AI Citation Tracker — ChatGPT & Perplexity Sources

AI Citation Tracker records which websites and URLs ChatGPT, Perplexity, Claude, and Gemini cite when they answer your buyer prompts, whether your domain and your competitors' domains are among them, and which sources were gained or lost since the last run. It is built for SEO, content, and digital PR teams doing AI citation tracking, and it costs USD 50.001 per 1,000 prompt checks (USD 0.050001 per prompt × engine) on the FREE tier, plus a USD 0.010001 start fee.

Free users: maximum 20 results per run.

Which websites do ChatGPT and Perplexity cite for my topic?

Give the tracker the questions your buyers ask, and each selected engine answers them with live web search; every cited URL is collected, grouped by domain, and checked against your domain and your competitors' domains. The run summary then ranks the most-cited domains, so you see the publications AI answers rely on in your niche.

{
  "brandDomain": "notion.so",
  "competitorDomains": ["evernote.com", "obsidian.md"],
  "prompts": [
    "What is the best note-taking and productivity app?",
    "Recommend the top productivity software for teams."
  ],
  "platforms": ["perplexity", "chatgpt", "claude", "gemini"],
  "trackingId": "notion-weekly",
  "maxPrompts": 25
}

Every prompt runs on every selected engine, so 2 prompts × 4 engines = 8 checks. With no prompts, four category prompts are generated from industry. The Console form starts with one prompt on ChatGPT for a quick, low-cost first run.

How do I track AI citations over time?

Schedule the tracker with the same prompts and the same trackingId. Each run compares every prompt × engine pair with its previous snapshot and labels it new, gained, lost, unchanged, or absent, and lists new_domains and lost_domains. History is kept in a named key-value store in your account (ai-citation-history-<trackingId>), so trends survive between runs until you delete it.

What does the AI citation tracker return?

One dataset row per prompt × engine check:

Field	What it tells you
brand_domain	The domain you want to see cited
engine, model_used	perplexity, chatgpt, claude, or gemini, and the neutral label fixed-ai-model (each engine uses one fixed model)
prompt	The question that was asked
brand_cited, brand_cited_urls	Whether your domain was cited, and which of your URLs
citation_urls, citation_domains, total_citations	Every cited URL, the unique cited domains, and the URL count
competitor_domains_cited	Competitor domains among the citations
new_domains, lost_domains	Domains cited now but not last run, and the reverse
previous_brand_cited	Last run's brand_cited, or null on the first run
citation_status	new, gained, lost, unchanged, absent, or error
error, checked_at	Failure reason (or null) and the check time

The SUMMARY record holds checks_total, brand_citation_rate, top_cited_domains (domain, count, share), citation_gap_domains (domains cited more often than yours), gained, lost, and stop_reason.

This row is from a real check that had a previous snapshot (model_used omitted):

{
  "brand_domain": "notion.so",
  "engine": "chatgpt",
  "prompt": "What is the best note-taking and productivity app?",
  "brand_cited": false,
  "brand_cited_urls": [],
  "citation_urls": [
    "https://appstested.com/best-note-taking-apps-2026/",
    "https://zapier.com/blog/best-note-taking-apps/"
  ],
  "citation_domains": ["appstested.com", "zapier.com"],
  "competitor_domains_cited": [],
  "total_citations": 2,
  "new_domains": ["appstested.com"],
  "lost_domains": ["softwareadvice.com"],
  "previous_brand_cited": false,
  "citation_status": "absent",
  "error": null,
  "checked_at": "2026-10-26T07:42:06.989Z"
}

Here the answer cited two review sites and not notion.so, and one source from the previous run dropped out: those review sites are the pitch list.

How much does AI citation tracking cost?

Pay-per-event prices, active since September 20, 2026, with no change scheduled. Model and web-search costs and Apify platform usage are included.

Event	FREE	BRONZE	SILVER	GOLD	PLATINUM	DIAMOND
prompt-engine-check: one successful prompt × engine (primary)	$0.050001	$0.04875	$0.0475	$0.04	$0.04	$0.04
apify-actor-start: run start	$0.010001	$0.00975	$0.0095	$0.008	$0.008	$0.008

The start fee is charged once per started GB of allocated memory; the default 256 MB run pays one start event. Free users: maximum 20 results per run.

• Store prefill (1 prompt on ChatGPT), FREE: $0.010001 + $0.050001 = $0.060002.
• Generated prompts on all engines (4 × 4 = 16 checks), GOLD: $0.008 + 16 × $0.04 = $0.648.

Failed checks are saved with citation_status: "error" and not charged. An answer that does not cite your domain is a valid, charged result. A maximum total charge set on the run stops the Actor when reached.

When should I use AI Citation Tracker, and when not?

Use it for a source-level question: which domains and pages do AI answers cite for these prompts, and is my site one of them? Choose a sibling Actor when the question is different:

• Is my brand mentioned, ranked, and described well in each answer? AEO Tracker.
• A quick yes/no brand check on four platforms: AI Brand Mentions Checker.
• Prioritized GEO recommendations and an HTML report: Generative Engine Optimization Tool.
• Share of voice and rank trends: AI Visibility Rank Tracker.
• Citations in Google AI Overviews: Google AI Overviews Scraper.

It is not a backlink crawler, a Google rank tracker, or a copy of anyone's logged-in chat session.

Use it from an AI assistant

AI Citation Tracker works from Claude, ChatGPT, Cursor, or any MCP client through Apify's MCP server. Add it as a single tool with https://mcp.apify.com/?tools=muhammadafzal/ai-citation-monitor, then ask, for example: "Which websites do Perplexity and ChatGPT cite when someone asks for the best CRM for small businesses, and is hubspot.com one of them?"

How do I run it from the API?

JavaScript (apify-client):

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('muhammadafzal/ai-citation-monitor').call({
  brandDomain: 'notion.so',
  competitorDomains: ['evernote.com'],
  prompts: ['What is the best note-taking and productivity app?'],
  trackingId: 'notion-weekly',
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
const summary = await client.keyValueStore(run.defaultKeyValueStoreId).getRecord('SUMMARY');
console.log(items.length, summary?.value.top_cited_domains);

Python (apify-client 3.x):

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("muhammadafzal/ai-citation-monitor").call(run_input={
    "brandDomain": "notion.so",
    "competitorDomains": ["evernote.com"],
    "prompts": ["What is the best note-taking and productivity app?"],
    "trackingId": "notion-weekly",
})
for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item["engine"], item["brand_cited"], item["citation_domains"])

Apify schedules, webhooks, and integrations with Make, Zapier, n8n, and Google Sheets can run weekly citation tracking per client, each with its own trackingId.

Limits and reliability

• Up to 200 prompts per run on four engines.
• Each answer request times out after 45 seconds without retry; a failed check is saved with an error.
• Answers are capped at 500 tokens, and ChatGPT, Claude, and Gemini get two web results per request (Perplexity uses its own search), which bounds how many sources an answer can cite.
• Domains are grouped by a simplified rule (last two labels, or three for suffixes such as co.uk), not the full public suffix list.
• Citations change between runs and with web results; treat a single lost as a signal to watch, not a verdict.

FAQ

What counts as an AI citation?

Any source URL attached to the answer as a citation or written in its text. Formatting depends on each engine's web grounding.

What are LLM citations?

LLM citations are the web sources a large language model links to when it answers with web search. This tracker collects them per prompt and engine.

Can I track Perplexity citations only?

Yes. Set platforms to ["perplexity"]. Each prompt then costs one check.

Why did my status change to lost when my site did not change?

Engines pick different sources between runs, and web results shift. Look at several runs before acting.

Does a missing citation mean AI engines ignore my brand?

No. An answer can mention a brand without linking it. AEO Tracker measures mentions.

Do I need an AI API key?

No. Model and search costs are paid by the Actor owner inside the event price.

Related Actors

• AEO Tracker: brand mentions, rank, sentiment, and own-site citations per answer engine.
• AI Brand Mentions Checker: one-question AI brand check across four platforms.
• Generative Engine Optimization Tool: GEO score, recommendations, and HTML report.
• AI Visibility Rank Tracker: AI share of voice and position changes over time.
• ChatGPT Brand Visibility Tracker: ChatGPT-style answers only.
• GEO Content Auditor: citation readiness scores for your own pages.

Legal and responsible use

AI Citation Tracker sends ordinary prompts to AI model APIs and records the answers and cited URLs; it does not access private accounts. AI answers and sources can be wrong or outdated, so verify important claims against the cited pages.

Last verified: 2026-10-27