To check which competitors ChatGPT and Perplexity name most often, put the same list of buyer questions to each engine and count how many answers name each brand. Do it per engine, since they differ. Keep the cited URLs while you're there, because those explain the count. Ten to twenty questions, repeated a few times, will give you an order you can start to trust.
I ran a small version of this on my own category on 2 November, so you can see what the output looks like and where it misleads.
My 11 answers
These came through our own check tool, which sends one fresh API request per question, with no account history. I looked for CueScout and three competitors.
| Engine | Answers | Profound | Peec AI | Otterly | CueScout |
|---|---|---|---|---|---|
| ChatGPT | 6 | 2 | 2 | 1 | 0 |
| Perplexity | 5 | 3 | 3 | 1 | 0 |
| Total | 11 | 5 | 5 | 2 | 0 |
So my own product was named in none of them. The questions weren't a designed sample: most were adapted from long queries in my Search Console, including one in French, and I asked one of them twice. With 11 answers, a single answer shifts a share by nine points, so this shows direction and nothing finer. A real benchmark needs more questions and more runs.
Still, three things showed up even at this size.
ChatGPT's numbers depended heavily on wording. It named no tracker at all for "tool to see if ChatGPT mentions your brand", and returned homepages for general brand monitoring products instead, as I describe in how to check if ChatGPT mentions your brand. When I asked for the "best AI search visibility tools for a B2B SaaS marketing team", it named Profound and Peec AI. If I'd tested only one of those wordings, I'd have reached the wrong conclusion about who ChatGPT favours.
My list had a blind spot. I only checked three names, so the tool could only report on those three. HubSpot wasn't on my list, and hubspot.com turned up in Perplexity's sources for 3 of the 5 questions. Its AEO product pages were among them. A competitor benchmark that only counts the names you supplied will miss whoever you forgot.
And the competitor set shifted with the question. For "a GEO content strategy platform that turns visibility gaps into a content calendar", Perplexity named none of my three, and Gauge had 3 of the 15 source URLs. Someone asking that question is comparing a different set of tools from the one I had in mind, so a single competitor list won't cover every question.
How to run it yourself
- Write 10 to 20 buyer questions without your brand name. Include the "best tool" kind and the "how do I" kind.
- Ask each on ChatGPT and Perplexity in a new session, with search on.
- Write down every brand named, not just the ones you were expecting.
- Save the cited URLs next to each answer.
- Repeat at least twice more on different days, then count per engine.
Share of voice is that count divided by the number of answers. If Profound appears in 3 of 5 Perplexity answers, that's 60%, and in a sample this small it means very little until it holds across runs. Being named isn't the same as being recommended, so if you have time, note who the answer actually tells the reader to pick.
What to do with the count
Open the pages cited in answers that name a competitor and not you. There are usually a few kinds:
- A third-party list that includes them. Find out who runs it and whether you can be added, which sometimes costs money.
- The competitor's own comparison or feature page. You can publish a fair equivalent.
- A forum thread or review. Those are harder, but they tell you what buyers say about the category.
If you want this done automatically
Trackers run the list for you and rank you against the competitors you name. CueScout's competitor analysis and share of voice views do that and keep the pages behind each answer. For a free look at one category, the AI Category Teardown shows who the answers name and who they recommend, and needs no signup. Other tools with competitor tracking are in our comparison.
Frequently asked questions
How do I check which competitors ChatGPT and Perplexity name most often for questions in our category?
Write 10 to 20 unbranded buyer questions, ask each on both engines with search on, and record every brand named. Count the answers naming each brand, per engine. Repeat over several runs before trusting the order, and keep the cited URLs so you can see which pages sit behind each name.
How can I find competitors that are outperforming me in ChatGPT?
Look for brands named in answers where you aren't. Sort them by how many of your buyer questions they appear in. Include names you didn't expect, since your AI competitors may differ from the companies your sales team runs into.
How do I track my share of voice in AI search versus competitors?
Divide the number of answers naming a brand by the number of answers you collected, per engine and over a fixed question list. Keep the list stable so that a change in the figure means a change in the answers. With a small list, one answer moves the figure a lot, so wait for it to hold across runs.
How many competitors can I track in AI search?
It varies by tool and plan, so check the limit. By hand, the limit is only your patience, because you log every name in the answer. Five to ten is enough for most categories, as long as you also watch for names that aren't on your list.