Google Autocomplete Keyword Expander

Turns one seed keyword into hundreds of real Google Autocomplete suggestions, over plain HTTP — no browser, no login.

Not a scrape — Google's own public suggest endpoint

Reads suggestqueries.google.com/complete/search — the same JSON endpoint Firefox's address bar calls for search suggestions. Unauthenticated, no anti-bot layer encountered, clean JSON response.

What makes this different

One seed becomes 150-350+ unique keywords. Google's endpoint answers only one query at a time; there is no bulk parameter. This Actor implements the standard keyword-harvesting technique on top of it: append every letter a-z and digit 0-9 to your seed (alphabet mode), or every question word (questions mode), fire all the variant queries concurrently, and deduplicate the combined suggestions into one list — live-tested at 307 unique keywords from the single seed "python tutorial".

Provenance kept, not thrown away. Each keyword's found_via field names which variant query surfaced it, so you can see whether a keyword came from the ...b branch or a how ... question — useful for understanding search intent clusters, not just getting a flat list.

Locale-aware question words. questions mode uses the right question words for the interface language (what/why/how in English, apa/kenapa/bagaimana in Indonesian, que/como/donde in Spanish) rather than only supporting English.

Input

Field	Type	Notes
seed	string	The keyword to expand.
mode	enum	alphabet (36 variants), questions (~10 variants), both, or seed_only.
hl	string	Interface language, e.g. en, id, es. Also picks the question-word list.
gl	string	Country code — changes which suggestions surface.
maxItems	integer	Cap on results; 0 returns everything harvested.
proxyConfiguration	object	Not needed for normal volumes.

Known limits

• No true "total suggestion count." Google's suggest endpoint returns up to ~10 suggestions per query; the expansion strategy widens coverage but a seed with genuinely few related searches will still return a short list regardless of mode.
• Duplicate suggestions across variants are merged, not double-counted — found_via lists every variant that surfaced a given keyword.

Local development

pip install -r requirements.txt
python test_local.py "python tutorial" --mode alphabet --out sample_output.json
python test_local.py "belajar python" --hl id --gl id --mode both

sample_output.json in this folder is real output from a live run, kept so the schema can be reviewed without running anything.