How-to · AI share of voice
To measure share of voice across ChatGPT and Perplexity, run the same set of customer questions on both engines on a fixed schedule, count how many answers name your brand and how many name each competitor you track, and divide your mentions by the total mentions of all tracked brands. Calculate it per engine first, because ChatGPT and Perplexity answer the same question differently and Perplexity cites its sources, then per prompt, then over time. helloAISearch does this automatically across ChatGPT, Claude, Gemini, and Perplexity and charts share of voice per engine and per prompt after every scan.
Free 3-day trial · No credit card · Cancel anytime
A share-of-voice number only means something if the tracking behind it meets these criteria. Whether you count by hand or use a tool, check them first.
Not web-listening for your name on other sites. The tool should send your customers' questions to ChatGPT, Claude, Gemini, and Perplexity itself and record what came back.
You need the sentence where you were named, or the sentence where a competitor was named instead. A visibility index without the text underneath is not actionable.
How often each engine names you versus each competitor, per prompt, over time. This is the number that tells you whether anything you changed worked.
The gap between "you were not named" and "here is the page to write" is where most tools stop. Look for a recommendation tied to a specific prompt.
Entry pricing under $60 a month, no seat minimums, and a feed you can get through with coffee rather than a dashboard you open quarterly.
Your prompts go to the engines on a schedule.
A competitor was named. You were not.
One specific page to write today.
Share of voice moves. Or it doesn't.
We run helloAISearch on ourselves. Between July 17, 2026 and August 31, 2026 it recorded 350 answers from ChatGPT, Claude, and Gemini across 7 category prompts; this prompt joined the tracked set on October 13, 2026, and helloAISearch has not yet been named in its answers. The “named” column is how many of those answers named each tool. Competitor descriptions are how the engines characterised them; we have not added claims of our own.
helloaisearch.com
Otterly AI
otterly.ai
Peec AI
peec.ai
Profound
tryprofound.com
Rankscale
rankscale.ai
Zero is our own number and we are publishing it. This page is step one of fixing it. If you want to see yours, type your website and the trial runs the same check on your brand.
See what the engines say about your brand.
Ten of your prompts on ChatGPT and Gemini, twice, with the actual answers and your brand highlighted.
Write 10 to 20 questions your buyers ask, in their words: "best [category] for [audience]", "alternatives to [competitor]", "how do I [job your product does]". Keep the set stable; changing prompts mid-stream breaks the trend.
Name the 3 to 7 brands you actually lose deals to. Share of voice is always relative to a list, so the list has to stay the same between runs.
Answers vary between runs, so one pass is a sample, not a measurement. Weekly is the minimum; daily for the prompts closest to a purchase. Run ChatGPT and Perplexity separately, never blended.
For each answer, record every tracked brand that was named. Your share of voice is your mentions divided by the total mentions of all tracked brands, for that engine over that period. Positions and sentiment are useful context, but the count is the number that trends.
Perplexity shows the sources behind each answer. The domains it cites when a competitor is named are the pages you need to be on, or to out-write. ChatGPT cites less often, so its share of voice rests more on the answer text.
A 20% share on ChatGPT and 0% on Perplexity is two different problems. Sort prompts by the gap between you and the leading competitor, write one page for the biggest gap, and re-run it after the next scan.
4.1% → 7.2% in 28 days
Write a comparison page for “alternatives to Competitor A”
This prompt came up 4× this week and Competitor A took every mention. A direct comparison page closes a 9-point Share of Voice gap. About an hour.
It depends on how many brands you track and how crowded the category is. With five tracked brands, an even split is 20% each; consistently above that on your core prompts means the engines treat you as a default answer. The trend matters more than the level: a rising share after you publish a page is the sign it worked.
They retrieve differently. Perplexity searches the web for every answer and cites what it found, so it rewards pages that rank and match the question. ChatGPT leans more on what the model already knows and browses less often. The same brand can lead on one and be absent on the other.
At least once a week per engine, and judge the trend over three or more weeks rather than any single run. helloAISearch scans every prompt weekly on Lite and Studio, adds ten daily prompts on Studio, and scans everything daily on Pro and Agency.
Yes, for a handful of prompts: one row per answer, one column per tracked brand, a 1 when it was named. It becomes unmanageable around 10 prompts across 4 engines on a weekly schedule, which is 160 answers a month. helloAISearch's free 3-day trial runs 10 prompts on ChatGPT and Gemini twice and computes share of voice for you.
Free 3-day trial · No credit card · Cancel anytime
10 of your prompts on ChatGPT and Gemini, twice, with the actual answers and your brand highlighted. Plans from $29 a month after.
hello, AI search.