One of four shapes: best-for, alternatives-to, cheapest or for-a-team-of-N. Reflexa builds every buyer-question set behind the AI Visibility Index from exactly these four, because each one forces ChatGPT, Claude and Perplexity to apply a different filter before it names anyone.
Best-for, alternatives-to, cheapest and for-a-team-of-N are the four shapes real buyers type before picking software, and each one asks the engine to do something different. A best-for question ("best CRM for a small sales team") asks for a category shortlist from scratch. An alternatives-to question ("alternatives to Salesforce that are easier to set up") anchors on one named company and asks who else belongs next to it. A cheapest question forces a price filter before any name comes out. A for-a-team-of-N question forces a size filter, since the right tool for a 5-person startup and a 50-person company is rarely the same answer.
| Question shape | Example | What it forces the engine to do |
|---|---|---|
| Best-for | "Best project management tool for a remote team" | Build a category shortlist with no brand given |
| Alternatives-to | "Alternatives to Asana" | Name companies comparable to the one already given |
| Cheapest | "Cheapest accounting software for a startup" | Filter by price before naming anyone |
| For-a-team-of-N | "Best CRM for a team of 50" | Filter by company size and likely feature needs |
Proof: these are the same four shapes behind the buyer questions in best AI visibility tools for B2B SaaS companies, where real SaaS buyer questions are described the same way, best-for, alternatives-to, cheapest and for-a-team-of-N.
Bottom line: a brand that only tracks "am I mentioned" is checking one shape of question at most; the other three can be failing silently.
Because the brand named in the question gets a free mention that has nothing to do with whether AI would have picked it on its own. If a buyer asks "alternatives to Asana," Asana appears in that answer by definition, so counting it as a win would inflate Asana's own score with a question that already gave away the answer. The Reflexa AI Visibility Index calls this an anchored question and excludes it from the named brand's own denominator, while still counting it normally for every other company that question surfaces.
Worked example: the Index's September 2026 project-management edition asked 100 fixed questions three times across three engines, 900 answers in total. Two of those questions were anchored on Asana ("alternatives to Asana" phrasing), producing 18 answers that were dropped from Asana's own denominator. Across the remaining 882 eligible answers, Asana was recommended in the top three in 359 of them, 359 ÷ 882 = 40.7%, reported as 41% Recommended (Reflexa AI Visibility Index, September 2026).
Do this: if you're building your own question set, write down which questions name a competitor outright and exclude those answers from that competitor's score, the same rule the Index uses, covered in full in how to measure AI visibility.
Not reliably, because the engine can only repeat the number it could fetch, and a lot of pricing pages don't give it one. ChatGPT, Claude and Perplexity read whatever pricing page or review site they can reach at the time of the question; a page that hides its price behind "contact us," or hasn't been updated since a price change, hands the engine a stale or missing number instead of a fact it can quote.
Why it matters: a cheapest question doesn't reward the cheapest tool, it rewards whichever tool made its price the easiest fact for AI to find and trust; the entity and freshness problems behind this are covered in why AI gets my pricing wrong.
Yes, because a for-a-team-of-N question adds a constraint most category pages never answer. "Best CRM for a team of 50" asks for a different shortlist than "best CRM," since seat limits, support tiers and onboarding effort start to matter at that size in a way a generic category page rarely spells out. A company whose own site only answers the generic version of the question gives AI nothing to quote when a buyer adds a team size to it.
Bottom line: a strong answer to "best for X" says nothing about whether the same company answers "best for a team of 5" or "best for a team of 200," and buyers ask all three.
A different page or section for each shape, not one generic product page trying to answer all four. The four map cleanly onto four kinds of content:
Do this: check which of the four your site currently answers with a real page, and treat the missing ones as your next content gaps, not your marketing copy gaps.
None of the steps below require buying anything:
Bottom line: a brand that checks only one question shape is measuring a quarter of what its buyers actually type.
Sources: Reflexa — Best AI visibility tools for B2B SaaS companies · Reflexa — How do I measure AI visibility? · Reflexa AI Visibility Index, September 2026
Best-for ("best CRM for a small sales team"), alternatives-to ("alternatives to Salesforce"), cheapest ("cheapest project management tool for a startup") and for-a-team-of-N ("best accounting software for a team of 50"). Each one forces ChatGPT, Claude and Perplexity to apply a different filter before naming anyone.
An anchored question names a brand outright, such as "alternatives to Asana," which hands that brand a free mention before the engine does any work. The Reflexa AI Visibility Index drops those answers from the named brand's own denominator so its score isn't inflated by a question that already gave away the answer.
Not reliably. ChatGPT, Claude and Perplexity read whatever pricing page or review site they can fetch at the time, and a page that hides its price behind "contact us" or hasn't been updated gives the engine stale or missing numbers to work with.
10 to 20 real questions split across the four shapes, asked three times each across ChatGPT, Claude and Perplexity, is enough to see a pattern. The Reflexa AI Visibility Index runs the same method at 100 questions and 900 observed answers per category.
Run the free check at dashboard.reflexatechnologies.com/audit, which asks your real buyer questions across ChatGPT, Claude and Perplexity and shows the raw answers, in minutes with no card needed.
The free check reads your pages the way ChatGPT, Claude and Perplexity read them — 23 structural checks, evidence attached, no card needed.