AI brand visibility checker for ChatGPT, Claude, Gemini and Perplexity

Show your client where ChatGPT, Claude, Gemini and Perplexity name a rival instead of them. Type the questions your client's customers ask, and get back the full dated answer, the exact pages it cited, the exact sentences it wrote about the brand, and its rank against up to 5 competitors. Real test on Moz, 10 questions on 4 assistants: Moz was named in 14 of 40 answers, Semrush in 36, and ChatGPT named Moz 2 times out of 10. Built for SEO agencies that report AI visibility to their clients. That full check (40 answers) lists at $2.705. $0.03 to $0.12 per answer depending on the assistant ($0.05 ChatGPT, $0.12 Claude, $0.07 Gemini, $0.03 Perplexity), no subscription, no minimum. Failed or empty answers are not charged (a $0.005 start fee applies once per run), and you can set a maximum cost per run.

• The full answer, not a score. Every row keeps the complete answer, the exact model name and the time it was fetched (UTC).
• The exact pages it cited. Not just the website: the full address of each page, its title, and which of the brand's own pages the assistant sent people to.
• The exact sentences about the brand. Every sentence of the answer that names the brand, copied word for word and counted by code (brandSentences). No tone score: you read them.
• The brand's rank, counted by code. Who is mentioned first: the brand or its competitors? Whole-word matching, case insensitive, with aliases. No AI "grades" the brand, so no invented sentiment.
• Answers from the official OpenAI, Anthropic, Google and Perplexity APIs with web search. These are not the app screens, so results can differ from what a person sees in each app (see "Honest limits" below).

What can you use it for?

• Check a client's AI visibility. Find out if ChatGPT, Claude, Gemini or Perplexity names the brand when someone asks "best CRM software for small business".
• Compare the brand with its competitors. See who gets mentioned first, how often, and who has the share of voice.
• Find the pages that feed the answers. The exact cited pages show which articles and review sites to get listed on, and which competitor page to beat.
• Read what the AI says about a client's brand. The exact sentences, side by side across assistants.
• Feed your client report. Schedule the same questions every week: each row is dated and carries the exact sentence and cited page, ready to paste into your own report. You get data rows and a Run summary, not a finished client report.
• Test a market before launch. Ask the buying questions of a new niche and see which brands already own the answers.

Running this for clients? Sillage Data also offers a done-for-you white-label report with your agency's logo, as a separate service (not part of this Actor). See how it works

How to use it in 3 steps

1. Enter the brand name (your client's), its website domain and up to 5 competitors. Choose one or several assistants under Answer engines (default: ChatGPT).
2. Write the questions your client's customers ask (1 to 50).
3. Click Start. The example input is a small run: 2 questions, 1 assistant, $0.105. Read the Overview table, then the Run summary.

Real run: a known brand that loses (29 Oct 2026)

Brand Moz, competitors Semrush, Ahrefs and SE Ranking, 10 SEO buying questions ("best keyword research tool", "best SEO reporting tool for clients"...), 1 answer each on ChatGPT, Claude, Gemini and Perplexity = 40 answers.

Brand	Named in
Semrush	36 of 40
Ahrefs	31 of 40
SE Ranking	25 of 40
Moz	14 of 40 (ChatGPT 2 of 10, Claude 7 of 10, Gemini 2 of 10, Perplexity 3 of 10)

Claude named Moz 7 times out of 10, the three others 2 or 3 times: one assistant is not the market. What to do with it: each row lists the pages the assistant cited, with the rank of each tracked name; read those pages to see what the assistant relied on when it named a rival. List price for this run: $0.005 + 10 x ($0.05 + $0.12 + $0.07 + $0.03) = $2.705. One answer per question on one day is a starting point, not a verdict: answers change. For a client, use 3 answers per question and schedule the run.

Real output example: one ChatGPT row (15 Oct 2026)

Input: brand HubSpot, competitors Salesforce and Zoho, question "best CRM software for small business". One row of the result, trimmed:

Field	Value
Model	gpt-6-luna (OpenAI API with web search, country US)
Brand mentioned	yes, 4 times
Brand rank	1 (Zoho is 2, Salesforce is not mentioned)
Brand site cited as a source	yes (hubspot.com)
Share of voice	0.5 (HubSpot 4 mentions, Zoho 4 mentions)
Cited sources	hubspot.com, pipedrive.com, zoho.com

{
  "checkedAt": "2026-10-15T10:03:49.171Z",
  "engine": "chatgpt",
  "model": "gpt-6-luna",
  "searchMode": "api_with_web_search",
  "question": "best CRM software for small business",
  "status": "ok",
  "brandMentioned": true,
  "brandMentionCount": 4,
  "brandRank": 1,
  "brandCitedAsSource": true,
  "brandShareOfVoice": 0.5,
  "competitors": [
    { "name": "Salesforce", "mentioned": false, "mentionCount": 0, "rank": null, "citedAsSource": false },
    { "name": "Zoho", "mentioned": true, "mentionCount": 4, "rank": 2, "citedAsSource": true }
  ],
  "searchQueries": [
    "best CRM software small business 2026 HubSpot Zoho Pipedrive pricing official",
    "site:hubspot.com/pricing CRM free sales hub pricing"
  ],
  "sources": [
    { "url": "https://www.hubspot.com/pricing/crm?tier=enterprise", "title": "Free Tools | HubSpot", "domain": "hubspot.com", "position": 1 },
    { "url": "https://www.pipedrive.com/en/pricing/professional-crm", "title": "CRM Pricing Plans | Affordable CRM Software Costs | Pipedrive", "domain": "pipedrive.com", "position": 2 },
    { "url": "https://www.zoho.com/en-us/crm/zohocrm-pricing.html?facet1=google-sheets", "title": "Zoho CRM Pricing and Editions - Free for 3 users", "domain": "zoho.com", "position": 3 }
  ],
  "answerText": "**Best default for most small businesses: HubSpot** ... (full answer, about 1,350 characters)"
}

brandSentences from the Moz run of 29 Oct 2026, question "best tool to check domain authority", copied word for word:

• Claude: "This is the original "Domain Authority" metric, so if you specifically want the real DA score, Moz is the source."
• Perplexity: "Try Moz’s free Domain Authority Checker." (Moz is the brand tracked.)

Compare ChatGPT, Claude, Gemini and Perplexity side by side

Add Claude, Gemini or Perplexity to the Answer engines input (engines) and every question is also answered by each, through the official API with web search. You get the same rows, with engine set to claude, gemini or perplexity, so you can put the assistants side by side: who names the brand first, who cites its site, which competitor is named instead.

• Same counting, same fields. Mentions, rank, competitors, share of voice, cited pages and brand sentences are counted by code, exactly as for ChatGPT. Filter the dataset on engine.
• Opt-in. The default is ChatGPT only. Pick any mix of the four.
• Billed per successful answer. One successful answer is one check at that assistant's price. Empty or failed answers are never charged.
• More assistants, more checks. Questions x answers per question x assistants = number of checks.
• Official APIs, not the apps. Answers can differ from what a person sees in each app. Gemini's API does not take a country: its rows say countryApplied: false.

Example input to compare all four:

{
  "brandName": "HubSpot",
  "brandDomain": "hubspot.com",
  "competitors": ["Salesforce", "Zoho"],
  "questions": ["best CRM software for small business"],
  "engines": ["chatgpt", "claude", "gemini", "perplexity"]
}

The Run summary then has one block per assistant under engines, with the most cited pages, the brand's own pages cited and the number of sentences about the brand.

Prices are listed in the next section.

Example: 10 questions, 1 answer each, on all four = $0.005 + 10 x ($0.05 + $0.12 + $0.07 + $0.03) = $2.705. See "How much does it cost" below for the Actor start fee and budget limits.

How much does it cost to check your brand in ChatGPT, Claude, Gemini and Perplexity?

You pay for these events, and nothing else. Apify platform usage (compute, storage) is included in these prices.

Event	Price
Actor start (once per run)	$0.005
ChatGPT check (one question, one successful answer, fully analyzed)	$0.05
Claude check (one question, one successful answer, fully analyzed)	$0.12
Gemini check (one question, one successful answer, fully analyzed)	$0.07
Perplexity check (one question, one successful answer, fully analyzed)	$0.03

• Example 1, the default input (2 questions, 1 answer each): $0.005 + 2 x $0.05 = $0.105.
• Example 2, a weekly check for one client (10 questions, 3 answers each, 30 checks): $0.005 + 30 x $0.05 = $1.505.
• Example 3, the same weekly check on all four assistants (120 checks): $0.005 + 30 x ($0.05 + $0.12 + $0.07 + $0.03) = $8.105.

On an Apify account with the $5 monthly credit, one ChatGPT-only run of 99 checks (for example 33 questions x 3 answers) costs $4.955, so it fits in that credit. Each assistant you add multiplies the number of checks. Empty or failed answers are free (the one-time Actor start fee still applies). Set a maximum cost per run in Apify: the Actor checks your budget before every paid call and stops cleanly before going over it.

FAQ

How do I check if ChatGPT, Claude, Gemini or Perplexity mentions my brand?

Run this Actor with the brand name and the questions its customers ask, and pick the assistants you want. Each row tells you if the brand is mentioned, how many times, in which position compared with its competitors, and which exact pages of its website are cited as sources. It also keeps the exact sentences written about the brand.

What is AI share of voice?

It is the share of mentions a brand gets compared with its competitors in AI answers. This Actor computes it per answer (brandShareOfVoice) and for the whole run (shareOfVoice in the summary), by counting names in the answer text.

How is this different from an AI visibility score?

A score is a number someone computed for you. This Actor stores the full answer, the date, the model and the cited sources, so you can check every number yourself.

Why is there no sentiment or tone score?

Because a tone score needs another AI to judge the answer, and that judgment can be wrong. Everything here is counted by code, so the same answer always gives the same numbers.

Is this the same as asking in the ChatGPT, Claude, Gemini or Perplexity app?

No. Answers come from the official OpenAI (ChatGPT), Anthropic (Claude), Google (Gemini) and Perplexity (Sonar) APIs with web search enabled, not from the app screens. The app also uses memory, personalization and other settings, so what a person sees can differ. Use this Actor to measure trends and compare brands under the same conditions.

Which AI assistants does it support?

ChatGPT, Claude, Gemini and Perplexity, through their official APIs with web search.

Does it track Google AI Overviews, AI Mode, Copilot or Grok?

Not yet. It tracks ChatGPT, Claude, Gemini and Perplexity, all through official APIs with web search. Ask for another one on the Issues tab.

Do I need an OpenAI, Anthropic, Google or Perplexity API key?

No. The Actor uses its own access. You only pay the price per check shown above.

What happens if an answer is empty or fails?

The row is saved with status set to empty or error and an errorMessage, and that check is not charged. If every check fails, the run is marked as failed.

Is it legal?

It uses the official OpenAI, Anthropic, Google and Perplexity APIs under their terms. It does not scrape the assistants' websites or any other website, and it does not collect personal data: the input is brand names and questions. The answers are AI-generated content, and each row says so in the note field.

Can I run it every week and send the data elsewhere?

Yes. Use Apify Schedules to run the same input every Monday and build a history. Send results to Google Sheets, Slack, Make, Zapier or n8n with Apify integrations, call it through the Apify API, or let an AI agent call it through the Apify MCP server (turn off includeFullAnswer for a compact output).

Output fields and honest limits

One row per question and answer.

Field	Meaning
checkedAt	When the answer was fetched (ISO 8601, UTC)
engine, model, searchMode	chatgpt, claude, gemini or perplexity, exact model name from the API, api_with_web_search
country, countryApplied	The country you chose, and whether the web search could use it (false if the engine refused it)
question, sampleIndex	The question and which repetition it is
status	ok, empty or error (only ok is charged)
brandMentioned, brandMentionCount	Is the brand in the answer, and how many times
brandRank	Order of first appearance among the brand and competitors (1 = first, null = absent)
brandFirstMentionChar	Character position of the first mention
brandCitedAsSource	The brand's domain (or a subdomain) is among the cited sources
brandInConsultedSources	The brand's domain is among all pages the search consulted
brandShareOfVoice	The brand's mentions divided by all tracked mentions in this answer
competitors	Same counts for each competitor
sources	Cited pages: full url, title, domain, position
brandCitedPages	The pages of the brand's site (url, title) that this answer cited
brandSentences, brandSentenceCount	The exact sentences of the answer that name the brand, and how many (cut by code, no tone score)
searchQueries	The web searches run to answer
answerText	The complete answer (if includeFullAnswer)
note	Reminder that this is the official API of the engine, with web search

The Run summary (key-value store record SUMMARY) adds, per engine: brand presence rate, average rank, cited-as-source rate, share of voice for every tracked name, competitor presence rates, the 10 most cited domains, the 15 most cited exact pages (topCitedPages), the brand's own pages cited (brandPagesCited) and the number of sentences about the brand.

Honest limits

• These are the official APIs with web search enabled, not the app interfaces. Answers can differ from what a person sees in the ChatGPT, Claude, Gemini or Perplexity app, which may use other models, memory or personalization.
• AI answers vary from one run to the next. Use 3 answers per question to measure a rate instead of trusting one answer.
• Rank is among the names you track, not among every brand in the answer.
• Name matching is whole word, with one tolerance. A name of 5+ letters also counts when glued to AI, .ai, HQ, .com, .io or App (Otterly = OtterlyAI = Otterly.ai). Plurals and other words do not count (Sillage is not Sillages). Add brandAliases for any other spelling.
• A competitor counts as "cited as source" when a cited domain label equals its name (for example zoho.com for "Zoho").
• On some Claude answers the web search may run without the country setting, and Gemini's API takes no country at all (the row field countryApplied says so).
• Perplexity always runs perplexity/sonar (Perplexity's own search model). A page counts as "cited" when the answer marks it as a source; if the answer marks none, every page found counts as cited. Compare its cited-as-source rate with its own history, not with other assistants.
• Gemini links its sources through Google redirects; the Actor tries to open each one to get the real page address; if that fails, the redirect link is kept.
• Not affiliated with OpenAI, Anthropic, Google or Perplexity. ChatGPT is a trademark of OpenAI, Claude of Anthropic, Gemini of Google, and Perplexity of Perplexity AI. Moz, Semrush, Ahrefs and SE Ranking are named only as a measured example.

Input settings

Field	What it does	Default
brandName	Brand to track (whole word, case insensitive). Required.	HubSpot (prefilled)
brandDomain	The brand's domain, to detect when its site is cited	hubspot.com
competitors	Up to 5 competitor names	Salesforce, Zoho
questions	1 to 50 questions. Required.	2 examples (prefilled)
brandAliases	Other spellings that count as the brand	empty
engines	Which assistants answer: any of chatgpt, claude, gemini, perplexity	chatgpt
chatgptModel, claudeModel, geminiModel	Exact API model name per assistant. Defaults: gpt-6-luna, claude-sonnet-5, gemini-3.5-flash. Each row stores the model that answered (model). Perplexity has no model field: it always uses perplexity/sonar	the defaults
samplesPerQuestion	Ask each question 1 to 3 times (answers vary, each answer is one check per assistant)	1
country	Approximate location sent to the web search	US
includeFullAnswer	Keep full answer text and all consulted URLs	true

Our other tools

• Flight Price Tracker: Fare Drop Alerts
• TikTok Influencer Finder by Engagement Rate
• SAM.gov Contract Opportunities Monitor & Alerts

Support

Found a bug or need another assistant? Open an issue on the Issues tab. We reply there. More tools: sillagedata.com/tools

Found this useful? Leave a quick review on the Reviews tab: it takes 10 seconds and helps us keep improving it.