AI Search Citation Monitor — Google AI Overview & AI Mode

Find out which websites Google's AI answers cite for the keywords you care about. Give the Actor a list of search queries and, optionally, the domains you want to track (your site, your clients, your competitors). For every query it fetches the Google results page, extracts the AI Overview block, and returns the answer text, every cited source with its real URL and domain, a yes/no hit for each target domain, and the organic top 10 for comparison.

Use it for GEO (generative engine optimization) reporting, AI-visibility tracking, content gap analysis, or simply to see whether Google's AI sends traffic to you or to someone else.

What you get

One dataset record per query and engine:

Field	Type	Description
query	string	The query you passed in
engine	google_ai_overview | google_ai_mode	Which AI surface was checked
countryCode, languageCode	string	Localization used for the search
hasAiAnswer	boolean	Whether an AI answer was present on the page
answerText	string | null	Plain text of the AI answer as served in the results page
answerMarkdown	string | null	The same answer as Markdown, with inline links resolved to real URLs
citations	{ position, url, domain, title, resolved }[]	Sources cited by the AI answer in order of appearance. domain is without www.
citedDomains	string[]	Unique cited domains, first-seen order
targetDomainHits	{ domain, cited, positions }[]	One entry per target domain from the input. Subdomains count as hits
organicTop10	{ position, url, domain, title, resolved }[]	Up to 10 organic results from the same page
extractionMethod	string | null	Which extraction path found the answer (useful when Google changes its markup)
serpUrl	string	The Google URL that was requested
fetchedAt	string	ISO 8601 timestamp
status	ok | failed | skipped	failed = blocked, network error or AI Mode unavailable; skipped = spending cap reached
error	string	Present only when status is not ok

The run always finishes as SUCCEEDED, even when some queries fail. Check status per record instead of the run status.

Input

Field	Type	Default	Description
queries	string[]	required	Search queries. One record per query and engine
targetDomains	string[]	[]	Domains to track, e.g. ["example.com", "competitor.io"]. Protocol, www. and paths are stripped
engines	string[]	["google_ai_overview"]	google_ai_overview and/or google_ai_mode (experimental, see below)
countryCode	string	us	Selects the Google domain (google.com, google.co.uk, google.de, ...) and the gl parameter
languageCode	string	en	Interface language (hl). Extraction is most reliable with en
maxQueries	integer	100	Safety cap on the number of queries processed
resolveUrls	boolean	true	Google serves result links as opaque redirect tokens. When on, the Actor follows each redirect (headers only) to get the real URL and domain
proxyGroup	auto | GOOGLE_SERP | RESIDENTIAL	auto	Which Apify Proxy group fetches the results page. auto prefers the Google SERP proxy
maxConcurrency	integer	3	Parallel result-page fetches (1–10)
debugSaveHtml	boolean	false	Store the raw HTML of every fetched page in the run's key-value store (HTML-0001.html, ...)

Example input:

{
  "queries": ["best project management software", "how to make sourdough bread"],
  "targetDomains": ["forbes.com", "kingarthurbaking.com"],
  "engines": ["google_ai_overview"],
  "countryCode": "us",
  "languageCode": "en"
}

Output example

{
  "query": "best project management software",
  "engine": "google_ai_overview",
  "countryCode": "us",
  "languageCode": "en",
  "hasAiAnswer": true,
  "answerText": "The top project management software platforms for 2026 depend on your team's size, workflow style, and budget:\nTop Software Picks\nClickUp\nBest for: All-in-one flexibility and custom workflows.\nKey features: 13+ task views (list, board, Gantt, timeline), built-in AI, and a generous free plan ...",
  "answerMarkdown": "The top project management software platforms for 2026 depend on your team's size, workflow style, and budget:\n\nTop Software Picks\n\n-   **ClickUp** [ClickUp](https://www.clickup.com/)\n    -   **Best for:** All-in-one flexibility and custom workflows.\n ...",
  "citations": [
    { "position": 1, "url": "https://www.clickup.com/", "domain": "clickup.com", "title": "ClickUp", "resolved": true },
    { "position": 2, "url": "https://www.forbes.com/advisor/business/software/best-project-management-software/", "domain": "forbes.com", "title": "10 Best Project Management Software of 2026 – Forbes Advisor", "resolved": true },
    { "position": 3, "url": "https://www.monday.com/", "domain": "monday.com", "title": "monday.com", "resolved": true },
    { "position": 4, "url": "https://www.reddit.com/r/projectmanagement/comments/1jwjw9s/what_project_management_tools_do_you_guys/", "domain": "reddit.com", "title": "What project management tools do you guys recommend? - Reddit", "resolved": true },
    { "position": 5, "url": "https://www.youtube.com/watch?v=v33m1aDCs1w&t=270", "domain": "youtube.com", "title": "Ranking the Best Project Management Tools of 2026", "resolved": true }
  ],
  "citedDomains": ["clickup.com", "forbes.com", "monday.com", "reddit.com", "youtube.com"],
  "targetDomainHits": [
    { "domain": "forbes.com", "cited": true, "positions": [2] },
    { "domain": "kingarthurbaking.com", "cited": false, "positions": [] }
  ],
  "organicTop10": [
    { "position": 1, "url": "https://www.icagile.com/resources/honest-review-of-6-personal-project-management-tools-with-kanban-view", "domain": "icagile.com", "title": "Honest Review of 6 Personal Project Management Tools ...", "resolved": true },
    { "position": 2, "url": "https://www.paymoapp.com/blog/project-management-software/", "domain": "paymoapp.com", "title": "20+ Best Project Management Software for 2026", "resolved": true }
  ],
  "extractionMethod": "heading:ai-overview",
  "serpUrl": "http://www.google.com/search?q=best+project+management+software&hl=en&gl=us&pws=0",
  "fetchedAt": "2026-10-15T04:40:12.345Z",
  "status": "ok"
}

A query where Google shows no AI Overview still returns status: "ok" with hasAiAnswer: false and the organic top 10, because "no AI answer for this keyword" is itself a monitoring result. A failed query looks like this and is not charged:

{
  "query": "best project management software",
  "engine": "google_ai_mode",
  "status": "failed",
  "error": "ai_mode_not_available",
  "hasAiAnswer": false,
  "citations": [],
  "organicTop10": [],
  "fetchedAt": "2026-10-15T04:40:12.345Z"
}

Pricing

Pay per event. You are charged only for records with status: "ok".

Event	Price	When
query-processed	$0.01	One query checked on one engine and the results page parsed (with or without an AI answer)

Monitoring 200 keywords once a day costs $1.00 per day. Failed queries (blocked page, network error, AI Mode unavailable) are free. You can cap spending per run with the Maximum total charge setting in Apify Console or the maxTotalChargeUsd API parameter; once the cap is reached, remaining queries are returned with status: "skipped".

Usage

Apify Console

Open the Actor, paste your keywords into Search queries, add your domains to Target domains, click Start, and download the dataset as JSON, CSV or Excel. Schedule the task daily or weekly to build a time series of AI citations.

JavaScript (apify-client)

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });

const run = await client.actor('fre.d/ai-search-citations').call({
  queries: ['best project management software', 'how to make sourdough bread'],
  targetDomains: ['forbes.com', 'kingarthurbaking.com'],
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const r of items) {
  if (r.status !== 'ok') { console.warn(r.query, r.error); continue; }
  console.log(r.query, r.hasAiAnswer ? r.citedDomains.join(', ') : '(no AI Overview)');
  for (const hit of r.targetDomainHits) if (hit.cited) console.log('  cited:', hit.domain, 'at', hit.positions);
}

Python (apify-client)

from apify_client import ApifyClient

client = ApifyClient(token=os.environ["APIFY_TOKEN"])

run = client.actor("fre.d/ai-search-citations").call(run_input={
    "queries": ["best project management software", "how to make sourdough bread"],
    "targetDomains": ["forbes.com", "kingarthurbaking.com"],
})

for r in client.dataset(run["defaultDatasetId"]).iterate_items():
    if r["status"] != "ok":
        print("failed:", r["query"], r["error"])
        continue
    print(r["query"], "->", r["citedDomains"] if r["hasAiAnswer"] else "(no AI Overview)")

AI agents via MCP

Add the Apify MCP server to your agent (Claude Desktop, Cursor, LangGraph, etc.) and allow this Actor:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com/?tools=fre.d/ai-search-citations",
      "headers": { "Authorization": "Bearer <APIFY_TOKEN>" }
    }
  }
}

The agent can then call the tool with {"queries": ["..."], "targetDomains": ["..."]} and read citedDomains, targetDomainHits and answerText directly from the result.

HTTP API

curl -X POST "https://api.apify.com/v2/acts/fre.d~ai-search-citations/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"queries": ["best project management software"], "targetDomains": ["forbes.com"]}'

How it works

1. Queries are de-duplicated and combined with the selected engines.
2. Each results page is fetched through Apify Proxy (Google SERP proxy by default, residential as fallback) with your country and language settings.
3. The AI Overview block is located by a chain of strategies: known container attributes, then the "AI Overview" heading and its enclosing block, then a text-structure fallback. The strategy that succeeded is reported in extractionMethod.
4. Google serves result links as opaque redirect tokens. With resolveUrls on, the Actor follows each redirect (reading only the Location header, never downloading the target page) to obtain the real URL and domain.
5. The answer is converted to plain text and Markdown, citations are de-duplicated and numbered, and target domains are matched against the cited domains (subdomains included).

Limitations and notes

• Not an official Google API. Results are scraped from the public results page and are a snapshot: AI Overviews vary by country, language, time of day and even between two requests. Treat single runs as samples and schedule repeated runs for trends.
• Partial answer text. Google serves only the collapsed part of the AI Overview in the HTML; text behind "Show more" is loaded later by the browser and is not included. Citations visible in the collapsed block are captured; sources that only appear after expanding may be missing.
• AI Mode is experimental. Google renders AI Mode (udm=50) in the browser. When the HTML does not contain an answer, the record is returned with status: "failed" and error: "ai_mode_not_available" and is not charged.
• Markup changes. Google changes its HTML frequently. If hasAiAnswer drops to false across all queries while Google clearly shows AI Overviews, open an issue; extractionMethod and debugSaveHtml exist to make the fix fast.
• No personal data. The Actor only processes public search result pages for the keywords you provide. It does not use cookies, Google accounts or logins, and it does not collect data about individuals. Keep your queries free of personal information.
• languageCode other than en works for fetching, but the AI Overview heading is recognised in a limited set of languages (en, ja, de, fr, es, it, pt, nl, zh, ko).

Support

Open an issue on the Actor's Issues tab in Apify Store. Feature requests (Bing Copilot, Perplexity, ChatGPT search, historical comparison between runs) are welcome.