Are you tracking visibility in ChatGPT, Perplexity, and Google AI Overviews? What does your setup look like?« Back to Previous Page

Honestly I went through the same phase, 30ish prompts in a sheet every Monday. It works for a month and then you hate your life.

What I do now is a mix. I use a paid tracker for the heavy lifting (Peec and Otterly are the ones I've had the best luck with, Profound if the client has budget) and a small sheet on top for the stuff the tools miss. Ahrefs Brand Radar and the Semrush AI toolkit are fine too if you already pay for those, but they felt thinner on prompt-level detail.

A few things I learned the hard way:

30 prompts is too few and also too many to run by hand. Go to 100-150, split by intent (problem-aware, comparison, "best X for Y", branded). Branded prompts will flatter you, so report them separately or they'll hide how weak you are on the non-branded ones.
Run each prompt multiple times. The same prompt on ChatGPT can give you a different set of brands on run 1 vs run 4. One run is basically noise. I run 3-5 and track the percentage of runs where we show up.
API results aren't the same as what a real user sees. If you build a script yourself, the answers can differ from the actual UI, especially when browsing is involved. Fine for trends, don't treat it as ground truth.

On metrics:

Mention rate is the one I look at first. It's simple, and clients get it right away.
Citation share (your cited URLs vs. competitors' for the same prompt set) is more useful for deciding what to do, because it tells you which pages are actually getting pulled in and which competitor pages are beating you.
I watch the trend, never the absolute number. "We went from 12% to 21% over 8 weeks on comparison prompts" is believable. "We have 21% visibility" means nothing on its own.
Referral traffic I trust the least as a KPI, only because the volume is tiny and a lot of AI traffic shows up as direct. Still worth setting up in GA4 though. Make a custom channel group with a regex for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai. The conversion rate on those visits is often way better than organic, so even low volume gets attention from leadership.

For Google AI Overviews, GSC still doesn't split them out, so you're stuck using Semrush/Ahrefs/SE Ranking style trackers to see whether you're in the overview for your keywords. Not perfect, but it's the only way I know of.

If you want to stay cheap, a Python script hitting the APIs and dumping to a sheet gets you maybe 70% of the way. I'd only pay for a tool once you're past a couple hundred prompts or need competitor benchmarking across several engines.