What Emphanic does
Four parts: the questions you are measured on, the measurement itself, the technical state of your site, and the changes that follow from it.
Prompt research
The questions you are measured on
We turn your services into real customer questions and put them again every week, unchanged, so measurements stay comparable.
- Twelve buying-intent questions per industry and city
- Up to 3 terms on Starter, up to 10 on Pro
- The same prompt set every week, so the trend holds
Visibility analytics
Your position across four assistants
A score from 0 to 100 per engine and overall, measured weekly across ChatGPT, Claude, Perplexity and Gemini, with competitors alongside.
- Score per engine, with its denominator
- Trend line across weeks, plus the competitor line
- Every answer readable verbatim
- Ranking: who does get named, and how often
Content audit
Why AI skips your site
Nine technical checks on your site, grouped by the question they answer, with the evidence we observed attached.
- Crawl access: may and can an AI crawler fetch your page
- Readability: is the content there without JavaScript
- Discoverability: canonical and sitemap
- Every outcome with the literal line or header as evidence
GEO optimisation
The change, not the advice
Every finding ships with the code or copy you need, plus the steps to deploy it. After that we check weekly whether it held.
- JSON-LD, llms.txt, robots.txt and rewritten copy
- Ordered by strength of evidence, then by effort
- Status per fix: open, done, regressed
- Weekly re-check with evidence
What we do not do yet
Two things competitors show and we deliberately do not, because we cannot evidence them honestly.
- Search volume per prompt. No reliable source exists for how often a question is put to an AI assistant. Putting an estimate next to a real measurement would make the rest of the report suspect.
- Sentiment of mentions. We count whether you are named, not in what tone. That needs a separate judgement per answer; until we can do it reliably, we do not measure it.