---
title: "Citorum metrics & methodology"
slug: "metrics"
canonical_path: "/metrics"
meta_title: "Metrics & Methodology — Citorum"
meta_description: "The four core GEO metrics Citorum tracks: GEO Index, share of citation, recommendation rate, and sentiment."
h1: "Metrics"
date: "2026-05-03"
---
<!-- key-facts -->
> **Title:** Metrics & Methodology — Citorum
>
> **Canonical URL:** https://trycitorum.com/metrics
>
> **Description:** The four core GEO metrics Citorum tracks: GEO Index, share of citation, recommendation rate, and sentiment.
>
> **Published:** 2026-05-03
<!-- /key-facts -->
# Metrics

**Direct Answer:** Citorum reports four core metrics — the Citorum GEO Index (a 0–100 composite score), share of citation, recommendation rate, and sentiment — plus an engine-by-engine breakdown across ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews.

## Citorum GEO Index

A single 0–100 score combining share of citation, recommendation rate, and sentiment, weighted by buyer-prompt volume. The number you put on the slide. Comparable across engines, prompt sets, and quarters; designed so movement of two points is real, movement of half a point is noise.

## Share of citation

Of every answer to a tracked prompt, what percentage cites or names your brand? Reported per engine, per day, with a 7-day and 28-day rolling average so the chart is readable on a Monday.

## Recommendation rate

Of answers where the model recommends a specific tool or vendor, what percentage recommend you? This is share of citation's stricter sibling — a citation can be incidental ("brands like X, Y, Z exist"); a recommendation is the model picking sides.

## Sentiment

Positive, neutral, or negative framing in the sentence that names your brand. Sentiment skew is what catches narrative drift before it becomes the consensus answer. If sentiment trends negative across two engines for the same week, something upstream changed and Citorum will tell you which source.

## Engine breakdown

Every metric, sliced by ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews; by geography and by persona where supported. A brand winning on Perplexity often loses on ChatGPT for the same prompt — per-engine measurement is the entire game.

## What we deliberately do not report

- **Rank.** There is no rank in an answer; there is a paragraph, and you are either in it or you are not. Anyone publishing a "ChatGPT rank" is publishing a number that does not exist.
- **Estimated traffic from AI.** The engines do not consistently emit click data, and the surface where you are quoted often does not produce a click at all. Reporting fake numbers is worse than reporting none.
- **Forecasted citation count.** We are not in the prediction business. We measure.

## Methodology details

- **Prompt selection.** You define the prompt set. Citorum proposes more from your category data, you approve. Prompts are versioned so historical comparisons stay honest when the set changes.
- **Engine queries.** Each engine is queried via its public surface or supported API, daily, with cache busting where the engine respects it.
- **Entity resolution.** Brand mentions are normalized to canonical entity names so "Acme", "Acme, Inc.", and "acme.com" are one count, not three.
- **Source attribution.** When the engine cites a URL or domain, Citorum keeps the link; when it does not, Citorum traces the most likely upstream source via overlap with the engine's known retrieval mix.

Sign up at [app.trycitorum.com/sign-up](https://app.trycitorum.com/sign-up) or visit [app.trycitorum.com](https://app.trycitorum.com).

## Frequently asked questions

**What is the Citorum GEO Index?**
A 0–100 composite score per brand, per engine, per day. Weighted sum of share of citation (50%), recommendation rate (30%), and sentiment (20%), normalized against the active competitive set. The number you put on a slide every Monday.

**Does Citorum publish a 'rank'?**
No. There is no rank inside an answer; there is a paragraph and you are either in it or you are not. Anyone publishing a ChatGPT rank is publishing a number that does not exist in the model's architecture.

**How is sentiment measured?**
Sentiment is bucketed into positive, neutral, hedged, or negative for the sentence in which your brand is named, then mapped to a 0–1 score. Bucket boundaries and inter-rater agreement statistics are published on request.

**How many prompts do I need before the metrics are stable?**
At least 25 active prompts per category for a usable signal, and 50+ for trend reading. Below 25, single-prompt swings dominate the number and the dashboard flags the category with a low-confidence badge.

**How does Citorum handle engine outages?**
When an engine fails to return a usable answer for a prompt, that prompt is excluded from the day's denominator for that engine, not counted as a zero. Otherwise an engine outage would look like a brand collapse.


<!-- structured-data -->
<script type="application/ld+json">{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://trycitorum.com/#organization","name":"Citorum","alternateName":"Citorum, of those who are cited","url":"https://trycitorum.com/","mainEntityOfPage":{"@id":"https://trycitorum.com/about#webpage"},"logo":"https://trycitorum.com/brand/logos/citorum-wordmark-ink-on-cream@2x.png","description":"Citorum tracks how often your brand is cited, recommended, and quoted across every major AI answer engine.","slogan":"Be the answer, not the runner-up.","foundingDate":"2025","email":"hello@trycitorum.com","sameAs":["https://app.trycitorum.com/"],"contactPoint":[{"@type":"ContactPoint","contactType":"customer support","email":"hello@trycitorum.com","url":"https://trycitorum.com/about","availableLanguage":["English"]},{"@type":"ContactPoint","contactType":"press","email":"press@trycitorum.com","url":"https://trycitorum.com/about"},{"@type":"ContactPoint","contactType":"privacy","email":"privacy@trycitorum.com","url":"https://trycitorum.com/privacy"}]},{"@type":"Person","@id":"https://trycitorum.com/about#editorial-team","name":"The Citorum editorial team","url":"https://trycitorum.com/about","jobTitle":"Editorial team","worksFor":{"@id":"https://trycitorum.com/#organization"},"knowsAbout":["Generative Engine Optimization","Share of citation","Retrieval-augmented generation","AI answer engines"]},{"@type":"WebSite","@id":"https://trycitorum.com/#website","url":"https://trycitorum.com/","name":"Citorum","publisher":{"@id":"https://trycitorum.com/#organization"},"inLanguage":"en-US"},{"@type":"SoftwareApplication","@id":"https://trycitorum.com/#software","name":"Citorum","applicationCategory":"BusinessApplication","operatingSystem":"Web","url":"https://trycitorum.com/","description":"Generative Engine Optimization (GEO) platform that monitors brand citations across ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews.","publisher":{"@id":"https://trycitorum.com/#organization"},"offers":{"@type":"Offer","url":"https://app.trycitorum.com/sign-up","category":"SaaS subscription"}}]}</script>
<script type="application/ld+json">{"@context":"https://schema.org","@type":"WebPage","@id":"https://trycitorum.com/metrics#webpage","url":"https://trycitorum.com/metrics","name":"Metrics & Methodology — Citorum","description":"The four core GEO metrics Citorum tracks: GEO Index, share of citation, recommendation rate, and sentiment.","isPartOf":{"@id":"https://trycitorum.com/#website"},"inLanguage":"en-US","dateModified":"2026-05-03"}</script>
<script type="application/ld+json">{"@context":"https://schema.org","@type":"BreadcrumbList","@id":"https://trycitorum.com/metrics#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://trycitorum.com/"},{"@type":"ListItem","position":2,"name":"Metrics","item":"https://trycitorum.com/metrics"}]}</script>
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","@id":"https://trycitorum.com/metrics#faq","mainEntity":[{"@type":"Question","name":"What is the Citorum GEO Index?","acceptedAnswer":{"@type":"Answer","text":"A 0–100 composite score per brand, per engine, per day. Weighted sum of share of citation (50%), recommendation rate (30%), and sentiment (20%), normalized against the active competitive set. The number you put on a slide every Monday."}},{"@type":"Question","name":"Does Citorum publish a 'rank'?","acceptedAnswer":{"@type":"Answer","text":"No. There is no rank inside an answer; there is a paragraph and you are either in it or you are not. Anyone publishing a ChatGPT rank is publishing a number that does not exist in the model's architecture."}},{"@type":"Question","name":"How is sentiment measured?","acceptedAnswer":{"@type":"Answer","text":"Sentiment is bucketed into positive, neutral, hedged, or negative for the sentence in which your brand is named, then mapped to a 0–1 score. Bucket boundaries and inter-rater agreement statistics are published on request."}},{"@type":"Question","name":"How many prompts do I need before the metrics are stable?","acceptedAnswer":{"@type":"Answer","text":"At least 25 active prompts per category for a usable signal, and 50+ for trend reading. Below 25, single-prompt swings dominate the number and the dashboard flags the category with a low-confidence badge."}},{"@type":"Question","name":"How does Citorum handle engine outages?","acceptedAnswer":{"@type":"Answer","text":"When an engine fails to return a usable answer for a prompt, that prompt is excluded from the day's denominator for that engine, not counted as a zero. Otherwise an engine outage would look like a brand collapse."}}]}</script>
<!-- /structured-data -->