Two measurement systems power Citeon: daily visibility scans that ask AI engines your customers' real questions, and first-party traffic tracking that counts the visitors and leads those engines send you. This page explains both, including what we cannot see — because a measurement tool that hides its limits isn't one.
Every day we put your tracked prompts — the questions buyers actually type — to each AI engine on your plan (ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews on Optimize). The answers are real product answers, with web search where the engine uses it. A judge model then extracts, per answer: whether you are mentioned, whether you are recommended, your position among named brands, every competitor named, and every source the answer cites.
| Score | 60% × mention rate + 40% × recommendation rate, a rolling 28-day average across all measured answers — one bad day never spikes it, one good day never fakes it. |
| Competitors | Found in the answers themselves: a brand named in two or more answers becomes a candidate; it only joins your ranked list after we crawl its own site and confirm it is a positioning peer — marketplaces and retailers stay visible but unranked. |
| Citations | The domains and pages each answer cites, counted per answer. A gap = a source AI leans on in answers that name competitors but not you. |
| Grounding | Fix suggestions are generated from a live crawl of your actual site plus verified business facts (including a live Google Business Profile check) — never from assumptions. |
What we refuse to do: a value we didn't measure is shown as not measured — never as zero. A skipped engine-day is a gap in the line, not a dip. Partial scans are labeled as partial. This rule is enforced in code, not in copy.
Citeon Pulse is a tiny first-party script: no cookies, no localStorage, no persistent identifier. Visitors are counted with a hash that rotates every day and is never stored raw. It works on any site, follows single-page-app navigation automatically, and excludes headless browsers and known bots.
| Any website | Paste one line into your site's <head> — <script defer src="https://www.citeon.ai/p.js" data-site="pk_…"></script> — with the key from your dashboard. Works on WordPress, Webflow, Wix, Shopify themes, custom sites; about two minutes. |
| Astro & Next.js | The citeon-astro / citeon-next packages inject it automatically when you install Citeon Connect — nothing extra to add. |
| Managed | On managed plans we install it for you as part of onboarding. |
| Signal | How we read it |
|---|---|
| Referrer | Visits carrying a referrer from an AI product are attributed to that engine: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com, grok.com, chat.deepseek.com, chat.mistral.ai and the long tail (duck.ai, you.com, poe.com…). bing.com counts as Copilot only on its chat paths. |
| UTM fallback | Most AI clicks strip the referrer. We also read the link parameters engines append — utm_source=chatgpt.com and equivalents — so referrer-less clicks still attribute. |
| Sessions | Same definition as GA4 and Plausible: a 30-minute inactivity window; the session's source is its first pageview. Internal navigation never starts a session. |
| Leads | Captured automatically, no configuration: phone-call taps, email clicks, WhatsApp, bookings on 11 known platforms (TheFork, OpenTable, Calendly, Treatwell, Booksy, Zenchef…), and form submits (search, newsletter and login forms excluded). Each lead is attributed to the visit — and therefore the engine — that produced it. |
| Engagement | Active seconds (visible tab, recent input), scroll depth, and interaction — measured per view, aggregated per engine. |
The honest limit: roughly 97% of AI-referred traffic carries no referrer at all. Links opened from mobile AI apps land in Direct; clicks from Google's AI Overviews are classified as organic search by every tool in the industry, including this one. Your AI numbers are therefore a floor, not a ceiling — the real influence is higher. We print this in the dashboard, not just here.
For comparison: tools that read Google Analytics inherit GA4's consent losses and only count leads you configured as GA4 key events; server-side crawler analytics require access to your CDN or server logs and track no leads at all. Citeon's tracking is first-party, cookieless, and complete on day one.
When Citeon says a fix is done, it re-fetches your live pages and verifies it — schema present, files served, meta updated — and stamps the card verified live with the date. Off-site work (reviews, citations) can't be auto-checked from a crawl, so it is never auto-stamped; we say so on the card instead.