How do I actually measure whether my company is being cited by ChatGPT or Perplexity, versus just guessing?

Why "I asked ChatGPT once" isn't data

Large language models don't return the same answer twice. Ask ChatGPT "best CRM for a 10-person sales team" today and you might see HubSpot and Pipedrive. Ask again tomorrow, or from a different account, or with a slightly reworded prompt, and the list shifts. Perplexity's answers also vary by session because it blends real-time web retrieval with model generation — a page indexed an hour ago can change the citation set. A single screenshot proves nothing except that your company showed up (or didn't) in one specific instance. Measurement means establishing a baseline you can repeat and compare over time, the same way you'd track keyword rankings in traditional SEO rather than checking Google once and calling it done.

Build a fixed prompt set tied to buyer intent

Start with 20 to 30 prompts that mirror how your actual buyers phrase problems, not your product name. If you sell CPQ software, that's prompts like "how do I quote a project with sub-assemblies" or "best CPQ tool for a small manufacturer," not "what is [YourCompany]." Run the same prompt set against ChatGPT, Perplexity, and Google AI Overviews every two to four weeks, logging three things for each: whether you appear at all, where in the response (first mentioned vs. buried in a list of five), and which underlying source the engine cited to justify including you — a review site, a comparison blog, your own documentation page. That source column matters more than most people realize, because it tells you where to focus content effort next.

Tools that automate the tracking

Doing this by hand across three engines and thirty prompts monthly is tedious enough that most teams stop after one cycle. Profound and Otterly.AI exist specifically to run prompt panels against major AI engines on a schedule and report share-of-voice and citation trends over time, similar to how Semrush or Ahrefs track organic rankings. Neither is free — Otterly's paid plans start in the range of $29 to $189+ per month depending on prompt volume as of their published pricing — but either removes the manual-screenshot problem entirely and gives you a trend line instead of a single data point.

What a real measurement cadence looks like

Pick your prompt set once per quarter and don't change it mid-cycle — swapping prompts every month makes the trend line meaningless. Track three numbers: citation rate (percentage of prompts where you appear at all), average position when you do appear, and the ratio of your own domain being cited versus a third-party review or directory citing you. A rising reliance on third-party citations usually means your own site isn't structured in a way these engines trust yet — thin pages, no clear answer-first content, missing schema markup. That's a content and structure problem, not a tooling problem, and no tracker fixes it for you; it just tells you it exists.

Where SalesARC fits

SalesARC's ArcClaw™ AEO service (Basic at $499/month, Standard at $1,499/month, Enterprise at $3,499/month) builds and restructures a company's content so it's the kind of source AI engines actually cite, and includes ongoing tracking against a defined prompt set so you're seeing trend data, not a one-off screenshot. It doesn't replace a tool like Profound for raw monitoring volume — it's built for founder-led B2B companies trying to get cited in the first place, then prove it's working. See ArcClaw™ AEO.