Keyword Ideas & Autocomplete Scraper (Google, YouTube, Amazon, Bing)

Enter a few seed keywords. The Actor collects what people actually type into the search boxes of Google, YouTube, Amazon and Bing. It returns one row per unique suggestion, merged across all four sources.

• Expansion modes: basic (seed only), alphabet (seed + a–z and 0–9), questions (how, what, why, when, where, which, who, can, does, is, are, should, will, best, cheap + seed vs / for / with / without / near me / or / and / like / to)
• Depth 2: every found suggestion is sent back once more to the source it came from, for deeper long-tail ideas
• Cross-source dedupe: sources lists every source that suggested the phrase, and sourcesCount counts them. A phrase that appears in 3–4 sources is a strong signal.
• Rule-based features (no AI, fully deterministic): wordCount, isQuestion, isCommercial, isLocal, containsSeed and an intent label (local > commercial > question > informational)

Field	Meaning
suggestion	Suggested search phrase (lowercased, whitespace normalized)
source / modifier / position	Where it was first found: the source, the modifier used (a, how, depth2:<parent>...) and its 1-based rank in that list
sources / sourcesCount	All sources that suggested this phrase for this seed
intent	local (near me, nearby, "in
depth	1 = from the seed, 2 = from expanding a suggestion
error	Set only on rows for a source that failed completely for that seed. These rows are not charged

Who it is for

• SEO and content teams building keyword lists and FAQ sections
• YouTube creators looking for video titles people already search for
• Amazon sellers and print-on-demand sellers looking for product and listing keywords
• PPC marketers collecting commercial and local modifiers

Input example

{"seeds": ["yoga mat", "standing desk"], "sources": ["google", "youtube", "amazon", "bing"],
 "expansionMode": "alphabet", "depth": 1, "maxSuggestionsPerSeed": 1000,
 "country": "us", "language": "en", "amazonMarketplace": "com"}

Output example

{"seed": "coffee grinder", "suggestion": "best coffee grinder for espresso", "source": "google",
 "sources": ["google", "youtube", "bing", "amazon"], "sourcesCount": 4, "modifier": "best", "position": 2,
 "depth": 1, "country": "us", "language": "en", "amazonMarketplace": "com", "wordCount": 5,
 "isQuestion": false, "isCommercial": true, "isLocal": false, "containsSeed": true,
 "intent": "commercial", "error": null}

Speed and volume (measured)

• 1 seed, questions mode, 4 sources: about 490 unique suggestions in 7 seconds
• 2 seeds, alphabet mode, 4 sources: about 2,000 unique suggestions in 21 seconds

Limits

• Each autocomplete list has at most about 10 entries (Bing up to 12), so volume comes from expansion modes and depth.
• country / language are passed to Google, YouTube and Bing, but Google may still mix in results that reflect the server's own location.
• Question modifiers are English words. For other languages, basic or alphabet usually work better.
• Amazon has no country parameter. Choose the marketplace with amazonMarketplace (19 marketplaces).
• No search volume or CPC numbers: these are real autocomplete phrases, not estimates.
• Only public autocomplete endpoints are used. No login and no API keys. No proxy is needed for normal batches. Requests are throttled (default 5 in parallel) and retried on HTTP 429 and 5xx errors.

Not affiliated with Google, YouTube, Amazon or Microsoft. By HAPPYSPROUTSTORE LLC.