is published in our changelog when it changes.

Known limitations

• Scores are the model's judgement, not an objective measurement. The same prompt can score differently across runs.
• Models can state things confidently that are wrong. Treat factual claims in rewrites as needing your verification.
• Very long prompts may produce truncated analyses; the app will tell you when that happens rather than showing a partial result silently.

What your prompts are not used for

We do not use your submitted prompts to train models, build marketing profiles, or populate a public library. Prompts are processed to produce your report and are not retained by us afterwards.