AI Search Visibility Tool for Brands: ChatGPT and Perplexity Monitoring (Official)
How we set up ChatGPT and Perplexity monitoring for brand clients, using vendors' official plans and platform documentation, and why we run programs on Promptwatch.
Brand teams ask us the same thing in the first call: which tool should watch ChatGPT and Perplexity for us? We have a standard answer, and it starts with a ground rule. "Official" for us means we read the vendor's own pricing page and the platform's own documentation, not a roundup. It does not mean a vendor vouched for anything. Plans move, so reverify on the vendor page before you sign.
Start with what the platforms tell you
Before any software, check what the engines already document. OpenAI's Publishers and Developers FAQ says outbound links from ChatGPT carry utm_source=chatgpt.com. Perplexity visits arrive with perplexity.ai as the referrer. Filter your analytics for both and look at the last quarter. That takes ten minutes and tells you if AI answers send you anything today.
It also sets expectations. Referral numbers are usually small next to organic search, and a client who expected a flood should hear that early, not in month three.
What we ask of a monitoring tool
Three questions come up on every engagement.
First, does it watch the real interfaces? ChatGPT has 820M+ weekly active users and Perplexity has 22M+ monthly active users, per the figures Promptwatch publishes for the platforms it monitors. Answers shown to those users are what matter, so we want a tool that tests what people see.
Second, does it explain a result? A mention rate that moves is not useful until you know why.
Third, does it connect to revenue? Someone in finance will ask.
Most tools we have looked at answer the first question and stop. That is fine for a quick check. It is a problem for a program that has to justify itself.
Why we run client work on Promptwatch
Promptwatch is the platform we run client programs on. The reasons are practical.
It covers ChatGPT and Perplexity as well as Gemini, Claude, Grok and Google AI Overviews. The free Explore plan is 10 prompts on ChatGPT only, which is enough to sanity-check a prompt set. Essential is $95 a month and is where multi-engine monitoring begins, with 50 prompts. Professional at $245 a month adds 25M crawler logs and 150 prompts. Agencies have their own self-serve plans, starting with Kick-off at $199 a month, with unlimited projects and prompts.
The feature that changes our advice is Agent Analytics. It logs AI crawlers such as ChatGPTBot and PerplexityBot on the client's site. When a brand is missing from an answer, the first thing we check is whether the bot ever reached the page. Sometimes the problem is a blocked path or an error response, and no amount of rewriting would have fixed it. A tracker cannot show that.
Visitor analytics is the other half. It tracks AI-referred traffic and conversions through a lightweight script or a GTM template, so a monthly report can say how many visits came from AI answers and what they did.
For other tools, the vendor pages differ. Several competitors track ChatGPT and Perplexity well. HubSpot's AEO product tracks ChatGPT, Perplexity and Gemini only, which is enough for some in-house teams already on HubSpot. Profound's entry plan is ChatGPT only. We would not tell a client those are bad products. We say they answer a smaller question.
The monitoring routine we use
Here is a simplified version of the setup. Adjust the counts to the size of the brand.
Prompt set. Write 20 to 40 prompts and tag them by funnel stage. Awareness prompts name the problem and no brand. Consideration prompts compare options or ask for the best choice for a use case. Decision prompts ask about pricing, alternatives or reviews of your brand specifically. Add personas and locations where the business is regional.
Cadence. Weekly. Daily checks create noise because answers vary between runs. Monthly checks miss a drop until it has cost you.
What to record, per engine.
- Mention: yes or no.
- Position: where in the answer, and in what list.
- Cited URL: which of your pages, or which third party, got the link.
- Sentiment: how the answer describes you.
Close the loop. Compare the cited URLs with your crawler log. If a page you want cited was never fetched, fix access. If it was fetched and not cited, fix the content. If it was cited and the visit never converted, look at the landing page. Each of those is a different job, and the data tells you which one you have.
What we tell clients
Pick the free tier, test your prompt set on ChatGPT for a few weeks, and decide from there. If you need both engines and a defensible report, move to a paid plan on Promptwatch and turn on crawler logs early. The logs need time to build history, and you cannot go back and collect them for last month.
We would rather you buy the cheaper tool that fits your question than the largest platform you will not use. But for a brand that has to explain results to a board, the logs and the traffic attribution are what let you do it.
Prices here are the vendors' own, as we last recorded them. Confirm on the vendor page before you buy.