Brand Share of Voice in AI Search - Category Benchmark

$0.08 per prompt benchmarked: a 10-prompt benchmark costs $0.80, and the free sample run costs no prompt fee.

When a buyer asks an AI assistant "what are the best project management tools?", which brands does it name, in what order, and which websites does it cite? This Actor runs that experiment for a whole product category and hands you the scoreboard: every brand ranked by mention rate, citation share and average position, split by AI engine, plus the websites the engines cite most.

You type a category. It writes the buyer questions, asks Perplexity and Google Gemini (ChatGPT optional), reads the brand names out of every answer, merges name variants ("HubSpot", "HubSpot CRM", "HubSpot Sales Hub" become one brand) and counts. No brand list needed, no API key needed.

Try it in one minute:

1. Press Start. The prefilled input benchmarks "project management software" with 3 prompts, and keeps Asana, Trello and Monday.com in the table even if nobody mentions them.
2. Open the Output tab for the brand ranking, or the key-value store record OUTPUT for the full report with the top-cited sources.
3. Tick Free sample run to see real output for 3 prompts with no prompt fee, then raise the number of prompts to 20-30 for a steadier picture.

More AI visibility tools from highbrow_fame

Tool	Question it answers	Price
AI Brand Monitor	Is my brand cited by ChatGPT, Perplexity and Gemini?	$0.05 per answer; full audit $4.80
AI Overview citation checker	Which sites does Google AI Overview cite for my keywords?	$0.03 per keyword
Page citation checker	Is this URL cited by ChatGPT, Perplexity or Gemini?	$0.12 per URL
Shopify AI Readiness Audit	Is my Shopify catalog ready for AI shopping assistants?	$0.01 per store

What you get

One dataset row per brand (recordType: "brand"), best first:

Field	Meaning
rank, brand	position in the ranking and the normalised brand name
mentionRate	share of all analysed answers that mention the brand (0-1)
citationShare	share of all cited source URLs that sit on the brand's own website (0-1)
avgPosition	average place of the brand in the answers that mention it (1 = named first)
engineSplit	mention rate per engine, for example { "perplexity-agent": 0.8, "gemini": 0.6 }
sampleAnswersCount	how many answers (prompts x engines x samples) the rates rest on
answersMentioning, promptsMentioned	raw counts behind the rates
domains	the brand's own domains found among the cited sources

A last row (recordType: "summary") describes the run: prompts requested, benchmarked and failed, engines, answers analysed, brands found, the top brand and the five most-cited domains. The key-value store record OUTPUT holds everything in one JSON (summary, brands, top-cited sources, status of every prompt), and TOP_SOURCES holds the sources table on its own.

Sample output

From a real 3-prompt run (Perplexity and Gemini, 1 sample each, 6 answers). Your numbers will differ: that is the point of measuring.

{
  "recordType": "brand",
  "category": "project management software",
  "rank": 2,
  "brand": "Asana",
  "mentionRate": 1,
  "citationShare": 0.015,
  "avgPosition": 1.83,
  "engineSplit": { "gemini": 1, "perplexity-agent": 1 },
  "sampleAnswersCount": 6,
  "answersMentioning": 6,
  "promptsMentioned": 3,
  "domains": ["asana.com"],
  "isSeedBrand": true,
  "country": "us",
  "language": "en"
}

{ "domain": "forbes.com", "citations": 4, "share": 0.061, "promptsCiting": 3, "engines": ["perplexity-agent"], "brandOwner": null }

The second record is one entry of the top-cited sources table: independent publishers and review sites are visible next to the brands' own websites, so you can see where to earn a mention, not only who already has one.

Input

• Category (required): a noun phrase such as "CRM software" or "running shoes".
• Brands to always include: optional. Your brand and your rivals; a brand nobody mentions is reported at 0%.
• Market (country) and Answer language: the market goes into the generated questions ("... in Germany"); the language is what the engines answer in. Generated questions are written in English; for another language write your own under Your own prompts.
• Number of prompts: 1-30, default 8. Samples: 1-5 per prompt and engine, default 2.
• AI engines: ChatGPT, Perplexity and Gemini by default (ChatGPT limits below).
• Free sample run: at most 3 prompts, 1 sample, Perplexity only, no prompt fee.
• Brands in the table, Prompts in parallel, and an optional OpenAI API key for uncapped ChatGPT.

Price

$0.08 per prompt benchmarked (pay per event). A prompt counts as benchmarked when at least one engine answered it and its answers were analysed; samples and engines do not change the price. A prompt that no engine answers is not charged, and neither is a prompt that the run's maximum charge had no room for. The default run (8 prompts) costs $0.64; a thorough one (30 prompts) $2.40. The free sample run charges no prompt fee; Apify's small run-start fee applies to every run on the platform.

Honest limits

• It measures what the engines say today, from a small sample. AI answers change between requests and between weeks. With 8 prompts and 2 samples you get a reliable first picture of who dominates, not a precise percentage. Use more prompts and samples, or run it monthly, to see movement.
• Brand names are read by a language model. One small model call per prompt lists the brands in each answer; it is accurate on mainstream categories and can miss or mis-split obscure names. If that call is unavailable the Actor falls back to simple rules (bold names and list items) and says so in the summary row (extraction).
• Merging is by name. Variants that start with the same brand name are merged ("Salesforce Sales Cloud" into "Salesforce"). Two different products that share a first word and never appear under the bare name stay separate.
• Citation share counts a brand's own website only. Sources are matched to brands by domain name (hubspot.com belongs to HubSpot). Third-party pages that talk about a brand do not count toward its share; they show up in the top-cited sources table instead.
• Generated questions are English. They are fixed templates (the same category always gives the same questions, so runs are comparable). Write your own prompts for other languages.
• Engines. ChatGPT, Perplexity and Gemini (with Google Search) run on our keys. ChatGPT with web search costs about 3x more to run, so on our key it answers the first 15 prompts of a run with 1 sample each (3 prompts on a free Apify plan); add your own OpenAI key and it runs on every prompt and sample on your own OpenAI bill. Gemini's source links are reported by host name only.
• Not affiliated with or endorsed by Perplexity, Google or OpenAI.

FAQ

Can I benchmark a very specific niche? Yes: "ergonomic office chairs for tall people" works as a category. The more specific it is, the fewer brands you will find and the more each prompt repeats.

Can I run it on a schedule? Yes. Use an Apify schedule and compare the brand rows from run to run, or read the dataset through the API.

Why is my brand missing? Add it under "Brands to always include" and it stays in the table at its true rate, 0% included.

If this Actor was useful, an honest review on its Store page helps other people find and judge it.