Amazon Autocomplete Keyword Scraper

Get real Amazon search suggestions from seed keywords for listing optimization, PPC research and product discovery. Supports the US, UK, Germany, France, Spain, Italy, Canada, Brazil, India and Japan. No API key is required.

Each marketplace uses an appropriate regional Amazon completion host, its own marketplace ID and language. India uses Amazon's EU completion service with the Indian marketplace ID and en_IN locale because the .in completion origin currently returns server errors. Use words customers search in that marketplace: for example, yoga mat in the US and yogamatte in Germany.

Input

{
  "seeds": ["yogamatte"],
  "marketplace": "de",
  "expansions": ["alphabet", "prefixes"],
  "maxResults": 500,
  "maxRequests": 53
}

Field	Default	Description
seeds	Required	1–1,000 keyword strings, each at most 200 characters. Repeated identical seeds are processed once.
marketplace	us	us, uk, de, fr, es, it, ca, br, in, jp.
expansions	["alphabet"]	alphabet: queries the seed, then seed a through seed z. prefixes: adds a seed through z seed, with a separating space. Use [] for the seed alone.
maxResults	10000	Hard cap on output rows across all seeds, from 1 to 100,000.
maxRequests	1000	Cap on distinct suggestion queries, from 1 to 10,000. Retries count as additional HTTP attempts, not additional distinct queries.
maxRetries	2	Up to 0–3 retries per query for rate limits, temporary server errors and network failures.
requestTimeoutSecs	20	Timeout for each HTTP attempt, including its response body, from 1 to 60 seconds.

One seed uses 1 query with no expansions, 27 with either expansion option, and 53 with both. The actual number of returned keywords depends on Amazon and the seed; an obscure or overly specific phrase may return none. No search-volume estimate is included.

Output

Each dataset row retains this format:

{
  "seed": "yogamatte",
  "keyword": "yogamatte rutschfest",
  "marketplace": "de",
  "source": "yogamatte"
}

Suggestions are deduplicated within each seed. The same keyword can appear for two different seeds so its provenance is preserved. source is the exact expanded query that returned the suggestion.

Run summary and troubleshooting

Open the SUMMARY record in the run's key-value store to inspect the endpoint, locale, output count, successful/empty/failed query counts, total HTTP attempts and a bounded sample of request diagnostics. Each sampled request includes its query, HTTP status and suggestion count or error code.

• complete: queries succeeded and keyword rows were delivered.
• empty: Amazon returned valid empty suggestion lists. Try a shorter phrase in the marketplace language.
• partial: some queries failed; other queries succeeded. Review the diagnostics before treating the dataset as complete.
• failed: every attempted query failed. The Actor fails the run instead of reporting a successful empty dataset.

HTTP_ERROR, TIMEOUT, NETWORK_ERROR, INVALID_JSON and INVALID_RESPONSE are reported separately. A blocked or invalid response is never counted as a valid empty suggestion list. Temporary failures are retried with backoff. If truncated is true, truncationReason tells you which limit stopped collection; increase that input limit to collect more.

For a quick connectivity test, use one seed, expansions: [], maxRequests: 1 and maxResults: 10. Avoid comparing English and German phrases as if they represented the same local query behavior.