Content readiness
Our proprietary rules based rubric assesses 13 content factors, including answer clarity, structure, extractability, and trust signals. Scores describe that rubric, not the internal ranking systems of AI providers. They are not calibrated probabilities or independently validated predictions of citations.
Where questions come from
Customer imports, public research suggestions, and content derived suggestions have different evidence. Public research uses provider supplied source references; references may support the research as a whole rather than prove that a specific question was asked. We do not have access to private conversations inside consumer AI platforms or global prompt volumes.
Available model tests
Prompt testing supports ChatGPT, Claude, Gemini, Perplexity, and Grok providers through configured API models. Our Microsoft Copilot pages offer guidance only; Proven Cite does not currently run tests in Copilot. Provider availability and the actual model used can vary.
API responses can differ from consumer apps because browsing, personalization, system instructions, and model versions differ. A test is a dated snapshot, not continuous observation of every user’s experience.
What counts as evidence
Untested questions stay labeled Not tested. Saved responses distinguish a brand mention, a detected domain link, and absence. A link in generated output is not proof that the provider retrieved or relied on that page. Review the full response and source before interpreting it.
A reproducible comparison
- Save the exact question, tested domain, provider, model where available, and date.
- Keep the full response and citation URL, not only a summary score.
- Document the content change and repeat the same question with the same model.
- Report both positive and negative results, sample size, and possible confounders.
- Measure leads and sales separately, and treat a change as a signal rather than proof that the edit caused it.
How to read results claims
Examples on this site are illustrative, not customer outcomes. When you weigh any AI visibility result, from us or another vendor, ask for dated before and after observations, the specific change that was made, and what else could explain the difference.
Sources and questions
See Google’s guidance on AI search features for search eligibility. This supports general SEO practices, not validation of our scoring weights. Questions about how we measure, or something you think we got wrong? Email hello@provencite.com.