Blibli Search Keyword Autocomplete & Suggester

Harvest the keyword suggestions Blibli's search box shows shoppers — expanded far beyond the handful you can see — each with a measured count of competing products so you know which terms you can actually rank for.

For merchants optimising product titles and marketplace SEO.

No login, no API key. HTTP-only — no browser.

⚠️ About "search volume"

Blibli does not publish search volume, and this actor does not invent one.

Measured: across 65 autocomplete suggestions from 10 prefixes, every volume-shaped field (productCount, count) was null and dsScore was 0.0 on every row. No endpoint in the search family exposes a volume, impression count or frequency.

What you get instead — both real, both measured:

• competingProducts — how many products Blibli actually returns for the keyword. This is the number that decides whether you can rank: laptop returns 58,744 competing products, laptop gaming 6,639, laptop advan 644.
• suggestPosition — rank inside Blibli's own autocomplete ordering, which comes from their relevance model. An ordinal popularity signal, not a volume.

If you need true search volume, no public Blibli surface provides it.

What it does that the search box doesn't

Typing a seed into Blibli shows you ~6 suggestions. This actor appends every letter a–z to the seed and collects each different slice of the suggestion index:

Query	Keywords found
laptop typed into the search box	6
this actor, expansionMode: alphabet	147

31 of those 147 don't even start with "laptop" — asus laptop, backpack laptop, tote bag laptop, tas laptop jinjing. Those are the terms you'd never have thought to check.

Output

{
  "keyword": "laptop hp official store",
  "seedKeyword": "laptop",
  "source": "autocomplete",
  "sourcePrefix": null,
  "suggestPosition": 5,
  "competingProducts": 335,
  "competitionLevel": "LOW",
  "wordCount": 4,
  "isLongTail": true,
  "containsSeed": true,
  "searchUrl": "https://www.blibli.com/cari/laptop%20hp%20official%20store"
}

Sort the dataset by competingProducts ascending and the top rows are your SEO opportunities — real demand (Blibli suggests them) with little competition:

Anti Aging Diskon hingga 60%        3 products   LOW    (curated)
sepatu pria kanky                  42 products   LOW    (related)
isolasi listrik nitto              85 products   LOW    (trending)
laptop hp official store          335 products   LOW    (autocomplete)

Input

{
  "seedKeywords": ["laptop", "sepatu pria"],
  "expansionMode": "alphabet",
  "includeCompetition": true,
  "includeTrending": true
}

Field	Description
seedKeywords	Terms to expand
expansionMode	none (1 req/seed), alphabet (27), alphanumeric (37), related (2), full (38)
includeCompetition	Measure competing products per keyword — 1 extra request each
includeTrending	Add marketplace-wide trending searches (works with no seeds)
includeCurated	Add Blibli's curated promo keywords, with their promo URLs
minWordCount	Set to 3 to harvest long-tail terms only
maxKeywordsPerSeed, maxConcurrency	Volume and politeness
proxyConfiguration	Residential + country ID recommended

Four keyword sources in one run

source	What it is
autocomplete	Suggestions for the bare seed
autocomplete-expansion	Suggestions for seed + a suffix (sourcePrefix records which)
related-search	Blibli's "refine your search" terms
trending	Marketplace-wide trending searches
curated	Editorially promoted keywords, with promoUrl

Everything is deduplicated across sources and seeds in a single run.

Known limits

• No search volume — see above.
• competitionLevel bands are a convenience from this actor, not a Blibli metric; competingProducts is the real number.
• Autocomplete reflects what an anonymous Indonesian visitor sees — the right baseline for marketplace SEO, but a logged-in shopper's box may differ.
• Suggestions are in Indonesian, as shoppers type them.
• With alphabet + competition on, a single seed is ~27 + N requests; use a residential proxy for large seed lists.

See CRAWLING_METHOD.md for the endpoint map, the searchTermPrefix parameter trap, the endpoints that return 500, and the measured expansion figures.