AliExpress Search Scraper | USD Prices and Orders

AliExpress Search Scraper collects AliExpress keyword search result cards. Supply search phrases and receive product links, images, USD prices, ratings, sold text and the page’s trade count. It reads anonymous AliExpress search pages without opening product or store pages.

Try a keyword search

The form is prefilled with usb c cable. Run it to collect up to 48 card occurrences from the first results page. A prefill is a starting example; API requests with omitted or empty keywords perform no searches and exit successfully with guidance.

A 5-row request:

{"keywords": ["usb c cable"], "maxItems": 5, "maxPages": 1}

Input

Field	Type	Default	Meaning
keywords	array	none	Search phrases, processed in input order. Repeated identical phrases are collapsed after whitespace normalization and counted in SUMMARY.
maxItems	integer	48	Rows per keyword after local filters; maximum 200. Zero collects no rows. The requested count is not guaranteed.
maxPages	integer	1	Pages per keyword, maximum 20. Zero collects no rows. Depth is source-limited and the requested count is not guaranteed.
sort	string	default	default, orders, price_asc, price_desc, rating. Source sorts order each delivered page. rating uses default, matching the measured leader mode, which did not return rating-ordered rows.
minPrice	number	none	Source minimum price in USD major units.
maxPrice	number	none	Source maximum price in USD major units.
freeShippingOnly	boolean	false	Source free-shipping refinement.
fourStarsAndUp	boolean	false	Source four-star refinement.
choiceOnly	boolean	false	Source Choice refinement.
shipFrom	string	none	Source ship-from refinement; US is the currently listed value.
minRating	number	none	Keep delivered rows with rating at least this value, from 0 to 5. Applied locally; missing ratings do not pass.
minOrders	integer	none	Keep delivered rows with page-state trade count at least this value. Applied locally; missing counts do not pass.
proxyConfiguration	object	none	Optional buyer proxy settings; default is Apify automatic datacenter, country US. Other countries are refused.

Odd or nonsense phrases can return unrelated recommended products; a matching query does not guarantee relevant results. Filters are the source’s own refinements, except minRating and minOrders, which filter delivered cards before the row limit. minRating and minOrders are applied after each page is fetched. Restrictive values can yield few or no rows while consuming the page and time limits; maxItems limits delivered rows, not pages fetched.

Output

Each row is one delivered card occurrence. Keywords retain input order, pages retain numeric order, and position is the card’s 1-based position on its page, before local filtering. Ranking overlap can repeat the same product on different pages; these occurrences are retained. Duplicate-free traversal is not promised.

The following excerpt is row 3 of retained run BiDOz1BkHO0oWriHj (build 0.1.4, 2026-10-19); fields not shown are omitted:

{
  "keyword": "usb c cable",
  "page": 1,
  "position": 3,
  "productId": "3256812800329908",
  "price": 0.33,
  "originalPrice": 4.13,
  "currency": "USD",
  "formattedPrice": "US $0.33",
  "rating": 4.9,
  "isTopRated": true,
  "orders": 88348,
  "soldText": "50,000+ sold",
  "deliveryDays": {"min": 6, "max": 11},
  "shipFrom": "CN"
}

Rows also carry scrapedAt (UTC), title, productUrl, imageUrl, additionalImages, discountPercent (integer percent), shippingText, tags (card label texts), sponsored, and promotionText. Prices are major currency units, never cents. Ratings use the source’s 0–5 scale; isTopRated means rating at least 4.8 and is null without a rating.

Source-absent values remain null; a missing list is empty. Hidden or renderer-suppressed original prices remain null. Cards displaying “FREE with any purchase” retain that promotion with null price, currency and formatted price. An unreadable displayed field is recorded as an extraction miss rather than silently treated as absent.

orders is the page-state trade count and can exceed rounded soldText. shipFrom comes from each card’s encoded page data. Neither is described as a displayed label. deliveryDays is the page’s delivery range in days; shippingText preserves available shipping and delivery labels.

Billing and partial results

Billing has an Actor Start event and one result event per delivered dataset row. The current Store pricing panel governs rates and tier discounts. Start charging is one event per GB of run memory, minimum one; the default 256 MB run incurs one start event, including when it returns no rows. There is no result charge for errors, empty answers, filtered-out cards or browser attempts. Rows obtained through our browser service are charged once, like other rows.

When the customer’s maximum charge binds, collection stops before another source request. Dataset delivery and charging failures stop collection because their outcome may be unconfirmed.

ERRORS in the run’s key-value store records unresolved keyword-pages, input index, page and reason, plus browser-service diagnostics carrying message and step. Errors never enter the result dataset. SUMMARY records rows, duplicate keywords, empty answers, price-null rows, per-keyword progress, transport activity (including first-attempt timeouts, hedge fires and spare wins) and timing phases. A partially completed run preserves its delivered rows; an all-attempts-failed run fails with guidance. True empty answers succeed without retrying them.

Access and limits

The default run memory is 256 MB. The primary path is browserless, using anonymous US datacenter egress. A currency-preference cookie is sent, and every priced card is verified as USD. A non-USD or unknown-currency priced card rejects the entire page; values are never converted.

The session rotates its proxy label and anonymous cookies after ten reads. A fresh identity’s first read, at start and after a ten-read rotation, may start one spare after 3.75 seconds on another fresh identity; the first verified answer wins and the other request is cancelled. Each attempt has a six-second timeout. Classified challenges receive at most two fresh-label retries; identity or transport misses receive one. A refused label is never replayed. Remaining eligible misses can use our own browser service, pinned to a specific build, with a cap of three calls and 240 seconds per run. A browser attempt adds no buyer charge. Non-USD misses receive fresh-label retries and then an error; they do not currently use the browser route. The source query and page must match before rows can be delivered. Login pages are not crossed and CAPTCHAs are not solved.

Collection has a 270-second deadline after proxy setup. Remaining work is recorded as TIME_LIMIT or NOT_ATTEMPTED. Default or empty-group proxy settings use US residential egress for the browser service. An explicit DATACENTER-only restriction or custom proxy URLs cannot use that service and end with BROWSER_PROXY_UNSUPPORTED; their primary route is preserved. Explicit useApifyProxy: false keeps the browser service on direct egress.

This Actor covers keyword search on aliexpress.com in US/USD. Product details, reviews, stores, image search, category browse and other regional storefronts are outside its scope. Page results and prices reflect an anonymous snapshot and can change between reads.

AliExpress’s robots.txt has no rule for the /w/ search path this Actor reads. AliExpress Terms of Use (version effective September 26, 2026) clause 3.2(a) states: “Systematic retrieval of Site Content from the Sites to create or compile, directly or indirectly, a collection, compilation, database or directory (whether through robots, spiders, automatic devices or manual processes) without written permission from AliExpress.com is prohibited.” The same clause states: “(b) you will not copy, reproduce, download, compile or otherwise use any Site Content for the purposes of operating a business that competes with AliExpress.com, or otherwise commercially exploiting the Site Content.” Consider those terms before use: AliExpress Terms of Use.

Help

For missing results, inspect ERRORS and SUMMARY, then try another keyword or fewer filters. Report the run URL, input and expected fields in the Actor’s Issues tab. Avoid sharing credentials or custom proxy passwords.