How can I check whether AI assistants like ChatGPT recommend my company to potential buyers?
To check whether AI assistants like ChatGPT recommend your company, write down the questions your buyers actually ask, put each one to ChatGPT, Claude, Perplexity and Gemini, and record three things: whether your company was named, whether it was cited, and who was named instead. Pivot AI Global runs this check as the Discoverability Baseline, across up to fifty buyer questions.
Start with the questions your buyers actually ask
The check is only as good as the questions you put in. Write them the way a buyer would type them, before they know your name: who handles this at our size, what it costs, who is good in our vertical, how long it takes to stand up, who to call first.
Leave your company name out of these. A buyer who already knows your name is past the point where the shortlist gets made.
Ask every assistant your buyers use
Buyers do not all use the same tool. Put each question to ChatGPT, Claude, Perplexity and Gemini, and keep a separate result for each one. An average across the four hides the one that is failing you.
What to record for each answer
For every question and every assistant, write down the same five things. Keep them in one place so you can compare them later.
| Record | What it tells you |
|---|---|
| Named | Whether your company appeared in the answer at all |
| Cited | Whether the assistant pointed to a page about your company |
| Named instead | Which companies the assistant put forward in your place |
| What it said about you | Who it thinks you are best for, what it got right, what it got wrong, and what it left out |
| Sources | The pages the assistant read before answering |
Add one question that names you
Ask each assistant one question with your company name in it, such as “What is [your company] known for?” This works as a control. It shows whether the assistant knows your company exists at all, which is a more basic problem than losing a comparison.
Lock the questions before you run them
Write the full list first, then do not change it. If the questions move from one check to the next, the results cannot be compared and you cannot tell whether anything you changed made a difference.
Look at what the answers were built from
The sources an assistant reads before it answers are your target list. A mention on a page that already shapes these answers moves more than a new page on your own site that nothing reads.
Assistants build a shortlist from text they can find, read and attribute. Guides that answer the buyer's question directly, client results written as plain sentences, and third-party roundups and directory listings all count. Case studies with the numbers inside an image or a PDF usually do not.
Run the same check again later
One check is a photograph. To know whether your work moved anything, run the identical questions again on the same assistants. Answer engines take roughly thirty to forty-five days to recrawl a changed site and start citing it, so leave time between runs.
Where a check by hand falls short
A few questions typed into one assistant on one afternoon is a start. It will not tell you how you compare with the companies you are actually measured against, which sources are doing the work, or which change to make first. Those need the full question list, run the same way each time, against named competitors.
How Pivot AI Global runs the check
Pivot AI Global runs this check as the Discoverability Baseline. The questions are written cold from a read of your site and locked before anything runs. Each one goes to ChatGPT, Claude, Perplexity and Gemini, and the same panel runs against three competitors you approve.
| Baseline | What you get |
|---|---|
| Buyer questions | Up to 50, written for your offer and locked before the run |
| Assistants | 4: ChatGPT, Claude, Perplexity and Gemini, each scored on its own |
| Competitors | 3 that you approve, run on the same questions |
| Sources | The 25 the assistants lean on most, and where your name is missing |
| Actions | 12, ranked by what moves the result most for the least work |
| Plan | A sequence for the first ninety days |
| Price | $3,000. Pivot AI Global runs the first Baseline for a company at no charge |
What to do with what you find
The findings matter less than the order you act on them. Start with the change most likely to get your company named, such as publishing the guide your buyers keep asking for or moving client results out of a deck and onto a page. Then run the same questions again and see what the assistants noticed.
To start a Baseline you give two things: your website and a work email. Pivot AI Global comes back with the segment you compete in, the buyers who would be searching, and the three competitors you are measured against, and you approve that scope before anything runs.
Want the check run for you?
The first Discoverability Baseline for a company is free. It takes your website and a work email to start.
Last reviewed: