Ask your real buyer questions to ChatGPT, Claude and Perplexity, then log every name each one gives back instead of yours; that list is who AI recommends over you. In one Reflexa evidence-log example, HubSpot led all three engines for a single CRM question, while Salesforce, Pipedrive and Zoho rotated in and out behind it (Reflexa AI Visibility Index, October 2026).
It means asking the exact question a buyer would ask, across all three engines, and writing down every name each one gives back that isn't you. That's a different object than a Google ranking or a mention count: ChatGPT, Claude and Perplexity answer "best X for Y" questions with a short, named shortlist, usually three to five brands, and every name on it that isn't you won the moment you wanted. Our comparison of the leading AI visibility tools covers ten vendors that help run this check at scale, but the method itself works with nothing more than the three chat windows.
Bottom line: the unit to track is the full named list per question per engine, not a single "are we mentioned" check.
It looks like a ranked list per engine, attached to one buyer question, not a percentage. Here is a real row from Reflexa's CRM evidence log, recorded October 2026, for the question "Which CRM would you recommend for lead management specifically?":
| Engine | Named, in order | A source it cited |
|---|---|---|
| ChatGPT | 1. HubSpot · 2. Pipedrive · 3. Zoho | blog.hubspot.com |
| Claude | 1. HubSpot · 2. Salesforce · 3. Pipedrive | www.saleshandy.com |
| Perplexity | 1. HubSpot · 2. Pipedrive · 3. Zoho | www.g2.com |
Proof: Reflexa AI Visibility Index — CRM: who AI recommends, October 2026, evidence log.
Read this row as a competitor, not as HubSpot. If you are Salesforce, you only made the list once, in Claude's answer, and Claude is the one engine that leaned toward feature detail ("built-in lead tracking, email marketing, automation") rather than a plain recommendation. If you are Zoho, you placed third in two engines out of three and never in Claude's. That is the actual, actionable difference between "we are invisible to AI" and "we are invisible on two of three engines, for this one question."
Why it matters: once you can see the ranked list per engine, you know exactly which engine to fix for, not just that you are losing.
Usually one of four reasons, and they are all checkable. The competitor's page is easier for a crawler to read than yours; it states a fact plainly instead of pitching it, which is what the engine actually quotes; a third-party source the engine trusts cites them and not you; or your own entity details are ambiguous enough that the engine named someone it was more certain about. Our five reasons ChatGPT recommends a competitor breaks each of these down with the free tool that checks it.
Do this: before assuming "they have a bigger brand," check the cited source behind the competitor's mention; it is usually the real reason.
Because each engine builds its answer from a different mix of training data and live search, and because the models sample rather than read off a fixed ranking. Independent testing of repeated identical prompts found that AI models rarely return the exact same brand list twice, matching lists in under 1 in 100 repeated runs. A competitor who dominates one engine's answers can barely register in another's, as the CRM row above shows for Salesforce and Claude.
Proof: SparkToro — New research: AIs are highly inconsistent when recommending brands or products.
Why it matters: one screenshot of "AI recommended our competitor" is a sample; you need the same question repeated across engines and over time before you react to it.
Most AI visibility tools summarize this into a percentage; only a few keep the actual ranked names you can read. Reflexa attaches the full ranked answer and its cited sources to every finding, per engine, on every plan including the free first check. Otterly.AI and Peec AI show the sources an engine cited for a prompt, a step beyond a bare score, but neither publishes a full ranked-answer log on its own public page. Profound, Semrush AI Toolkit, Scrunch AI and Ahrefs Brand Radar each report a score, a percentage or a share-of-voice figure as the primary output (each vendor's own pricing or documentation page, September 2026). See the full seven-tool table in Profound alternatives.
You can see your own competitor list without buying anything yet:
Ask the Reflexa agent which comparison pages name your competitors and not you: it lists them from the sources the engines retrieve, with the page behind each one, on any paid plan.
Usually one of four reasons: the competitor's page is easier for crawlers to read, it states a fact plainly instead of pitching it, it is cited by a third-party source the engine trusts, or your own entity details are ambiguous enough that the engine named someone clearer instead.
No. Independent testing of repeated identical prompts found AI models rarely return the exact same brand list twice, so a single run is a sample, not a trend (SparkToro, 2026).
A fixed set of your real buyer questions, asked across all three engines and repeated on a schedule, not a one-off prompt. One answer can be noise; a tracked set repeated weekly shows a real shift.
Yes. Run the free check at dashboard.reflexatechnologies.com/audit. It reads your public pages the way ChatGPT, Claude and Perplexity read them and runs 23 structural checks — crawler access, what is readable without JavaScript, what says who you are — and shows what to change, in order. No card, nothing installed.
The free check reads your pages the way ChatGPT, Claude and Perplexity read them — 23 structural checks, evidence attached, no card needed.