Google AI Mode API: AI Mode Answers, Citations & Sources avatar

Google AI Mode API: AI Mode Answers, Citations & Sources

Pricing

from $8.00 / 1,000 google ai mode answers

Go to Apify Store
Google AI Mode API: AI Mode Answers, Citations & Sources

Google AI Mode API: AI Mode Answers, Citations & Sources

Google AI Mode scraper and MCP tool: send queries, get Google AI Mode's answer as JSON: markdown text and every cited source (URL, domain, title, snippet). No Google account, no browser. $0.008 per answer, plus $0.001 per run; empty or failed answers are free.

Pricing

from $8.00 / 1,000 google ai mode answers

Rating

0.0

(0)

Developer

PlainRows

PlainRows

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

7 days ago

Last modified

Features

Limited permissions

Share

Give it a list of queries and get one row per query: Google AI Mode's answer (Markdown) and every source it cites (rank, URL, domain, title, snippet, source name), plus the domains it links to.

A Google AI Mode scraper alternative: no browser, no proxies, no CAPTCHAs, no Google account. Queries run five at a time through the data source's priority queue; one answer took under 40 seconds in our test. You only pay for queries that got an answer: $0.008 per answer, plus $0.001 per run.

Where the answers come from: this Actor gets Google AI Mode's answers through a commercial data API that collects them from Google. Answers vary with the time and the location, so they can differ from what you see in your own browser.

Typical uses

  • AI search visibility (GEO / AEO): which sites does Google AI Mode cite for the queries your buyers type? Is yours one of them?
  • Citation research: sources and citedDomains show the pages Google AI Mode relies on, in order.
  • Content gaps: read what the answer says, and which competitors it cites, before writing your own page.
  • Answers with sources for your app or agent: Google AI Mode as plain JSON, with the citations kept apart from the text.
  • Weekly reports: schedule a run and send the rows to Google Sheets, Looker Studio, Slack, n8n, Make or Zapier.

What you get for each query

FieldDescription
queryThe query asked
answer, answerLengthGoogle AI Mode's answer as Markdown (headings, lists, tables, links)
sources, sourcesCountEach cited source: rank, url, domain, title, snippet, sourceName (up to 50, each URL once)
citedDomainsDomains cited, in order, without duplicates: sources first, then links in the answer (images excluded)
productsProducts shown as shopping cards in the answer (rank, name, price, url): the answer text itself omits their names
refinementChipsFollow-up refinements Google suggests, when the data source returns them (empty in our tests so far)
checkUrl, answeredAtA link that opens the query in Google AI Mode, and when it answered
found, status, messageANSWERED for a billed row; free rows say why: EMPTY_ANSWER, UNREADABLE, REJECTED, PROVIDER_ERROR, NOT_CHECKED, FREE_PLAN_LIMIT, FREE_PLAN_DAILY_QUOTA
country, languageCode, checkedAtLocation and language of the answer; when the run started

Missing values are null or empty lists, never guessed.

Input

{
"queries": ["best standing desk for small spaces", "Which CRM do small agencies recommend in 2026?"],
"country": "US",
"languageCode": "en"
}
  • queries: up to 200 queries per run, 1 to 500 characters each. Each query is one answer.
  • country (default US): US, GB, CA, AU, IE, IN, DE, AT, CH, FR, BE, ES, MX, IT, NL, PT, BR, PL, SE, JP.
  • languageCode (optional): two-letter code; default is the main language of the country.

Google search operators (site:, intitle:, inurl:, filetype: and the like) are not supported: those queries are skipped before any request, listed in the run summary, and never charged as answers (the $0.001 run start still applies). A word followed by a colon that looks like an operator (for example cache:) is treated the same way: rephrase it. Write them as plain questions instead.

Pricing

Pay per event, no subscription:

  • Google AI Mode answer: $0.008 per query that got a non-empty answer, at every Apify plan, free plan included.
  • Actor start: $0.001 per run.

Examples on a paid plan: 100 queries in one run cost $0.801 ($0.80 + $0.001) if all of them get an answer. Tracking 30 queries every day for 30 days (900 answers, 30 runs) costs $7.23.

Empty answers, queries the data source refused or failed, answers that came back in an unexpected format, and queries not reached (budget or timeout) get a free row with found: false and the reason. An answer can take up to 3 minutes: no query starts in the last 4 minutes before the run's timeout. The run never spends more than the maximum cost you set.

Free Apify plan: each run asks the first query only, so you can test the output, and only 5 runs per account and UTC day get answers (later runs that day ask nothing); queries not asked get a free row saying why. A paid Apify plan removes both limits.

Good to know

  • Answers vary from one run to the next, even for the same query: track trends over several runs, not a single answer.
  • Some sources point to Google itself (for example Google Shopping product pages): they are kept as cited, with the domain google.com.
  • An answer without sources is still an answer and is billed; sources is then empty.
  • When most answers come back empty, the run stops early to protect your budget (each query asked costs the data source money even when the answer is empty); the run summary says so. Rephrase the queries as questions Google answers in AI Mode.
  • For one table of brand mentions across ChatGPT, Gemini and Google AI Mode, use AI Visibility Tracker; for ChatGPT alone, ChatGPT Search Scraper; for Perplexity, Perplexity Search API.
  • Not affiliated with Google.

What this is not

  • Not Google's AI Overviews on the regular results page: this is the separate AI Mode answer.
  • Not a conversation: one query, one fresh answer; no follow-up questions.
  • Not a rank tracker for Google's blue links: for organic results use a SERP tool.

Run summary

Each run stores a summary in the key-value store (OUTPUT): queries asked, answers billed, empty answers (queriesEmptyAnswer), unreadable answers (queriesUnreadable), queries refused (operators, too long) and anything skipped because of the free-plan or budget limit.

Use it from code, automations or AI agents

API (one call, results returned directly):

curl -X POST "https://api.apify.com/v2/acts/plainrows~google-ai-mode-api/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
-H "Content-Type: application/json" \
-d '{"queries": ["best standing desk for small spaces"]}'

n8n, Make, Zapier: use the Apify integration, pick plainrows/google-ai-mode-api, schedule it and send the rows to Google Sheets or Slack.

AI agents (MCP tool): add https://mcp.apify.com?tools=plainrows/google-ai-mode-api to Claude, Cursor or any MCP client, then ask "What does Google AI Mode answer to these queries, and which sites does it cite?".

FAQ

Is this a Google AI Mode API? Yes: one call per list of queries, JSON rows back with the answer and its sources, usable from the Apify API, n8n, Make, Zapier or as an MCP tool.

How do I see if Google AI Mode cites my site? Look for your domain in citedDomains (and its position in sources.rank).

Do I pay for empty answers or errors? No. Only queries with a non-empty answer are charged ($0.008), plus $0.001 per run.

More PlainRows tools

Same pay-per-result pricing, same clean JSON, built to chain together in n8n, Make, Zapier or AI agents.

Support

Missing a field or found a bug? Open an issue on this page. Issues are read and answered.

⌂ 换网址