AI Brand Visibility Tracker — ChatGPT & Google AI Overviews

Find out whether Google AI Overviews and ChatGPT recommend your brand when customers ask about your category. The actor asks your prompts in a real browser, exactly as a visitor sees them — the AI Overview on the Google results page and, as a beta option, ChatGPT without login, with its live web search — and measures your brand visibility: mention, rank among all brands named, share of voice against your competitors, tone, and whether your own website is cited as a source. Every cited link is resolved to its real URL.

• Real answers, not an LLM API. Most AI visibility trackers send your prompt to a model API, which has no live web search and no AI Overview. This actor reads the answers your customers actually get.
• 43 fields per row: full answer text and markdown, every cited source (URL, domain, title, publisher, snippet), the brands the engine itself recognized with their websites, your brand's position, context and sentiment, competitors named and not named, and more.
• Brand report per run: mention rate, average position, share of voice per engine, most cited domains, prompts where you are missing, and changes since your previous run.
• Pay only for answers: a check that fails, or a Google search that shows no AI Overview, costs nothing.

ChatGPT is in beta. Google AI Overviews is the default engine and is fast and reliable. ChatGPT is an explicit option: from Apify's servers it currently accepts only some browser sessions. On 14 Oct 2026, 158 of 488 ChatGPT attempts (32 %) were accepted in 50 test runs; retries still returned an answer for 158 of 162 ChatGPT checks, but a check can take several minutes, and runs limited to 5 minutes got a ChatGPT answer in 8 of 10. A ChatGPT check without an answer is never charged.

Who uses it

• SEO and GEO teams (generative engine optimization): track whether ChatGPT and Google's AI answers mention your brand for "best X", "X alternatives" and "X vs Y" questions, and which sites they cite so you know where to get featured.
• Marketing and brand managers: share of voice in AI search against named competitors, by engine and country, with a monthly or weekly trend from scheduled runs.
• Agencies: one run per client with the client's prompts, brand aliases and competitors; export JSON, CSV or Excel, or call it from the Apify API.
• Product and PR teams: see the exact sentences AI engines write about your product, their tone, and the sources they rely on.

How it compares (measured 14 Oct 2026)

We read the published input and output schemas of the AI-visibility and AI-answer actors in Apify Store through the Apify API, and measured our own success rate on the Apify platform. No third-party actor was run.

	This actor	LLM-API brand trackers	ChatGPT / AI Overview answer scrapers
Source of the answer	Real Google results page and ChatGPT (beta) in a browser	Model API (no live web search, no AI Overview)	Real answer
Brand metrics (mention, position, share of voice, sentiment)	Yes, per row and in a run report	Yes	No
Cited sources with real URLs	Yes: median 18 per ChatGPT answer, 7 per AI Overview (13–14 Oct, n=261 and 145)	Some	Some
Brands the engine itself recognized	Yes, with their websites	No	Only one of them
Country and language	5 countries, 3 languages	Missing in 2 of the 5 most used	Partly
Output fields declared in the schema	43	0–20 (7 of 11 actors declare none)	0–18
Fields filled per row (measured)	36 per row (median of 306 answers, 14 Oct 2026)	at most 20 (all declared fields; not measured, no public sample)	at most 18
Price of the main event	$0.02 per AI Overview; $0.083 per ChatGPT answer (beta)	$0.05–$0.30 per check	$0.20 per ChatGPT answer; $0.003 per AI Overview without brand metrics
Failed checks charged	Never	Varies	Varies

In 50 test runs on the Apify platform with varied inputs (5 countries, 3 languages, 1 to 48 checks per run, 14 Oct 2026) 50 of 50 runs succeeded, charges matched the delivered answers in all 50, no row was duplicated, and 315 of 319 checks returned a result (306 answers plus 9 Google searches that showed no AI Overview; the other 4 were refused by the source after every retry and were not charged). In the morning of the same day, before ChatGPT's acceptance rate dropped, 95–100 % of ChatGPT checks returned an answer (n=20 per proxy route, three routes) and 60 of 60 Google checks were read, 95 % of them with an AI Overview. The brand mentioned in the prompt was found in 89 % of ChatGPT answers and 92 % of AI Overviews; a made-up control brand was found in 0 of 407 answers.

What you get

Group	Fields
Check	rowType, checkId (stable across runs), engine, prompt, country, language, repetition, checkedAt, billable, attempts, responseMs
Answer	answerText, answerMarkdown (ChatGPT), answerLength, answerSource, aiOverviewPresent and aiOverviewLabel (Google), answerFormat, imageCount, citeMarkerCount
Your brand	brandMentioned, brandMentionCount, brandPosition, brandContext, brandSentiment, brandShareOfVoice, brandInEntities, brandEntityCategory, brandCitedAsSource, brandCitations
Competitors	competitorsMentioned (position, mentions, context, sentiment), competitorsNotMentioned, competitorsCitedAsSource, brandsRanking (every brand named, in order), totalBrandsMentioned
Sources	citations (position, url, domain, title, publisher, snippet, kind), citationCount, citedDomains, hiddenSourceCount (Google "+N"), links written in the answer (ChatGPT)
Entities	entities: brands, products and organizations the engine recognized, with category and website
Errors	error with a code (blocked, no-answer, source-changed, timeout, input) and a plain-English message; error rows are never charged

Every row has every field; a value the engine did not give is null, never missing. The brand report is saved as BRAND_REPORT in the run's key-value store, and checks that could not be completed are listed in ERRORS.

Pricing

Pay per event:

Event	Price
Google AI Overview checked	$0.02 (not charged when Google shows no AI Overview)
ChatGPT answer checked (beta)	$0.083 (not charged when ChatGPT gives no answer)
Brand report (once per run, only if at least one answer was read)	$0.10
Actor start	$0.00005 per GB of memory ($0.0001 per run at the default 2 GB)

Example: 20 prompts on Google AI Overviews in 1 country with the report = $0.50 at most; adding ChatGPT for the same 20 prompts adds up to $1.66. Failed checks, duplicates and Google pages without an AI Overview are never charged. If you set a maximum cost per run, the actor checks the budget before opening each browser and stops cleanly when it is reached.

How to use it

1. Enter your brand and, optionally, aliases and your own website domains.
2. Paste the prompts your customers ask, one per line ("best crm for small business", "HubSpot alternatives", "HubSpot vs Salesforce").
3. Add competitors (optional aliases after a vertical bar: Salesforce | Sales Cloud).
4. Pick the engines and countries and run. The default run (1 prompt on Google AI Overviews) took a median of 16 seconds on the Apify platform and finished within 2 minutes in 8 of 8 runs, the slowest in 41 seconds (n=8, 14 Oct 2026). Each check runs in a real browser, 2 at a time by default. With ChatGPT (beta) selected, a run can take several minutes: give it a timeout of at least 15 minutes.
5. Schedule the same input weekly: the report shows the changes since the previous run.

Input example

{
  "brand": "HubSpot",
  "brandDomains": ["hubspot.com"],
  "prompts": ["best crm for small business", "HubSpot alternatives", "HubSpot vs Salesforce"],
  "competitors": ["Salesforce", "Zoho", "Pipedrive"],
  "engines": ["google-aio", "chatgpt"],
  "countries": ["us", "gb"],
  "runsPerPrompt": 1,
  "includeReport": true
}

Output example (abridged)

{
  "rowType": "prompt-check",
  "engine": "chatgpt",
  "prompt": "best crm for small business",
  "country": "us",
  "billable": true,
  "brandMentioned": true,
  "brandPosition": 1,
  "brandSentiment": "neutral",
  "brandShareOfVoice": 0.3,
  "brandInEntities": true,
  "brandCitedAsSource": true,
  "competitorsMentioned": [{ "name": "Zoho", "position": 2, "mentionCount": 4 }, { "name": "Pipedrive", "position": 3, "mentionCount": 3 }],
  "competitorsNotMentioned": ["Salesforce"],
  "entities": [{ "name": "HubSpot CRM", "category": "software", "position": 1, "url": "https://www.hubspot.com/products/crm" }],
  "citationCount": 29,
  "citedDomains": ["bestcrmpick.com", "fitsmallbusiness.com", "hubspot.com", "techradar.com", "zoho.com"]
}

Good to know

• Engines: Google AI Overviews (default) and ChatGPT (beta: public, without login, with web search; see the note at the top for its measured answer rate). Google AI Mode, Gemini, Copilot and Perplexity are not included: the last two require an account, and the actor never logs in anywhere.
• Countries: United States, United Kingdom, Mexico, Spain and Germany, in English, Spanish or German. The country sets Google's country and language and the browser's locale and time zone; ChatGPT answers follow the language of your prompt.
• AI answers vary. The same prompt can name different sources on different days. In our tests the brand-mention verdict for the same ChatGPT prompt matched 93 % of the time while the cited domains changed a lot; use Repetitions per prompt (2–3) for steadier citation data.
• Sentiment is a word-list heuristic in English, Spanish and German. Use it as a signal, not as a verdict.
• Google does not show an AI Overview for every search. Those rows are delivered with aiOverviewPresent: false so you can see it, and are not charged.
• No CAPTCHA solving: when an engine asks for a verification, the actor drops that connection and retries with a fresh one; it never solves or bypasses challenges.
• Timeouts: the actor stops before your run timeout and keeps everything delivered; resurrect the run to finish the remaining checks without paying twice.

FAQ

Does it use the OpenAI API? No. It reads the public ChatGPT page in a real browser, so you see the same answer, sources and recommended brands as a visitor.

Why is a check not charged? Errors, empty answers and Google searches without an AI Overview are free. Only rows with billable: true are charged.

Can I compare runs over time? Yes. checkId is stable for the same prompt, engine, country and language, and the brand report includes the changes since your previous run with the same brand, engines and countries.