Best platform to monitor share of voice across ChatGPT, Gemini and Perplexity: our practitioner view
How we monitor share of voice across ChatGPT, Gemini and Perplexity for clients: what to define first, which reports to trust, and why we run it on Promptwatch.
Share of voice is the number clients ask for first and understand least. In classic SEO it had a settled definition. In AI search it does not, and two platforms can report different numbers for the same brand on the same day without either being wrong. Before we pick a platform for a client, we agree on what the number means.
Here is how we do it across ChatGPT, Gemini and Perplexity, and what we look for in the platform underneath.
Define the pieces before you chart them
Four decisions shape the number.
The prompt set. Share of voice is only as meaningful as the questions behind it. We build the set from real sales conversations, then tag each prompt as organic, brand or competitor comparison. Branded prompts always flatter a brand, so we report them separately.
The competitor set. A share of 30 percent against two rivals is not the same as 30 percent against twelve. We fix the list at the start, and use Promptwatch's auto-discovered competitors only to find names we missed.
What counts as a mention. The Brand Book in Promptwatch carries analysis instructions that decide this, plus brand aliases. It sounds fussy until a client has a product line with three names and the numbers are wrong for a month.
The engines. We confirm that each plan actually covers the engines the client cares about. Some entry plans on other platforms leave out one of the three. In Promptwatch, every paid plan includes all active models, and ChatGPT (Search), Gemini and Perplexity are monitored through the real interface. The free Explore plan is ChatGPT only.
One number is not enough
Promptwatch reports Visibility Score, Share of Voice, Sentiment (a per-response score from 1 to 100, with a weekly AI summary) and Position in Answer. We use all four. A brand can have a decent share of voice and a poor position, named last in every list, and a client who sees only the first number will be pleased for the wrong reasons.
We also never report a blended figure across engines on its own. The heatmap compares any two dimensions, so brands against models is a standard view in our reports. It is common to see a client strong on Perplexity and nearly absent on Gemini, and the plan for each is different.
Time and noise
AI answers vary between runs. Monitors in Promptwatch run daily per model, country and language, so we get a distribution rather than a single draw. We use A/B period comparison for monthly reviews and ignore day-to-day movement. A rule we hold clients to: nobody reacts to a change smaller than what we see between two identical runs.
From score to explanation
Share of voice tells you where you stand. It does not tell you why. For the why, we open citation analytics, which show which pages and domains the engines used, and Agent Analytics, which shows whether AI crawlers can reach the client's pages. Crawler logs start on the Professional plan ($245/mo), and the Essential plan ($95/mo) has no logs, so for clients who need diagnosis we start on Professional. Visitor analytics then show whether visibility turns into AI-referred traffic and conversions, which is the part finance asks about.
Reporting
Clients want a deck. Promptwatch can export any dashboard as a PDF, schedule PDF reports, export Excel or CSV with a column picker, and connect to Looker Studio. Custom reports start on Professional and are on every agency plan.
Other platforms we have seen
AthenaHQ tracks many models but meters by credit. Scrunch AI's entry plan covers four engines and leaves out Gemini. Profound is sold as an enterprise agent program. Ahrefs and Semrush have AI modules for teams already in those suites, and HubSpot AEO and LLM Pulse are low-cost starting points. Each is reasonable in its niche. For client programs we use Promptwatch because share of voice sits beside citations, crawler logs and traffic, so one conversation covers what, why and so what.
Limits
Share of voice is a measurement, not a guarantee of revenue. We say that in every first meeting.
First steps
Fix the prompt set and the competitor list. Set the Brand Book. Run two weeks of daily data. Review share by engine, then position, then the citations behind the weakest engine. You can try this on the free Explore plan for ChatGPT, or a 7-day trial of Promptwatch for all three engines.