AI Visibility Audit: ChatGPT, Perplexity & Google Brand Report

Produce a brand visibility report from up to 20 buyer questions on 3 engines. Compare your brand with competitors, inspect the sources AI engines cite, and get a transparent visibility score, competitor gaps and a prioritised content review list.

Supply questions yourself, or provide a public website and category to generate questions in one low-cost model call. The engine answers and any question-generation call use Apify’s gateway and are billed to the buyer separately from the audit event.

What you get

Field	What it holds
Audit row	Delivery status, visibilityScore, usableAnswers, requestedAnswers, resolvedAnswers and reportRecord.
Answer rows	Original answer text, engine, model, source URLs, matched brands and returned AI usage. Failure rows are free.
REPORT.html	One standalone page with headline metrics, engine and brand tables, competitor gaps, recommendations, questions and answer excerpts.
REPORT / SUMMARY	JSON evidence, score components and formula, coverage, question-generation provenance, traffic, notes and charges.
CHANGES	With projectName: mention gains/losses, rate moves, cited sites gained/lost and tracked competitors appearing, per question and engine.
Named project store	Immutable RUN records plus LATEST; the exact name and run record are in REPORT.monitoring.

What people use it for

• Agency reports: Share a single HTML file explaining where a brand is visible and what the sample supports.
• Competitive audits: Find buyer questions where specified competitors are mentioned or cited without the brand.
• Content reviews: Inspect cited page titles and headings alongside topic-specific recommendations.
• Repeat measurements: Use a project name and consistent questions to record comparable changes.

How to use it

1. Enter your brand, website and category, plus competitors as names and domains.
2. Supply up to 20 questions or leave questions empty to generate them from the website and category. Select up to three engines.
3. Set the event budget to at least $4.00, allow for gateway AI usage separately, and start.
4. Open REPORT.html. Check coverage, questions, score formula and evidence before sharing the report.

Input

Field	Example	Notes
brand	Example Workspace	Required.
website	https://example.com	Required for generated questions; optional for supplied questions. Also identifies the subject’s citation domain.
category	Project management software for small teams	Required with a website when generating questions.
competitors	Example Planner | planner.example	Up to 10. Unlisted brands are not guessed.
questions	Buyer questions, one per line	Optional; up to 20 distinct questions. Supplied questions skip generation.
questionCount	20	Number generated when questions are absent; 1–20.
engines	openai, perplexity, google-ai-overview	Select 1–3 of the six supported engines.
country / language	US / en	Location where supported; Google result language and question-generation context.
projectName	Example project	Optional history and comparisons.
maxCitedPages	8	0–12 direct public page reads for evidence.

Output

Each result is one row in the run's dataset (download it as JSON, CSV, Excel, XML or HTML, or read it through the API). An example row (values shortened):

{
  "kind": "audit",
  "status": "delivered",
  "brand": "Example Workspace",
  "visibilityScore": 82.0,
  "usableAnswers": 5,
  "requestedAnswers": 6,
  "resolvedAnswers": 6,
  "engines": [
    "openai",
    "perplexity",
    "google-ai-overview"
  ],
  "reportRecord": "REPORT.html",
  "note": "Report saved. AI usage is billed separately by the gateway.",
  "charged": true
}

This example is the saved audit row from a free offline replay on 17 October 2026 of recorded engine answers from 12 October: 5 usable answers from 6 requests, one without a readable Google overview. Brand names and site addresses have been replaced. The 82.0 score is calculated from that sample; it is not a new live measurement. Page-reading fixtures were used for offline report tests. SUMMARY records the questions, exact score components, coverage and charges.

Pricing

• $4.00 per delivered audit, for up to 20 questions on up to three engines. There are no per-answer events.
• The audit event is charged only after saving the report, with at least one usable answer and at least 80% of requests resolved as an answer or a reported unreadable Google overview. Smaller successful audits use the same price.
• Below that threshold, the diagnostic report and answer rows have no audit event. Failed or absent answers never lower the score by being treated as zero visibility.
• All model usage is separate, billed to the buyer through Apify’s gateway. Question generation uses one newest-Gemini-Flash call without a search plugin. Engine calls use their configured web search.
• No residential proxy is used for cited-page evidence. Google Overview uses the platform’s SERP route; its cost is included in the audit event.
• Measured 17 October 2026: a 3-question audit on Perplexity delivered 3 of 3 answers and its report; the AI usage was about $0.08 on the buyer's account.

What typical jobs cost at the list price:

Job	Cost
20 questions × 3 engines, supplied questions	$4.00 plus up to 60 engine calls billed by the gateway where applicable
20 generated questions × 3 engines	$4.00 plus one generation call and engine usage
2 supplied questions × 3 engines	$4.00 plus engine usage
No usable answers or under 80% resolved requests	$0 audit events; gateway calls may still cost money

The run’s maximum cost limits Actor event charges. Gateway model usage is billed separately and is not capped by this event limit. Failed or cancelled model calls can still incur gateway charges. Check your gateway/account budget as well.

Tips

• Use unbranded buyer questions to measure discovery rather than prompting engines to name the brand.
• Keep and review the generated questions before relying on a recurring comparison. Use the same saved questions in later runs.
• Track a consistent competitor set; mention order is relative to those brands.
• For ongoing larger question sets or repeated samples, use the per-answer tracker family.

Use it through the API

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("zaiq/ai-visibility-audit").call(run_input={"brand": "Example Workspace", "website": "https://example.com", "category": "Project management software for small teams", "competitors": ["Example Planner | planner.example"], "questions": ["Which project management tools suit a small remote team?"], "engines": ["openai", "perplexity", "google-ai-overview"]})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(row)

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('zaiq/ai-visibility-audit').call({"brand": "Example Workspace", "website": "https://example.com", "category": "Project management software for small teams", "competitors": ["Example Planner | planner.example"], "questions": ["Which project management tools suit a small remote team?"], "engines": ["openai", "perplexity", "google-ai-overview"]});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

HTTP (one call, results back)

curl -X POST 'https://api.apify.com/v2/acts/zaiq~ai-visibility-audit/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>' \
  -H 'Content-Type: application/json' \
  -d '{"brand": "Example Workspace", "website": "https://example.com", "category": "Project management software for small teams", "competitors": ["Example Planner | planner.example"], "questions": ["Which project management tools suit a small remote team?"], "engines": ["openai", "perplexity", "google-ai-overview"]}'

Integrations and schedules

• Schedules: run it every hour, day or week from the Apify Console (Schedules) and keep a history of results.
• Make, Zapier, n8n, Google Sheets, Slack, Airtable: connect it through Apify's integrations, or send each finished run to a webhook.
• Datasets: results stay in the run's dataset; export them or read them by API whenever you need them.

Use it with AI agents (MCP)

Add Apify's MCP server (https://mcp.apify.com) to Claude, ChatGPT, Cursor, VS Code or any MCP client. The agent can find this Actor and run it for you: ask it for "audit a brand against its competitors and create an AI visibility report with cited-page evidence". AI agents can also pay for runs directly through Apify's agentic payments.

Limits

• API answers use vendor models with web search. They can differ from consumer apps because of personalisation, interface behaviour and model settings.
• A mention is a boundary-matched brand name or supplied domain in the answer. A citation is a link to the brand domain or its subdomains. Name-only citation matching is a heuristic; supply a domain for more reliable matches.
• Citation share in the tracker means the fraction of usable answers citing a brand, not its fraction of all source links. Failed answers and unreadable overviews are excluded from visibility rates; coverage is shown separately.
• The first tracked brand is the report subject. Other supplied brands are competitors. Ordinary trackers do not infer unnamed competitors. Shopping reports also use brands returned by the existing product parser.
• Recommendations use bounded page titles, headings and descriptions. They do not prove the brand lacks a page, verify factual accuracy, or promise improved rankings. Blocked, robots-disallowed and unread pages are labelled.
• Project history compares matching question/engine pairs only when successful repeat identities, counts and models match. Changing the tracked brands, country, language or Actor family creates a new baseline. An all-failed run is retained but does not replace the last usable baseline.
• Use a unique project name per engagement and schedule its runs sequentially. Concurrent runs share a last-writer-wins LATEST pointer. Named stores retain immutable run records until you remove them or your account retention policy applies.
• There is no login or CAPTCHA solving. The cited-page reader uses direct public connections, checks robots.txt, and stops using a host after a 429, 401 or 403. It reads at most 12 cited pages and uses no residential route.
• At most 20 questions, three engines and one sample per question/engine. Engine work is bounded to 900 seconds and evidence reading to 120 seconds; unfinished work is reported.
• Question generation requires readable website text and a category. A blocked site, incomplete model response or brand-leading generated question results in a free diagnostic audit; supply your own questions to proceed.
• The score is 100 × (0.60 × mention rate + 0.30 × citation rate + 0.10 × first-mention rate), with usable answers as the denominator. It is a transparent sample index, not an industry benchmark.

FAQ

Does this log into the AI apps?

No. Model answers come through the gateway’s official model APIs with web search; Google Overview comes from its results page.

Are AI costs included in $4.00?

No. Gateway usage is separate and can exceed the Actor event price.

Can I get a report without generated questions?

Yes. Supply questions; no question-generation call is made.

Does a zero score mean the engine failed?

No. Zero means usable answers gave the brand no measured visibility. When there are no usable answers, the score is null and displayed as Not measured.

Can this guarantee more AI citations?

No. Recommendations are evidence for a content review, not ranking guarantees.

Related Actors from Zaiq

• ai-answer-tracker: Ongoing monitoring with per-answer pricing and repeated samples.
• ai-shopping-tracker: Product recommendation and shopping visibility tracking.

Support

Questions, a site that changed, or a field you need? Open an issue on this Actor's Issues tab: we answer there, usually within a day.

---

Also found as: AI visibility audit, AI brand report, AI competitor audit, GEO audit, AI citation report.