AI SHORTLIST / METHODOLOGY

How we measure who AI recommends

151 B2B software categories, 1749 buyer prompts, 5246 answers from ChatGPT, Perplexity and Gemini. Latest answer collected Oct 29, 2026, 3:03 PM ET.

1. Which questions we ask

For every category we write 11 to 12 questions the way a buyer types them when they are already choosing a tool, not when they are learning about the space. We skip "what is" and "how does it work" questions, because those answers rarely end in a shortlist. Every prompt names a real buying situation: a budget, a team size or segment, a tool the buyer wants to leave, or two tools they are comparing.

2. Why these are the money prompts

We rank each prompt by how close its wording is to a purchase. A buyer asking about price or alternatives to their current tool is days from a decision. A buyer asking for "the best tool" in general is weeks away. The ranking uses the wording of the prompt only, with the same rules for every category, so nothing is hand-picked to make a brand look better or worse.

  1. 1Direct buy decision"should we buy X", "buy X for a 50-person team"
  2. 2Pricing or budget shortlist"cheapest EHR for independent practices"
  3. 3Switching, alternatives to a known tool"Hootsuite alternatives for agencies"
  4. 4Head-to-head comparison"athenahealth vs eClinicalWorks"
  5. 5Comparison shopping"top ambulatory EHR platforms compared"
  6. 6Vendor selection question"which platform should a 10-provider clinic pick"
  7. 7Best tool for a specific segment"best EHR for a 10-provider medical group"
  8. 8Category best-of"best social media management tool"
  9. 9Problem to tool, criteria-led"how to automate prior authorization"

3. How we ask

Each prompt is sent once to ChatGPT, Perplexity and Gemini in a fresh session with no login, chat history or personalization, in US English. We store the full answer, the sources the engine cited, and the exact time it was collected. Every page shows when its data was collected, in US Eastern time.

4. How we count

  • A mention is one answer that names the brand. A brand named three times in one answer still counts once.
  • Share of voice is a brand's mentions divided by all successful answers in the category.
  • Position is the order in which an answer first names each brand: the first brand named is #1. Buyers read top down, so being named first or in the top 3 matters as much as being named at all. Average position counts only the answers where the brand appears.
  • Product names are merged with their company (for example, a product line and its parent brand), and generic words that look like names, such as "Billing", are removed.
  • Cited domains are the websites the engines linked as sources. They show which pages AI reads before it builds a shortlist.

5. Shinlar AI Visibility Score

Every brand in every category gets one number from 0 to 100. It is computed from the same answers shown on the page, with the same rules for every brand, and nothing is adjusted by hand. It adds up four parts:

  1. 50 ptsShare of answersMentions divided by all successful answers in the category. Named in every answer = 50.
  2. 20 ptsPosition when namedAverage position #1 = 20, falling in a straight line to 0 at #6 or lower. Multiplied by the evidence factor.
  3. 20 ptsEngine coverageHow many of ChatGPT, Perplexity and Gemini named the brand at least once, out of 3. Multiplied by the evidence factor.
  4. 10 ptsOwn site cited10 if any engine linked the brand's own domain as a source in this category, otherwise 0.

Evidence factor = min(1, mentions / (answers / 3)). Position and coverage only count in full once a brand is named in at least a third of the answers, so a single lucky #1 spot cannot lift a brand that AI rarely mentions. The total is rounded to a whole number.

score = 50 × (mentions / answers)
+ 20 × max(0, (6 − avg position) / 5) × evidence
+ 20 × (engines that named the brand / 3) × evidence
+ 10 if the brand's own domain was cited as a source

  • Strong visibility: 70 to 100.
  • Partial visibility: 40 to 69.
  • Low visibility: 0 to 39.

A brand that no answer names can score at most 10, and only if an engine cited its website. The score compares brands inside one category; it is not a traffic or revenue ranking.

6. Limits

AI answers change from day to day and from person to person. This is a dated snapshot of what a new buyer sees, not a traffic ranking and not a guarantee of what any single user will see. We re-run categories to keep the data fresh, and every refresh gets a new timestamp.

Back to AI Shortlist