Google Autocomplete Scraper — keyword suggestions from Google, YouTube, Amazon and Bing

Google Autocomplete Scraper extracts keyword suggestions from google.com's autocomplete — and, in the same run, from YouTube, Amazon and Bing — for any seed keyword, in any language and country. It sends the queries the search boxes send, parses the JSON they answer with, and saves one clean row per suggestion: engine, seed keyword, query variant, suggestion text, position, Google's relevance score and type, and a ready-to-open search URL. No API key, no login, no browser.

Type espresso machine and Google answers with "espresso machine with grinder", "espresso machine near me", "espresso machine repair near me"… That is what people are actually typing. Turn on the a–z expansion and the Actor asks for espresso machine a … espresso machine z; add questions (how espresso machine, why …), prepositions (espresso machine for, … without), comparisons (… vs, … alternative) or your own modifiers, and a single seed grows into a few hundred long-tail keywords. Drill-down sends every suggestion back as a new seed for one, two or three more levels.

What you get

Field	Description
engine	google, youtube, amazon or bing
keyword	The seed keyword you entered
query	The exact string sent to the engine (seed, or seed + modifier)
modifier, expansion	Which modifier produced the query (a, how, vs, your custom word) and which set it belongs to (none, alphabet, numbers, questions, prepositions, comparisons, custom, drill-down)
level	0 for your queries, 1–3 for drill-down levels
suggestion	The suggestion text
position	1-based rank in the engine's answer
relevance	Google's relevance score (higher = shown first); null on the other engines
suggestionType	Google: QUERY or NAVIGATION (a site, e.g. https://www.youtube.com/); Amazon: KEYWORD
subtypes	Google/YouTube numeric subtype codes as returned (e.g. [512, 433]), null elsewhere
navigationTitle	Site name of a Google NAVIGATION row
wordCount	Words in the suggestion
market, language, country, department	Locale used (en-US, amazon.com.mx…) and the Amazon department alias
searchUrl	Search page for the suggestion on that engine
scrapedAt	ISO timestamp

A SUMMARY record in the key-value store lists, per engine, how many queries were answered, how many suggestions were listed, saved, dropped as exact matches or skipped as duplicates, plus the proxy rung and browser header profile used.

Use cases

• Long-tail keyword research: seed + a–z + questions on Google gives the phrases people complete, not the ones a tool guesses.
• Content and FAQ planning: the questions expansion returns "how", "why", "can" phrasings for articles, People-Also-Ask sections and chatbot intents.
• YouTube titles and tags: YouTube's own suggestions for a topic, by country.
• Amazon listings and PPC: what shoppers type in the search box of twelve marketplaces, optionally inside one department (electronics, stripbooks…).
• Brand and reputation monitoring: watch which completions appear next to a brand (brand vs, brand alternative, is brand …).
• Datasets for AI: thousands of real query completions per topic for classifiers, intent models or RAG evaluation.

How to use

1. Enter one or more seed keywords.
2. Pick the engines (default Google).
3. Optionally choose expansions, add custom modifiers (a * marks where the seed goes: mejor * 2026), set drill-down levels, the language/country and, for Amazon, the marketplace and department.
4. Set max suggestions. The run stops exactly there.

{
  "keywords": ["espresso machine", "milk frother"],
  "engines": ["google", "youtube", "amazon"],
  "expansions": ["alphabet", "questions"],
  "customModifiers": ["best * 2026", "under 500"],
  "language": "en",
  "country": "US",
  "amazonMarketplace": "com",
  "amazonDepartment": "aps",
  "maxItems": 1000
}

Example row:

{
  "engine": "google",
  "keyword": "espresso machine",
  "query": "espresso machine w",
  "modifier": "w",
  "expansion": "alphabet",
  "level": 0,
  "suggestion": "espresso machine with grinder",
  "position": 1,
  "relevance": 601,
  "suggestionType": "QUERY",
  "subtypes": [512, 433],
  "navigationTitle": null,
  "wordCount": 4,
  "market": "en-US",
  "language": "en",
  "country": "US",
  "department": null,
  "searchUrl": "https://www.google.com/search?q=espresso%20machine%20with%20grinder&hl=en&gl=us",
  "scrapedAt": "2026-10-20T03:12:44.118Z"
}

How many suggestions to expect

Per query, Google returns up to 15 suggestions (with relevance and type), YouTube about 14, Amazon 10 and Bing up to about 23. With Unique suggestions only (default) each suggestion is saved once per engine, so a seed with the a–z expansion typically yields 150–300 unique Google keywords; the engine's echo of the query itself is dropped unless you turn on Keep exact matches. Max suggestions per keyword and engine keeps big seed lists balanced.

Pricing

Pay per result: one keyword-suggestion event per suggestion saved, nothing else. See the Pricing tab for current prices; paid Apify plans get lower rates. Your plan's monthly limit and the run's max charge are both respected.

Compared with alternatives

• Google Ads Keyword Planner / paid SEO suites give search volumes and CPC, which autocomplete does not expose; this Actor gives you the actual completions Google, YouTube, Amazon and Bing show, in bulk and per country. Use both: generate here, size there.
• Single-engine autocomplete actors cover Google or YouTube or Amazon; here one input and one dataset cover the four, with the same expansion and drill-down logic and Google's relevance score kept.
• Typing by hand shows 10 suggestions at a time and never the a–z or question variants.

For the next steps of a keyword workflow, YouTube Search Scraper ranks the videos competing for a keyword and Google Trends Scraper shows its interest over time.

Integrations

Run it from the Apify Console, on a schedule, or through the API (Node.js, Python, cURL) and export the dataset as JSON, CSV, Excel or XML. Connect it to n8n, Make, Zapier, Google Sheets or a webhook with the Apify integrations, and expose it to AI agents and LLM tools through Apify's MCP server.

FAQ

Is it legal? The Actor reads the public suggestion endpoints the search boxes use, without logging in; suggestions are aggregated query text, not personal data. You are responsible for how you use the output and for complying with the terms of each engine.

Do I need a proxy? Apify datacenter proxy is on by default and should be enough. If an engine answers with a block (HTTP 403/429 or a captcha page) the Actor rotates browser header profiles and, after two blocks in a row, switches itself to RESIDENTIAL proxy in the chosen country for the rest of the run. Local or custom-proxy runs work too.

Are there limits? Each query returns at most 10–23 suggestions, which is why the expansions exist. Google and YouTube stop suggesting for very long or nonsense queries; the summary tells you how many queries came back empty. The run is capped by maxItems and your max charge.

Does it give search volumes? No — autocomplete engines do not publish them. Google's relevance is the only ranking signal available and it is kept.

Which languages? Any language code accepted by Google/YouTube/Bing (es, pt-BR, de, ja…) and any two-letter country. The built-in question/preposition words are English; use Custom modifiers for other languages (cómo *, * barata).

Why fewer rows than queries × 15? Unique suggestions only is on: the same keyword returned by seed a and seed b is saved once. Turn it off to keep one row per query variant.

Changelog

• 0.1 (October 2026) — Google (relevance, type, subtypes), YouTube, Amazon (12 marketplaces, department alias), Bing; a–z, 0–9, questions, prepositions, comparisons, custom modifiers with *; drill-down up to 3 levels; unique/exact-match filters; per-keyword cap; proxy and header-profile ladder; SUMMARY record.

---

En español

Google Autocomplete Scraper extrae las sugerencias de autocompletado de Google, YouTube, Amazon y Bing para cualquier palabra clave semilla, en cualquier idioma y país, sin API key. Cada fila trae el motor, la semilla, la consulta enviada, la sugerencia, su posición, la relevancia y el tipo (Google) y la URL de búsqueda. Las expansiones a–z, 0–9, preguntas, preposiciones y comparaciones multiplican las consultas (cafetera espresso a … z); con modificadores propios puedes usar palabras en español (mejor * 2026, * barata, cómo *) y con los niveles de profundización cada sugerencia vuelve a consultarse como nueva semilla. Amazon admite doce tiendas (amazon.com, .es, .com.mx, .com.br, .de, .co.uk…) y un departamento. Consulta la pestaña Pricing para ver los precios vigentes; los planes de pago de Apify tienen tarifas más bajas. El proxy de datacenter de Apify basta y, si un motor bloquea, el Actor pasa solo a proxy residencial.