Google AI Overview API — AIO Scraper & GEO Tracker

A Google AI Overview API that scrapes the generative answers Google now shows at the top of search results — for any query, from any country, at scale. This Actor fetches a live Google SERP, detects whether an AI Overview (AIO) rendered, and extracts both the full AI Overview answer text and the list of domains Google cited to generate it.

Use it as a drop-in AI Overview scraper for Generative Engine Optimization (GEO) tools, SEO agencies, brand monitors, and in-house marketing teams that need to answer:

• Is my brand mentioned in the Google AI Overview for these queries?
• Which domains does Google cite as authorities on my topic?
• How does the AIO answer change week over week?
• Which queries don't trigger an AI Overview at all?

Why use this Google AI Overview API

Every SERP API on the market treats the Google AI Overview as an afterthought — a bolt-on field on their generic Google endpoint. None of them give you the clean, AIO-native data shape this Google AI Overview scraper does:

	SerpApi	Serper	Scrapfly	This Actor
AIO field	bolt-on	bolt-on	bolt-on	first-class
Clean AIO text (no boilerplate)	no	no	no	yes
Deduplicated cited domains	no	no	no	yes
Country-accurate residential IPs	yes	yes	yes	yes
Distinguishes "not available for this search"	no	no	no	yes

Input

{
  "queries": [
    "what is retrieval augmented generation",
    "benefits of intermittent fasting",
    "best project management software"
  ],
  "country": "US",
  "language": "en",
  "device": "desktop"
}

• queries (required, array of strings) — up to 30 queries per run. This is an intentional ceiling (see Limitations).
• country (ISO-2, default US) — drives both Google's gl parameter and the proxy exit geo. Use matching language for best results.
• language (default en) — Google hl parameter.
• device (desktop or mobile, default desktop) — mobile SERPs ship different AIO content than desktop.

Output (one row per query)

{
  "query": "what is retrieval augmented generation",
  "country": "US",
  "language": "en",
  "device": "desktop",
  "appeared": true,
  "text": "Retrieval-Augmented Generation (RAG) is an AI framework that improves Large Language Model (LLM) accuracy by retrieving data from external, trusted knowledge bases...",
  "citations": [
    {"position": 1, "domain": "aws.amazon.com", "url": "https://aws.amazon.com/what-is/retrieval-augmented-generation/", "title": ""},
    {"position": 2, "domain": "blogs.nvidia.com", "url": "https://blogs.nvidia.com/blog/what-is-retrieval-augmented-generation/", "title": ""},
    {"position": 3, "domain": "research.ibm.com", "url": "https://research.ibm.com/blog/retrieval-augmented-generation-RAG", "title": ""}
  ],
  "status": "ok",
  "final_url": "https://www.google.com/search?q=what+is+retrieval+augmented+generation&hl=en&gl=us&pws=0",
  "raw_html_bytes": 966644,
  "fetch_seconds": 8.6,
  "fetched_at": "2026-04-13T23:47:01Z",
  "notes": []
}

Fields

Field	Type	Meaning
appeared	bool	Whether an AI Overview rendered for this query. false if Google explicitly said "an AI Overview is not available for this search" or if the query didn't trigger one.
text	string	The cleaned AI Overview answer text, with boilerplate and fallback phrases stripped.
citations[]	array	Deduplicated list of external domains Google cited, in order of first appearance in the panel. Google assets (gstatic, googleapis, etc.) are excluded.
status	string	ok | captcha | sorry | empty | error.
raw_html_bytes	int	Size of the raw SERP we scraped (for diagnostics).
fetch_seconds	float	Wall-clock time for the individual scrape.

Common use cases for the Google AI Overview API

• Brand presence monitoring — run 500 branded queries daily against the Google AI Overview API and alert when your domain disappears from or enters an AIO.
• GEO competitor research — which domains does Google cite for your target queries? Build a Generative Engine Optimization citation graph across a keyword set.
• Content planning — which topics trigger an AI Overview vs. don't? AIO-triggered queries need a different SEO strategy than organic-only queries.
• LLM discoverability benchmarking — snapshot AI Overview answers weekly and diff them to track how Google's generative summary of your industry evolves.
• AI search training data — collect real generative answers + source citations for RAG research.

Tips

• 30 queries per run is the cap. Schedule the Actor to run hourly against rolling batches for tracking at scale — that's up to 720 queries/day per schedule, which covers virtually every GEO monitoring workload.
• Country matters more than you'd think. The same query from US vs GB vs DE returns different AIO content. If you're tracking regional brand presence, run per-country.
• Queries that reliably trigger AIO: "what is", "how to", "benefits of", "best X for Y", informational and comparison queries. Transactional queries (buy / price / near me) rarely trigger AIO.
• Mobile ≠ desktop. Mobile AIO is shorter and often has different citation order.

Limitations

• Why 30 queries per run (not 100). Google's AI Overview endpoint applies a behavioral rate limit that kicks in after ~40 sequential SERP fetches from the same session, regardless of IP rotation or proxy quality. Every AIO scraper on the market hits this wall — most hide it by silently returning partial data. We cap at 30 and deliver 100% AIO capture under the cap so your dataset is clean. For larger keyword sets, use scheduled runs.
• AIO is a moving target. Google changes the rendered DOM regularly; we monitor parse accuracy continuously and patch within the hour of any drift detected.
• No historical data ships with the Actor itself. Every run is a fresh fetch. For historical tracking + diffs + alerting, see our upcoming hosted API at aio-tracker.com.

Pricing

Pay-per-result. You're charged only for successful AI Overview fetches (appeared: true). Queries that don't trigger an AIO, queries that hit anti-bot pages, and errors are free.

Support

Issues, feature requests, and custom needs: file them on the Actor's issues tab. Wholesale pricing available for teams running 100k+ queries/month — DM for rates.