If we cannot show how a result was produced, we do not call it a measurement.
An AI answer is an observation made under specific conditions. xNoted keeps those conditions attached, so your team can review the evidence, repeat the run and disagree with our interpretation when needed.
The free scan checks readiness, not ranking
The scanner inspects public foundations that can help a machine understand and cite a website: secure delivery, titles and descriptions, primary headings, canonical URLs, declared language, structured data, semantic structure, robots policy, international alternates and useful content depth.
A readiness score says whether those foundations are present. It does not say that a brand currently appears in ChatGPT, Gemini or another answer product.
A paid run keeps the original evidence
Where supported, monitoring uses documented APIs and search-grounding tools. Every observation stores the question, language, market, engine, model, answer, citations, execution time and relevant configuration. Consumer chat interfaces are not treated as stable APIs.
The full answer remains available next to any summary metric. A reviewer should never have to trust a percentage without being able to inspect the records beneath it.
Metrics are summaries of a declared question set
- Mention rate
- The share of sampled answers that meaningfully include the brand.
- Citation share
- The share of relevant citations pointing to the brand’s controlled web properties.
- Competitive share
- Brand mentions compared with the named competitor set in the same runs.
- Factual accuracy
- Whether material claims match facts approved by the customer.
- Volatility
- How much outcomes vary across repeated samples and time periods.
Interpretation and observation stay separate
“The brand was absent from this answer” is an observation. “Publish a comparison page” is an interpretation. Reports distinguish the two and attach recommendations to specific evidence, so a customer can accept, reject or revise the proposed action.
What xNoted will not promise
No provider controls the output of an independent answer engine. We do not guarantee a mention, citation, ranking, revenue result or permanent model behaviour. We improve public evidence, measure repeated samples and make uncertainty visible.
Connector methods are versioned. If a provider changes its model or access method, historical results keep the configuration under which they were recorded.