AI Visibility Tracker: ChatGPT, Perplexity, Gemini

Find out whether AI assistants recommend your brand. Give it your brand, a few competitors and the questions your buyers ask. For every question it asks ChatGPT, Perplexity, Gemini and Claude, then tells you who got named, in what order, which websites were cited, and your share of voice on each engine.

ChatGPT and Gemini answers are the ones the apps actually show people, with their citations, not a cheaper API model standing in for them.

What do you get?

One row per answer (prompt x engine):

• Mentioned or not, for your brand and each competitor
• Position: who was named first, second, third
• List rank: your place in the answer's main "top picks" list (#1, #2, ...)
• Cited: whether the answer linked to your website, and which of your pages
• Sources: every URL and domain the answer cited
• Sentiment (experimental): positive, negative, mixed or neutral, from a word list (see below)
• Other brands named (ChatGPT): brands in the answer that you aren't tracking yet
• The full answer text, plus the web searches the AI ran where the engine shows them

One summary row per run, per engine and across all engines:

• Share of voice: of the answers that named any tracked brand, how many named you
• Mention rate: % of answers naming you
• Citation rate: % of answers citing your site
• Average position when you are named
• Top cited domains: the sites the AI leans on for your topic, the place to get coverage
• Answers that failed or were skipped, per engine, so you always know how complete the numbers are

One share-of-voice row per engine x brand, ready for a chart or a sheet.

Why this one?

• Four engines in one run, a fixed price per answer. No API keys, no AI usage billed on top.
• Real app answers for ChatGPT and Gemini, collected by our data provider the way a person sees them, citations included.
• Your country. ChatGPT, Gemini and Perplexity answer as if asked from the country you pick (191 countries).
• Steadier numbers. AI answers change from one ask to the next. Set "Repeats per prompt" to 2-5 and the summary averages over them.
• Compare with last time. Give the run a project name and the summary shows how each number moved since your last run with that name. Schedule it weekly and you have a tracker.
• Your prompts, exactly. Nothing is padded with template questions.
• Failures are free and visible. A failed or empty answer is a free row with the reason, and the summary counts them per engine. A run never says "done" while quietly missing answers.

How do I check if ChatGPT mentions my brand?

1. Type your brand, with its website after a comma: HubSpot, hubspot.com.
2. Add up to 20 competitors the same way, one per line.
3. Add the questions your buyers ask, one per line, up to 100 per run.
4. Pick the engines (all four by default), the country and the language.
5. Run it. Open the Summary tab for share of voice, Answers for every answer.

How do I track AI visibility every week?

Save your input as a task, set a project name (for example Acme weekly), and schedule the task weekly in Apify. Each run's summary includes previousRun and changeSinceLastRunPts, the change in percentage points for share of voice, mention rate and citation rate on every engine. The last summary per project is kept in a store in your own Apify account called ai-visibility-tracker-history; delete it any time.

How much does it cost?

$0.008 per ChatGPT or Gemini answer, $0.02 per Perplexity or Claude answer, everything included. One answer = one prompt on one engine (times your repeats).

Run	Answers	Cost
1 prompt, 4 engines	4	$0.056
10 prompts, 4 engines	40	$0.56
10 prompts, ChatGPT and Gemini only	20	$0.16
25 prompts, 4 engines, 3 repeats	300	$4.20

Not charged: failed answers, empty answers, prompts we can't send, anything past your spending limit or the run's time limit, and the summary and share-of-voice rows. Set a maximum cost per run in the run options and the Actor stops asking before it would go over.

Free Apify accounts get up to 8 answers per run (two prompts on four engines). A paid Apify plan has no such limit. One run takes up to 1,000 answers.

Input example

{
  "brand": "HubSpot, hubspot.com",
  "competitors": ["Zoho CRM, zoho.com, Zoho", "Pipedrive, pipedrive.com", "Salesforce, salesforce.com"],
  "prompts": ["best CRM for small business", "which CRM is easiest for a 5-person sales team"],
  "engines": ["chatgpt", "perplexity", "gemini", "claude"],
  "country": "US",
  "language": "en",
  "samplesPerPrompt": 1,
  "projectName": "HubSpot weekly"
}

Output example

An answer row from a real run (shortened):

{
  "rowType": "answer",
  "prompt": "best CRM for small business",
  "engine": "ChatGPT",
  "status": "ok",
  "brandMentioned": true,
  "brandPosition": 1,
  "brandListRank": 1,
  "brandCited": true,
  "brandCitedUrls": ["https://www.hubspot.com/products/crm"],
  "brandSentiment": "positive",
  "brandsMentioned": ["HubSpot", "Zoho CRM", "Pipedrive", "Salesforce"],
  "otherBrandsNamed": ["Freshsales", "Monday CRM"],
  "citedDomains": ["hubspot.com", "zoho.com", "pipedrive.com", "freshworks.com"],
  "charged": true
}

The summary row from the same run (2 prompts on 4 engines, shortened):

{
  "rowType": "summary",
  "brand": "HubSpot",
  "answers": 8,
  "shareOfVoicePct": {"ChatGPT": 28.6, "Perplexity": 28.6, "Gemini": 33.3, "Claude": 28.6, "All engines": 29.6},
  "mentionRatePct": {"ChatGPT": 100, "Perplexity": 100, "Gemini": 100, "Claude": 100, "All engines": 100},
  "citationRatePct": {"ChatGPT": 100, "Perplexity": 50, "Gemini": 0, "Claude": 0, "All engines": 37.5},
  "avgPosition": {"ChatGPT": 1, "Perplexity": 1.5, "Gemini": 1.5, "Claude": 1, "All engines": 1.25}
}

How to read the numbers

• Share of voice = answers naming the brand / answers naming any tracked brand. It depends on which competitors you list, so keep the list the same between runs you compare.
• Mention rate = answers naming the brand / all answers that came back.
• Position = order of first mention among your brand and your competitors. Brands you didn't list don't count here (see "Other brands named").
• List rank = the brand's entry number in the answer's main list. The main list is the one whose entries name the most of your tracked brands (a "top picks" list, numbered or not); only the entries' own title lines count, not their sub-points. Empty when the brand has no entry of its own in that list, for example when it is only named in the opening sentence or inside another entry.
• What counts as text: what a reader sees. Image captions and link addresses are ignored, so a brand named only in a source URL is not "mentioned"; a link's visible text does count.
• Brand matching: whole words, any case, on the name, the other names you give, and the website. HubSpot matches "HubSpot's" but not "HubSpotter". If your brand is also a common word, add a more specific name ("monday CRM", "Square payments").
• Sentiment is experimental. It comes from a list of positive and negative words in the sentences that name the brand. In our hand check of 40 answers it matched a careful reading about 6 times in 10, mostly by calling praise "neutral" when the praise uses words outside its list. Use it to spot answers worth reading, not as a score.

Engines

Engine	What you get	Country
ChatGPT	The chatgpt.com answer as people see it, with citations and the brands it detected	Yes
Gemini	The gemini.google.com answer as people see it, with citations	Yes
Perplexity	Perplexity Sonar with live web search, via the official API	Yes
Claude	Claude Haiku from its own knowledge, via the official API, no live web search, so no citations	No

Limitations

• AI answers vary between asks. Use repeats for numbers you will act on.
• Claude answers from what the model learned, not from today's web, so it shows no sources and ignores the country.
• Prompts are capped at 500 characters (the AI APIs' limit).
• The ChatGPT and Gemini app answers are collected by our data provider; we never log in to or use those apps ourselves.

FAQ

Where does the data come from?

From DataForSEO's AI Optimization API. ChatGPT and Gemini answers come from its LLM Scraper, which reads the answers the consumer apps show. Perplexity and Claude come through its LLM Responses endpoint, which calls the official Perplexity and Anthropic APIs.

Is it legal to track what AI assistants say about my brand?

The answers come from DataForSEO, a licensed data provider: ChatGPT and Gemini as their apps show them, Perplexity and Claude through their official APIs. The Actor never logs in to any AI service and collects no personal data, only what the engines say about brands. Check your own use against your local rules.

Can I use it with the API, AI agents or MCP?

Yes. Call it through the Apify API with the same JSON input and read the default dataset, or add it as a tool in Apify's MCP server. The summary row is the easiest thing for an agent to read.

Does it collect personal data?

No. It sends your questions to AI engines and reads their answers about brands.

What happens if a run times out or restarts?

Answers already saved are kept and not charged again; the run picks up where it stopped. If the time limit is close, the remaining answers are listed as skipped (free) and the run still finishes with a summary.

More from Meridian Labs

• Backlink Checker: backlinks, referring domains, domain rank and spam score for up to 1,000 domains or URLs per run.
• Competitor Keyword Research: every keyword a website ranks for in Google, with position, ranking page, search volume and estimated traffic.
• Keyword Search Volume & CPC API + Google Trends: monthly search volume, CPC, competition, difficulty and intent for keyword lists.
• Google Jobs Scraper: Salaries, Apply Links & API: every job Google Jobs shows for a search, up to 200, with full descriptions, every apply link and salary normalised to yearly and hourly. $1.50 per 1,000 jobs.
• Google Trends Scraper & API: interest over time, by region and rising related searches for hundreds of keywords, as a maintained pytrends alternative. $5 per 1,000 keyword lookups.
• Facebook Ads Library Scraper: every ad a competitor runs on Facebook and Instagram, with days running and a plain-English summary per advertiser. $0.10 per 1,000 ads.
• YouTube Transcript API & Scraper: full transcripts with timestamps in the video's original language, for videos and Shorts. $3 per 1,000 transcripts; videos without captions are free.
• Greenhouse, Lever & Ashby Jobs Scraper: open jobs from companies on six hiring systems, with salary normalised to min, max, currency and period. $2 per 1,000 jobs.
• Instagram Transcript Scraper: Reels Transcript API: what is said in each Instagram Reel, as text with timestamps and SRT subtitles, in its original language. $3 per 1,000 reels; music-only reels free.

Support

Something looks wrong, or you need a field we don't return? Open an issue on the Issues tab with the input you used. We read every one.

If this saved you time, a rating helps others find it; if something's off, open an issue and we'll fix it fast.