Google Autocomplete Keyword Scraper

What does Google Autocomplete Keyword Scraper do?

Google Autocomplete Keyword Scraper is a keyword research tool that turns a few seed keywords into hundreds of long-tail keyword ideas straight from Google autocomplete, the suggestions people see as they type in the search box. It types each seed the way keyword tools do (with a–z, 0–9, question words and prepositions), can dig recursively into the keywords it finds, and can also ask YouTube, Google Shopping, Amazon and Bing for their suggestions, for any country and language.

Try it with the prefilled seed: open the Actor, click Start, and about 50 keyword ideas for "coffee maker" are ready in a few seconds. Because it runs on the Apify platform, you also get API access, scheduled runs, integrations (Google Sheets, Make, Zapier, Slack, webhooks) and run monitoring.

Why use Google autocomplete for keyword research?

• Real searches, not guesses. Autocomplete shows what people actually type into Google, so every idea is a query someone searches for.
• Long-tail keyword ideas at scale. One seed with a–z and question words gives about 400–500 unique ideas on Google; recursion finds thousands.
• Content and FAQ planning. Question expansions (how, what, why, can, is…) surface the questions your audience asks, ready for blog posts, FAQs and "People also ask"-style sections.
• Platform-specific ideas. YouTube suggestions for video topics, Google Shopping and Amazon suggestions for product and e-commerce keywords, Bing for a second search engine's view.
• Local SEO and international markets. Pick the country and language to see what people search in Germany, India or Brazil, in their own language.
• Cheap and fast. Plain HTTP requests, no browser: thousands of ideas per minute.

How to find keyword ideas with Google autocomplete

1. Type one or more seed keywords into the Input tab, one per line, such as coffee maker or running shoes for women.
2. Pick the suggestion sources (Google web search is on by default).
3. Choose the expansions: a–z and question words are on by default; add 0–9 and prepositions for more ideas.
4. Optionally set recursive depth to 1 or 2 to also expand the keywords found.
5. Set the country and language, then click Start. Download the keyword ideas when the run finishes.

You can also paste a search results link, such as https://www.google.com/search?q=cold+brew, a YouTube, Amazon or Bing search link, with or without https://: its search words become the seed.

Input

Everything is set on the Input tab. In JSON, the input looks like this:

{
    "keywords": ["coffee maker", "cold brew"],
    "sources": ["google", "youtube", "amazon"],
    "addAlphabet": true,
    "addNumbers": false,
    "addQuestions": true,
    "addPrepositions": false,
    "recursiveDepth": 0,
    "maxKeywordsPerSeed": 500,
    "country": "us",
    "language": "en"
}

Field	What it does
keywords	Seed keywords, one per line. Search results links work too: their search words are used. API callers can also send {"keyword": "..."} or {"url": "..."} objects.
sources	Where suggestions come from: google (web search, the default), youtube, google-shopping, amazon and bing. Each source gives its own rows.
addAlphabet	Types the seed followed by each letter, coffee maker a … coffee maker z (26 lookups). Default: on.
addNumbers	Types the seed followed by each digit, coffee maker 0 … coffee maker 9 (10 lookups). Default: off.
addQuestions	Types how, what, why, when, where, which, who, can, is, are, does and will before the seed (12 lookups). Default: on.
addPrepositions	Types the seed followed by for, with, without, near, to, vs, versus, like, or and (10 lookups). Default: off.
recursiveDepth	0 types the seed and its expansions only. 1 also types every new keyword found; 2 does that once more. Default: 0.
maxKeywordsPerSeed	Stops after this many unique keywords for each seed on each source. Default: 500.
country	Two-letter country code such as us, gb, de or in (uk works too). For Amazon it picks the store. Default: us.
language	Language code such as en, de, ar or pt-BR. Default: en.
proxyConfiguration	Leave empty. The services answer the Actor directly; if one refuses, it retries through Apify's datacenter and then residential proxies on its own.

Output

Each unique keyword idea becomes one row. Here is a real example from a run with seed "air fryer":

{
    "keyword": "air fryer asparagus",
    "source": "google",
    "position": 1,
    "seed": "air fryer",
    "query": "air fryer a",
    "expansionType": "alphabet",
    "expansion": "a",
    "depth": 0,
    "relevance": 1250,
    "suggestionType": "QUERY",
    "wordCount": 3,
    "country": "US",
    "language": "en",
    "scrapedAt": "2026-10-20T14:19:30+00:00"
}

And one from Amazon's German store, where relevance is not published:

{
    "keyword": "makita kaffeemaschine akku 18v",
    "source": "amazon",
    "position": 1,
    "seed": "kaffeemaschine",
    "query": "kaffeemaschine a",
    "expansionType": "alphabet",
    "expansion": "a",
    "depth": 0,
    "relevance": null,
    "suggestionType": null,
    "wordCount": 4,
    "country": "DE",
    "language": "de",
    "scrapedAt": "2026-10-20T14:19:23+00:00"
}

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

Data fields

Field	Description
keyword	The suggested keyword, as the service wrote it (Google and YouTube write in lower case).
source	google, youtube, google-shopping, amazon or bing.
position	Where the suggestion appeared in the list for its query, starting at 1.
seed	The seed keyword it was found for.
query	The exact text typed to get it, such as coffee maker a or how coffee maker.
expansionType	seed, alphabet, number, question, preposition or recursive.
expansion	The letter, digit or word added; for recursive, the keyword it grew from; null for the seed itself.
depth	0 for the seed and its expansions, 1 or 2 for recursive levels.
relevance	Google's own relevance score for the suggestion (higher is stronger, typically 550–1300). Google, YouTube and Google Shopping only; null for Amazon and Bing.
suggestionType	Google's suggestion type, almost always QUERY. null for Amazon and Bing.
wordCount	Number of words in the keyword, handy for filtering long-tail ideas.
country	The country the suggestions are for, such as US.
language	The language asked for, such as en.
scrapedAt	When the row was saved, in UTC.

Deduplication: each keyword is saved once per source across the whole run, including across seeds and expansions, ignoring case. Website suggestions (links such as https://www.facebook.com/) are left out, as they are not keywords.

How much does it cost to find keyword ideas?

The Actor uses pay-per-event pricing: you pay per keyword idea saved, and nothing for lookups that return nothing new. See the Pricing tab for the exact price.

A default run (one seed, Google, a–z and question words) saves up to 500 keyword ideas in about 5–10 seconds. Seeds that return nothing are never charged, and if you set a maximum cost for a run, the Actor stops as soon as it is reached.

Tips for keyword research with autocomplete

• Start broad, then go deep. Use short seeds (coffee maker) with a–z and questions first; turn on recursion when you need thousands of ideas.
• Questions for content, prepositions for intent. Question words find FAQ topics; for, with, vs and near find use cases, comparisons and local intent.
• Numbers for products. addNumbers finds model numbers, sizes and years, best combined with Google Shopping or Amazon.
• Use maxKeywordsPerSeed to control cost. Recursion multiplies lookups quickly; the limit stops each seed as soon as it is reached.
• Match language to the seed. Write seeds in the language you set: kaffeemaschine with de/de, not coffee maker.
• Schedule runs to watch how suggestions change, for example around seasonal events.

FAQ, disclaimers and support

Does it give search volume? No. Autocomplete shows which queries people type, ordered by Google's relevance, but Google does not publish search volumes there. position and relevance tell you which suggestions Google ranks highest.

Why are Google Images and Google News not sources? Their suggestion boxes return the same suggestions as Google web search (checked October 2026), so they would only add duplicate rows. YouTube and Google Shopping do return their own suggestions, and are included.

Which Amazon stores are supported? us, ca, mx, br, gb, de, fr, it, es, nl, se, pl, tr, be, eg, ae, sa, in, jp, au and sg. For other countries Amazon is skipped and noted in the FAILED_INPUTS record, while the other sources still run.

Why do some suggestions look odd? They are what the services suggest. Very rare or made-up seeds can get unrelated suggestions from Google, and Amazon sometimes suggests products that do not start with your seed.

What happens to seeds that return nothing? They are not saved as rows, so you are not charged for them. Each one is listed with its reason (no suggestions, every suggestion already saved for an earlier seed, unusable entry, or refused by the service) in the FAILED_INPUTS record of the run's key-value store, and the run's status message gives the counts. A run fails only when every seed was refused by the services.

Is it legal to scrape autocomplete suggestions? The Actor reads the public suggestion lists that Google, YouTube, Amazon and Bing show to anyone typing in their search boxes. Suggestions are aggregated search queries, not personal data. You are still responsible for making sure your use complies with each service's terms and with the laws that apply to you. If in doubt, ask a lawyer.

Found a problem or need a custom solution? Open an issue on the Issues tab and we will look into it. Custom keyword research tools and data feeds are available on request.