Amazon Autocomplete Keyword Expander

Turn a handful of seed phrases into the real keyword ideas Amazon's own search box suggests. This actor reads Amazon's public autocomplete service across 23 marketplaces, can sweep several departments in one run, and expands each seed through eight different query axes so you see far more than the ten suggestions a single lookup returns. No account, cookies, or proxy setup is required.

What this actor does

• 23 Amazon marketplaces — from amazon.com and amazon.co.uk to amazon.co.jp, amazon.com.au and amazon.in
• Multi-market and multi-department runs — compare the same seed across countries and browse aliases such as electronics, kitchen, fashion or books
• Eight expansion axes — suffix A-Z, prefix A-Z, 0-9, questions, comparisons, prepositions, buyer modifiers and your own custom suffixes
• Recursive discovery — optionally treat newly found suggestions as fresh seeds, up to three hops deep
• Two response dialects — enhanced (thumbnails, widget cards, predictive text) or organic keyword responses, or both in one pass
• Filters and ranking — keep questions or long-tail phrases, require or exclude terms, cap results per seed and overall, and order by source, alphabet or the relative volume proxy
• Deduplication with provenance — one row per unique phrase per seed, marketplace and department, showing every query variant that surfaced it
• Empty fields are omitted

Output per suggestion

• marketplace — countryCode, domain, marketplaceId, currency, language
• seedKeyword — the original seed this row belongs to
• query / prefix — the exact query text sent to Amazon
• suggestion — the suggested keyword
• department — the Amazon department alias used
• locale — the marketplace language, e.g. en-US
• suggestionType — Amazon's response type, e.g. KEYWORD or WIDGET; SEED for an echoed seed
• suggType — Amazon's finer subtype, e.g. KeywordSuggestion
• expansionType / expansionToken — seed, suffixAlphabet, prefixAlphabet, digits, question, comparison, preposition, buyerModifier, custom or recursive, plus the letter, digit or word used
• expansionDepth / expansionParent — recursion level and the suggestion that led to this query
• position / amazonRank — 1-based slot in the response that returned the phrase
• bestRank / rank — best slot across all queries, and final rank inside the seed group
• volumeProxy — how many distinct query variants surfaced the phrase; a relative signal, not monthly search volume
• surfacedByQueries — up to 20 query variants that returned it
• searchUrl — ready-to-open Amazon search URL for the suggestion
• wordCount, isQuestion, containsSeed — shape helpers for filtering and analysis
• thumbnailUrl — product thumbnail when the enhanced response supplies one
• widget — optional widget object: id, template, slot, byline, refinement, title, subtitle, imageUrl, imageShape, linkUrl, items[]
• predictiveText — Amazon's predictive completion for the query, when present
• strategyId / strategyApiType / candidateSources / refTag / prior / ghost / help — Amazon's own suggestion metadata
• responseId / shuffled — request correlation id and whether Amazon randomized the response order
• sourceUrl — the exact autocomplete request URL
• recordType: "suggestion", scrapedAt

Input

Field	Type	Default	Description
keywords	array	–	Seed keywords or phrases to expand (required)
marketplace	string	US	Single marketplace; used when marketplaces is empty
marketplaces	array	[]	Optional codes for a multi-market sweep, e.g. ["US","GB","DE"]
departments	array	[]	Department aliases, e.g. ["aps","kitchen","electronics"]
expandAlphabet	bool	false	Append A-Z to each query
expandPrefixAlphabet	bool	false	Prepend A-Z to each query
expandNumbers	bool	false	Append 0-9 to each query
expandQuestions	bool	false	Add locale-aware question stems
expandComparisons	bool	false	Add vs / or / versus / and style tokens
expandPrepositions	bool	false	Add for / with / without / under style tokens
expandBuyerModifiers	bool	false	Add best / cheap / for men style modifiers
customSuffixes	array	[]	Your own suffixes
maxDepth	int	0	Recursion depth, 0-3
queryMode	string	enhanced	enhanced / organic / both
includeWidgets	bool	false	Emit widget rows when available
includeSeed	bool	false	Emit one SEED row per seed
dedupe	bool	true	Merge repeats and add volume/provenance fields
resultOrder	string	source	source / alphabetical / volume
mustContainSeed	bool	false	Drop suggestions without the seed text
includeTerms	array	[]	Keep suggestions containing at least one term
excludeTerms	array	[]	Drop suggestions containing any term
onlyQuestions	bool	false	Keep question-shaped suggestions
onlyLongTail	bool	false	Keep suggestions with 3+ words
minWordCount	int	0	Drop shorter suggestions
maxSuggestions	int	500	Overall record cap (1-10,000)
maxSuggestionsPerKeyword	int	0	Per seed/market/department cap; 0 = no cap
maxQueries	int	400	Overall request cap (1-20,000)
maxConcurrency	int	4	Parallel requests (1-10)
requestDelayMs	int	100	Pause between requests (0-5,000 ms)

Example: one seed, one market

{
  "keywords": ["wireless earbuds"],
  "marketplace": "US",
  "maxSuggestions": 25
}

Example: full keyword-research sweep

{
  "keywords": ["yoga mat"],
  "marketplace": "US",
  "expandAlphabet": true,
  "expandPrefixAlphabet": true,
  "expandQuestions": true,
  "expandComparisons": true,
  "expandPrepositions": true,
  "expandBuyerModifiers": true,
  "maxSuggestions": 800,
  "maxQueries": 400,
  "resultOrder": "volume"
}

Example: compare countries and departments

{
  "keywords": ["running shoes"],
  "marketplaces": ["US", "GB", "DE", "JP"],
  "departments": ["aps", "shoes", "fashion"],
  "maxSuggestionsPerKeyword": 150,
  "maxSuggestions": 1200
}

Example: questions and long-tail only, with thumbnails

{
  "keywords": ["air purifier"],
  "marketplace": "US",
  "expandQuestions": true,
  "onlyQuestions": true,
  "onlyLongTail": true,
  "queryMode": "both",
  "includeWidgets": true,
  "maxSuggestions": 200
}

Use cases

• Amazon SEO and PPC — build exact-match keyword lists from the phrases shoppers actually type into Amazon
• Listing optimization — find long-tail phrasing for titles, bullets and backend search terms
• Niche research — expand one product idea into hundreds of related buying-intent phrases
• International expansion — compare autocomplete demand signals for the same seed in different marketplaces
• Category discovery — sweep departments to see how suggestions change between broad and scoped searches
• Content planning — turn question and comparison suggestions into buying guides and FAQ content

FAQ

Is this a search-volume tool?

No. Amazon does not publish monthly search volume, and this actor does not invent one. position is the order Amazon returned at request time, and volumeProxy counts how many distinct query variants surfaced the same phrase. Both are relative signals for prioritisation, not absolute demand numbers.

Why do I only get about ten suggestions per query?

Amazon's autocomplete response is capped at roughly ten keyword suggestions per request. The expansion axes work around that cap by sending related queries (yoga mat a, a yoga mat, best yoga mat, and so on), then merging and deduplicating the results.

What is the difference between organic and enhanced?

The organic dialect returns the plain keyword list. The enhanced dialect returns that list plus optional thumbnails, widget cards and predictive text when Amazon provides them. Both requests each dialect and unions the results, which costs twice as many requests but can surface extra phrases.

Why are thumbnails missing for some rows or markets?

Thumbnails come from Amazon's enhanced responses and are mostly present on the US and sometimes Canada marketplaces. Other markets usually return none. The field is omitted when Amazon does not supply an image; no placeholder or synthetic image is ever added.

How does the India marketplace work?

Amazon's own India completion host is occasionally unavailable. When that happens, the actor automatically retries the same Indian marketplace through Amazon's UK and UAE completion hosts, keeping the Indian marketplace id and locale. The run status message tells you which host was used.

Why did my run stop before the result count I expected?

Amazon caps each raw response, and the actor also enforces your maxSuggestions, maxSuggestionsPerKeyword, maxQueries and maxDepth limits. The run status message names the limit that stopped it, so you can raise that specific cap and run again.

Why do some department aliases return unrelated suggestions?

Amazon sometimes returns a generic cross-department fallback instead of an empty list. The record keeps the department alias that produced it, so you can see exactly which scope returned the phrase.

Is authentication, a proxy or a browser required?

No. The autocomplete service is public JSON. The actor calls it directly with no login, cookies, API key, browser rendering or proxy configuration.

Is this affiliated with Amazon?

No. This is an independent third-party actor that reads Amazon's public autocomplete responses.

Limitations

• Amazon returns at most about ten keyword suggestions per query, so breadth comes from the expansion axes and the query cap.
• Thumbnails and widget rows are only available from enhanced responses and are sparse outside the US marketplace.
• volumeProxy is a count of distinct query variants, not search volume.
• Department scoping is Amazon's own behaviour; an unrelated alias can return a generic fallback row.
• Recursion and many expansion axes multiply request counts. Use maxQueries, maxSuggestions and maxSuggestionsPerKeyword to keep runs bounded.