Out nowThe Citation EconomyBuy on Amazon →

AI Visibility · Methodology

How AI Visibility is calculated

AI Visibility measures one thing: when buyers ask AI engines about a software category, how often is your brand named? Everything you'd need to check or challenge a score is on this page.

40,335

buyer questions

164

categories scored

3

AI engines

9,104

brands scored

01

How the score is built

Buyers use AI in two different moments: when they don't know what to use yet, and when they're comparing a shortlist. We measure each one separately, then combine them.

Discovery rate · 60%

The share of a category's discovery questions (“what should I use for…”) where at least one engine names the brand.

Evaluation rate · 40%

The share of the category's evaluation questions (“X vs Y”, “best for teams like mine”) where it's named.

AI Visibility = 60% × discovery rate + 40% × evaluation rate

Discovery carries more weight because that's where the shortlist forms; a brand that isn't named there rarely makes it into the comparison. Each rate is a plain share of questions, not curved against competitors, so a score says what a brand has actually won, and the leader can still improve.

Worked example: the top brands in CRM, out of 193 scored there.

BrandDiscoveryEvaluationAI Visibility
Salesforce80% (118/148)57% (33/58)71
Hubspot79% (117/148)57% (33/58)70
Zoho CRM60% (89/148)55% (32/58)58
Pipedrive57% (85/148)55% (32/58)57
Microsoft Dynamics 36549% (72/148)55% (32/58)51
  • Scored per category. A brand is scored in its primary category, or its best-scoring category if the primary has too few questions. Brand pages let you switch between categories.
  • Minimum sample. A category needs at least 10 discovery and 5 evaluation questions before it's scored.
  • Never compared across categories. A CRM is only ranked against other CRMs.
  • Tiers. Frequently Cited 40+ · Emerging 15–39 · Rarely Cited 1–14. Most brands are named in only a small share of answers, so most scores are low. That's the finding, not a flaw.
02

What we ask

Each category has its own fixed set of buyer questions: 29,077 discovery and 11,258 evaluation questions across all categories; 164 categories have brands scored. They're written the way buyers actually ask, vary by team size, industry and use case, and never mention a vendor in a discovery question. Because the set is fixed, when a brand moves, the answers moved, not the questions.

Discovery · 148 questions in CRM

  • “CRM for marketing managers to enhance lead management”
  • “tools to automate data entry for customer records”
  • “CRM tools that assist in managing event registrations for marketers”
  • “how to analyze customer data for better decision-making”
  • “what's effective for managing client relationships in real estate”
  • “software to analyze customer demographics for targeted marketing”

Evaluation · 58 questions in CRM

  • “what are the common pitfalls when implementing a CRM system”
  • “how to measure ROI from CRM investments”
  • “what are the integration requirements for CRM systems”
  • “what data security features are essential in CRM tools”
  • “how to measure the ROI of CRM implementations”
  • “what role does automation play in CRM software selection”
03

Which answers count

Every question is put to each of these engines through their APIs:

  • ClaudeClaude Sonnet 4.6, answering from the model
  • ChatGPTGPT-5.2 with live web search
  • GeminiGemini 2.5 Flash, answering from the model

All answers count, whatever sources they draw on. For each question we use each engine's most recent answer, so a question counts once per engine, not once per time it was asked. A brand counts as named for a question when at least one engine's latest answer names it. 106,400 answers are on record.

Engines are pooled with equal weight. We don't weight by any engine's market share, because shares shift month to month and we'd rather the score stayed comparable.

04

How often it's refreshed

Questions are re-asked on a rolling basis, every weekday, with questions that have never been answered going first. Scores are recomputed daily from whatever the latest answers are, and a weekly snapshot powers the movement arrows on the Citation Index.

A rank only moves when answers move: the questions and the formula stay fixed between methodology changes, and every change is listed below.

05

Matching names to brands

AI answers mention brand names, not database records, so every name has to be matched to a brand before it counts. Three rules keep that honest:

  • Validated per category. A brand only appears in a category's rankings after a check that it actually belongs there.
  • Common words need a domain. Brands named after ordinary words (“Base”, “Range”) only count when the answer names their domain, not the word.
  • Same-name products are checked. When two products share a name (Mend the mental-health app vs Mend.io), we verify which one the answers mean against third-party category data, and we correct the brand's categories when they're wrong. In November 2026 we corrected 165 brands this way.

Where this can still go wrong: a name that matches a different product and isn't caught yet. If you think a score is counting someone else's mentions, claim your brand and tell us.

06

What's not in the score

  • Position in the answer. Named first and named fifth count the same today. We plan to add it.
  • Links and citations. We count whether you're named, not whether the answer links to your site.
  • Sentiment. A mention counts whether it's positive or negative.
  • Review ratings and your website. Your brand's fundamentals (how clearly AI understands what you do, third-party coverage, site structure) are scored separately as AEO Readiness and never feed AI Visibility.
07

Changes

12 Nov 2026

AI Visibility is now built from the two rates above. The previous score mixed question coverage, engine count and an answer-position term that didn't separate brands, so most leading brands scored 75–79. Most scores moved down 4–8 points with this change. That reflects the new formula, not a drop in visibility.

See it applied: the Citation Index ranks every scored brand, and every brand page shows its discovery and evaluation counts.