Google Autocomplete Scraper — Keyword Ideas, YouTube & Bing

Turn a few seed keywords into hundreds of long-tail keyword ideas from autocomplete: what people start typing in Google Search, YouTube, Google Shopping, Bing, DuckDuckGo and eBay. The actor queries each engine's suggest box for the seed, the seed + a–z, the seed + 0–9, question words (how, what, why, can, is, best ...) and modifiers (vs, near me, for, with ...), optionally goes several levels deep, and saves one row per suggestion with its engine, position, the query that produced it, country and language. Use it for SEO content plans, PPC keyword and negative-keyword lists, YouTube video ideas, product naming and marketplace listings.

It uses the engines' public suggest endpoints: no browser, no login, no API key. Runs are fast (about 150 queries and 2,000 suggestions in 40 seconds) and you pay only for the suggestions you get.

What data does it extract?

Field	Description
keyword	The autocomplete suggestion, e.g. air fryer chicken breast
source	Engine: google, youtube, googleShopping, bing, duckduckgo or ebay
seed	Your seed keyword that led to this suggestion
position	Position of the suggestion in the engine's list for that query (1 = top)
query	The exact text typed into the engine, e.g. air fryer c
expansion	How the query was built: seed, letter, number, question, modifier or recursive
modifier	The letter, digit or word added (c, 2, how, vs), else null
modifierPosition	after or before the seed
depth	1 for suggestions of the seed and its expansions, 2+ for recursive levels
country, language	Market and interface language used for the query
relevance	Google's own relevance score (Google, YouTube, Shopping), e.g. 1250, 601
suggestionType, suggestionSubtypes	Google's suggestion type (QUERY, ...) and internal subtype codes
ebayCategories	eBay category matches for the suggestion, when eBay shows them ([{id, name}])
wordCount	Number of words, handy for filtering long-tail keywords
suggestionId	`source
scrapedAt	UTC timestamp

With monitoring on, rows also carry changeType (new).

How to scrape Google autocomplete suggestions

1. Enter one or more Seed keywords, e.g. air fryer, learn python.
2. Pick the Engines (Google Search, YouTube, Bing, Google Shopping, DuckDuckGo, eBay).
3. Choose expansions: A-Z (on by default), 0-9, Questions & modifiers, and the Recursive depth.
4. Choose one or more Countries (and optionally a language), set Max suggestions, and click Start.

Each suggestion is saved once per engine and country, even if several queries or seeds return it.

Input example

{
  "keywords": ["air fryer"],
  "sources": ["google", "youtube", "bing"],
  "expandLetters": true,
  "expandQuestions": true,
  "depth": 1,
  "countries": ["us"],
  "maxItems": 1000
}

Output example

{
  "keyword": "air fryer chicken breast",
  "seed": "air fryer",
  "source": "youtube",
  "position": 3,
  "query": "air fryer c",
  "expansion": "letter",
  "modifier": "c",
  "modifierPosition": "after",
  "depth": 1,
  "country": "us",
  "language": "en",
  "relevance": 1050,
  "suggestionType": "QUERY",
  "suggestionSubtypes": [512, 433],
  "ebayCategories": null,
  "wordCount": 4,
  "suggestionId": "youtube|us|air fryer chicken breast",
  "scrapedAt": "2026-10-11T22:33:00Z"
}

How many keyword ideas do I get per seed?

Engines return 7-25 suggestions per query (Google, YouTube and Shopping 15, Bing up to 25, eBay 10, DuckDuckGo 7-8). Per seed and engine, roughly:

Expansion	Queries	Suggestions (after removing duplicates)
Seed only	1	7-25
A-Z	27	200-550
A-Z + 0-9	37	280-700
Questions & modifiers	+24	+150-350 on Google/YouTube, +10-50 on Bing, DuckDuckGo, eBay
Recursive depth 2	+1 per suggestion found	about 10x more

A real run with air fryer, A-Z and all six engines returned 2,000 suggestions (the Max suggestions cap) from 155 queries: Bing 544, YouTube 389, Google Shopping 378, Google 350, eBay 209, DuckDuckGo 130.

How much does it cost?

Pay per result: $0.0007 per suggestion ($0.70 per 1,000), plus Apify platform usage of about $0.006 per 1,000 suggestions (datacenter proxy, compute and storage). No start fee, no subscription. Examples: one seed on Google with A-Z (~350 suggestions) costs about $0.25; one seed on all six engines with A-Z (~2,000) about $1.41. Duplicates within a run are removed before saving, so they are never charged. Set Max suggestions to cap the cost.

How to get YouTube, Bing, eBay and Google Shopping autocomplete keywords

Select the engines in Engines. YouTube suggestions show what people search for on YouTube (great for video titles and tags), Google Shopping and eBay show buyer and product phrases, Bing and DuckDuckGo give a second opinion on web searches. Every row says which engine it came from, so you can compare them or filter in a spreadsheet.

Does it work for other countries and languages?

Yes. Pick one or more Countries: the actor sets Google's country (gl) and language (hl), Bing's market (de-DE), DuckDuckGo's region (uk-en) and eBay's site (ebay.de, ebay.co.uk ...). Leave Language empty to use each country's main language, or set it (e.g. fr for French in Switzerland). Question words are localized for German, French, Spanish, Italian, Portuguese and Dutch. eBay has no site for Brazil, India, Japan, Mexico, New Zealand or Sweden, so eBay is skipped there.

How does recursive expansion work?

With Recursive depth 2, every suggestion found at depth 1 is typed into the same engine again and its own suggestions are saved with depth: 2 (3 and 4 go further). Each keyword is queried only once per engine and country, and every suggestion is saved only once. Each level multiplies the number of queries by about ten, so use Max suggestions to cap the run.

Does it need a login, an API key or a browser?

No. It uses the same public suggest endpoints the search boxes use, through Apify's datacenter proxy with a fresh IP for every retry. If the datacenter IPs keep getting blocked, it switches to Apify's residential proxy for the rest of the run (if your plan includes it).

Monitor changes on a schedule

Turn on Only new or changed items and save the input as a task with a Schedule (e.g. daily): each run outputs only suggestions (engine + country + keyword) that earlier runs of this watch did not return, with changeType: "new", so you see new trending searches for your seeds or brand as they appear. Suggestions seen before are not output and not charged. Every run sends all queries of your settings; Max suggestions caps the new rows per run. Use Monitoring key to keep several schedules on the same keywords separate. Engines rotate some suggestions and localize a few by the IP's location, so a handful of "new" rows per run is normal.

Why do results differ between runs?

Autocomplete is live: engines update suggestions all the time, personalize some by location (e.g. "coffee shops nyc") and test variants. Two runs a few minutes apart usually agree on most suggestions but not all.

Is it legal to scrape autocomplete suggestions?

The actor collects the public search suggestions every visitor sees in the search box; it does not log in or access anything private. Suggestions sometimes contain names of public figures, brands or places, exactly as the engines show them. You are responsible for how you use the data and for complying with the engines' terms and applicable law.

Integrations and API

Use the results in Google Sheets, Zapier, Make, n8n, or call the actor from Python/Node with the Apify client. See the API tab for ready-made snippets.

Support

Found a bug or need another engine? Open an issue in the Issues tab; it is usually answered within a day.