AliExpress Reviews Scraper — AI complaints, no login · $1.5/1k

Part of LeoWorks AliExpress tools: search → rank (AliExpress Search Scraper) · reviews → complaints (AliExpress Reviews Scraper).

Scrape AliExpress reviews and see what buyers really complain about. Give it AliExpress product URLs or IDs and get every review's text, English translation, stars, date, buyer country, option bought and photos — optionally with AI complaint types and sentiment, free star, date and keyword filters, and a per-product report — an AliExpress reviews scraper with built-in AliExpress review analysis. No login or cookies: 20 reviews classified cost about $0.04, and the free $5 monthly Apify credit covers about 2,500 classified reviews.

Independent tool — not affiliated with, endorsed by or sponsored by AliExpress or Alibaba Group. The name is used only to describe the data source.

Why use it

• Dropshipper or sourcing agent — read the critical reviews of a product before you buy stock: set Maximum stars to 3 and the Actor reads on until it has them.
• Brand owner — see which complaint types (quality/defect, size/fit, delivery, packaging) drive bad reviews on your AliExpress listing.
• Supplier comparison — put 3 listings side by side in REPORT_PRODUCTS.csv: complaint mix, worst options, buyer countries.
• Quality assurance — find 5-star reviews that still report a defect, and photo evidence of problems.
• Customer service — a queue of 1–2 star complaints and seller-service complaints to act on.
• Developer or AI agent — reviews and labels as structured data through the API or MCP.

Performance at a glance

	Result
Complaint type	93% on 30 reviews it had not seen before (a small set, so read it as a rough guide)
Sentiment	87% (same set)
Speed	1,000 reviews (893 classified) in about 8 min on the default 256 MB (measured 2026-10-05)
Cost	$1.50 per 1,000 reviews collected plus $0.50 per 1,000 classified — $2 per 1,000 classified reviews

Details and limits are under Accuracy and limits below.

Compliance and privacy

• Public pages, no login. It reads publicly visible review pages; no AliExpress account, cookies or login are needed.
• No reviewer names or avatars. They are not collected. Only the review text (or its English translation) and the star rating are sent for classification.
• Nothing is kept unless you ask. If you set Monitor key, the Actor saves only summary figures (average rating, review count, newest review date) in a key-value store named monitor-aliexpress-reviews in your own Apify account, to compare with the next run. No review text, names or IDs are stored.
• You stay responsible for the use. Follow AliExpress' terms and local laws, and do not use personal data. This is not legal advice.

See it work

Real run: 1,000 reviews of one AliExpress earbuds listing, Oct 2026 — 893 classified, 51% with a complaint, 41 five-star reviews with a reported problem.

What can the AliExpress Reviews Scraper do?

Why filters matter on AliExpress: the reviews come in AliExpress' own order, which leans heavily to 5 stars. In our test, the first 100 text reviews of an earbuds listing averaged 4.91 stars, while AliExpress' own figure for all 25,872 reviews is 4.1 stars with 17.1% negative. Set Maximum stars to 3 and the Actor reads on until it has the critical reviews — reading is free; you pay only for the reviews it keeps.

• Reviews — original text + AliExpress' English translation, stars (1–5), date, country, option (color/size), photos, aspect ratings (e.g. "Durability: good"), helpful votes, follow-up reviews, and AliExpress' own "AI-written" and "featured" marks
• AI classification (optional) — complaint types (quality/defect, size/fit, delivery, packaging, price/value, customer service, no effect, skin/body reaction, other), sentiment, purchase motive and up to 10 custom labels in plain language, with probabilities; flat columns for spreadsheets
• What AliExpress knows about all reviews — total, average, star counts, negative rate, its topic tags ("poor sound quality" — 901 reviews) and its aspect panel (Durability: 22% poor / 56% ok / 22% good), in the summary at no extra request
• Per product — a one-line summary, complaint mix, worst options, buyer countries, monthly trend and last 90 days vs before, star–text mismatches → REPORT_PRODUCTS.csv
• Reviews to act on — 1–2-star complaints, skin or body reactions and seller-service complaints → ACTION_QUEUE.csv
• Free filters — stars, date, keywords and buyer countries; star-only and repeated reviews skipped — none of them stored or charged

You are	Use it to
Dropshipper or sourcing team	Read the 1–3-star reviews of a product before you list it: what breaks, which option is worst, which countries complain
Seller	Watch your listings every week: new complaints, star–text mismatches, reviews to answer
Comparing suppliers	Run 3 listings of the same product and compare REPORT_PRODUCTS.csv side by side
AI agent or dashboard	Structured reviews and labels instead of raw text

Works from Claude, Cursor or Claude Code — add it as a tool through the Apify MCP server and ask in plain language (setup in Use with Claude, Cursor or Claude Code below).

How to scrape AliExpress reviews

1. Paste AliExpress product URLs (any site — aliexpress.com, aliexpress.us, ko.aliexpress.com …) or product IDs. To find products first, use our AliExpress Search Scraper and pass its productUrls.
2. Set Reviews per product (the form starts at 20 for a quick first run).
3. Optional: Maximum stars (e.g. 3), Only reviews newer than (6 months), Review must mention one of (broken, battery), Buyer countries, up to 10 Custom labels.
4. Click Start. 2 products × 100 reviews take about 80 seconds (about 8 minutes per 1,000 reviews — start with 20–100 and raise it once the result looks right).
5. Open the Reviews view, or download REPORT_PRODUCTS.csv and ACTION_QUEUE.csv (Output tab).

Recipes

Goal	Input
Check a product before sourcing	{"productUrls": ["…"], "reviewsPerProduct": 50, "maxStars": 3}
Photo evidence of defects	{"productUrls": ["…"], "reviewFilter": "with_photos", "maxStars": 2}
Compare 3 suppliers	{"productUrls": ["…", "…", "…"], "reviewsPerProduct": 200} → compare REPORT_PRODUCTS.csv
Weekly monitoring	{"productUrls": ["…"], "reviewsNewerThan": "7 days"} + an Apify Schedule

Pricing

Pay only for what you collect — no subscription.

Event	Price	When
review	$0.0015	One AliExpress review collected (text, English translation, rating, date, country, option, photos).
review-judged	$0.0005	Complaint types, sentiment and purchase motive for one review (only when classification is on).

How much does it cost to scrape AliExpress reviews?

That is $1.50 per 1,000 reviews plus $0.50 per 1,000 classifications — $2 per 1,000 classified reviews.

Cost examples

Run	Cost
20 reviews of one product, classified	$0.04
1,000 reviews, classified	$2.00
1,000 reviews, not classified	$1.50
10 competitor products × 200 reviews, classified	$4.00

With the free $5 monthly Apify credit you can collect about 2,500 classified reviews or 3,333 reviews without classification.

Failed or unknown products are reported with an error row and not charged. Star-only reviews, reviews outside your filters and repeated reviews are skipped and not charged (turn on Include star-only reviews to collect those). If you set a maximum cost per run, the Actor stops cleanly when it is reached (classification is switched off first when the budget only covers collection).

Is it free to try?

Yes. The form defaults collect and classify 20 reviews of one product for about $0.04 (10 seconds), and the free $5 monthly Apify credit covers about 2,500 classified reviews or 3,333 reviews without classification (see the cost examples above).

Input and output examples

Output sample

Real rows from run lwSrLcHd9JBKwwveK (2026-10-11, 2 earbuds listings × 100 text reviews, one custom label "arrived broken or damaged").

stars	country	original (shortened)	AliExpress translation (shortened)	complaints	sentiment	motive	custom
1	US	прийшло повністю пошкрябане і наче…	It arrived completely scratched and as if it had already been used.	quality_defect	negative (1.00)	unknown	broken (0.91)
5	PE	tuve que retirarlo en centro de distribución…	I had to pick it up at the distribution center because the delivery drivers are not doing their job…	delivery	negative (0.97) — star–text mismatch	unknown	no
5	UA	якість на тройочку з мінусом. звук громкий…	the quality is below average. the sound is loud. If you need quality, buy more expensive options.	quality_defect	negative (0.97) — star–text mismatch	unknown	no
5	KR	퀄리티가 넘 좋고 음질도 넘 만족스러워요! 다만 귓구멍에 착 앵기진않네요…	The quality is really good. The sound quality is also very satisfying! However, it doesn't fit snugly in the ear hole…	size_fit	positive (1.00)	unknown	no
5	FR	top super écouteurs et le son bien…	Super headphones and good sound. I am satisfied, this is the second earphone I receive…	none	positive (1.00)	repurchase	no
5	US	Все прийшло.Подарунок виграв на Алиєкспрес	Everything has arrived. The gift was won on Aliexpress.	none	positive (0.98)	gift	no

The same run's product summary (REPORT_PRODUCTS.csv):

product	summary	collected avg	AliExpress avg (all reviews)	AliExpress negative rate	top negative topic (all reviews)
1005008549911526	98 reviews classified, 28% with a complaint — mostly quality/defect (17) and other (6).	4.91	4.1 (25,872 reviews)	17.1%	poor sound quality (901)
1005010748356463	99 reviews classified, 21% with a complaint — mostly quality/defect (11) and other (5); worst option: Color:Green Ships From:China Mainland (24%).	4.93	4.6 (970 reviews)	3.5%	average sound quality (42)

With Maximum stars 3 on the first listing (run ubfBvnkFdqaFIZJ3K), the Actor read 62 pages (1,240 reviews), kept the first 50 reviews of 1–3 stars — 92% with a complaint, mostly quality/defect (43) — and charged only those 50. 22 went to ACTION_QUEUE.csv, for example "The product did not work. It does not turn on or make any sound" (1 star, CL) and "I received a broken box without the earphones" (1 star, PL).

Input example

The form default — one product, its first 20 text reviews in AliExpress' order, classified (about $0.04, 10 seconds):

{
  "productUrls": ["https://www.aliexpress.com/item/1005008549911526.html"],
  "reviewsPerProduct": 20
}

The critical reviews of two listings, last 6 months, with a custom label:

{
  "productUrls": ["https://www.aliexpress.com/item/1005008549911526.html", "1005010748356463"],
  "reviewsPerProduct": 100,
  "maxStars": 3,
  "reviewsNewerThan": "6 months",
  "customLabels": ["arrived broken or damaged"]
}

Output files and columns

{
  "type": "review",
  "productId": "1005008549911526",
  "reviewId": "60093737103906642",
  "rating": 5,
  "date": "2025-11-15",
  "country": "US",
  "text": "Все прийшло.Подарунок виграв на Алиєкспрес",
  "textEnglish": "Everything has arrived. The gift was won on Aliexpress.",
  "language": "other",
  "option": "Color:black",
  "images": ["https://ae-pic-a1.aliexpress-media.com/kf/A4d5c8b0dedb6489ebd7b6d2cfa37a710f.jpg", "…"],
  "imageCount": 2,
  "aspects": [{ "aspect": "Quality of sound", "value": "Fast", "rating": "good" }, { "aspect": "Durability", "value": "Fast", "rating": "good" }, { "aspect": "User Friendly", "value": "Good", "rating": "good" }],
  "helpfulVotes": 0,
  "unhelpfulVotes": 0,
  "aiWritten": false,
  "featured": false,
  "additionalReview": null,
  "hasFollowUp": false,
  "complaintTypes": "none",
  "sentiment": "positive",
  "motive": "gift",
  "customLabelsMatched": "",
  "starTextMismatch": false,
  "labels": {
    "complaint": [{ "label": "none", "labelKo": "불만 없음", "probability": 0.97 }],
    "sentiment": { "label": "positive", "labelKo": "긍정", "probability": 0.98, "confidence": 0.97 },
    "motive": { "label": "gift", "labelKo": "선물", "probability": 0.85, "confidence": 0.82 },
    "custom": [{ "label": "arrived broken or damaged", "probability": 0.04, "matched": false }]
  },
  "classifiedFrom": "english_translation",
  "productUrl": "https://www.aliexpress.com/item/1005008549911526.html"
}

Reviews written in other languages are classified from AliExpress' own English translation (classifiedFrom). Reviewer names and avatars are not collected. aspects[].value is AliExpress' text, which is often generic ("Durability: Fast"); rating is the poor / ok / good score behind it. aiWritten and featured are marks AliExpress puts on the review, not our judgment. starTextMismatch is true for 4–5 stars with a clearly negative text (probability ≥ 0.7) or 1–2 stars with a clearly positive one.

Summary by product (REPORT, REPORT_PRODUCTS.csv)

Each run also saves, at no extra charge:

• REPORT (JSON) — per product and overall: headline (one-sentence verdict: top complaint and its share, worst option, 90-day trend), with monitorKey the change since the last run (previousRunAt, ratingDelta, newReviews, newCriticalIssues), summary, allReviews (AliExpress' figures for all reviews: total, averageRating, stars, negativeRate, photoReviews, followUpReviews, topics, aspects), and for the reviews collected: star distribution, complaint rate and mix, sentiment shares, purchase motives, custom label shares, worstOptions, countries, monthly, last90Days / before90Days, starTextMismatches, actionReviews, lowSample (fewer than 20 classified) and the 3 strongest complaint reviews.
• REPORT_PRODUCTS.csv — one row per product: productId, productUrl, headline, summary, reviewsCollected, classified, lowSample, averageRating, complaintRate, negativeShare, positiveShare, topComplaints, topMotive, worstOption, starTextMismatches, actionReviews, last90DaysComplaintRate, before90DaysComplaintRate, allReviewsTotal, allReviewsAverageRating, allReviewsNegativeRate, allReviewsTopNegativeTopics.
• ACTION_QUEUE.csv — reviews to act on, lowest stars first: productId, rating, date, reason, complaintTypes, sentiment, option, country, text, textEnglish, images, reviewId, productUrl.
• SUMMARY — counts, and a status per product (ok, not_found, error, scan_limit).

Photo evidence (beta)

A complaint is stronger with a photo. Turn on Judge review photos and the Actor looks at the photos buyers attached and checks whether a photo really shows the damage they describe. It is off by default and adds no extra charge; it adds a few seconds per photo review.

Setting	What it does
Judge review photos	Only reviews with a complaint or 3 stars or fewer (recommended) or All reviews that have photos
Photos judged per review	The first 1–3 photos of each review (2 by default; at most 2 with "All reviews that have photos")
Photo size	Longest side requested from the photo server: 512, 768 or 1024 pixels

Each judged review gets photoEvidence:

Value	Meaning
supports_complaint_product	A real product photo shows a defect on the product itself, and the review complains about it
supports_complaint_packaging	A real product photo shows damaged packaging (box, bag, film), and the review complains about damage
damage_seen_text_silent	The photo shows damage, but the review text does not complain about it (for example a praise review with a torn box)
no_evidence	A real product photo, no visible damage
not_product_photo	A screenshot, a catalog image, a label, a blank image or something unrelated

Product defects with a photo go to the top of ACTION_QUEUE.csv with the reason photo-backed defect; damaged packaging comes right after with photo-backed packaging damage (a supplier claim and a shipping problem need different action). The queue has the columns photoEvidence and photoUrl. The dataset view Photo evidence (beta) lists them, and photoStatus says why a review was not judged (no_photo, skipped_by_rule, too_small, download_failed, judge_failed).

• Beta: We scanned 1,193 review photos from 48 AliExpress products (mostly electronics and accessories) and a person looked at every photo the Actor flagged: 16 of 16 photos marked supports_complaint_product and 33 of 33 marked supports_complaint_packaging really showed damage the review complains about (95% lower bounds: 81% and 90%). Three of the 33 packaging cases were damage to the product itself, so the product-or-packaging split is right about 91% of the time. It finds roughly half of the real product defects (50%) and about four in five packaging cases (79%) by our reference labels, so it can miss defects, and these figures only cover the photos it flagged. We have not measured Google or Amazon photos. Use it to find evidence quickly, not as proof for a claim.
• Privacy: photos are downloaded only while they are judged and are sent to TypeSafe (the AI model provider) for that purpose. This Actor does not store them, and no question is about people. The dataset keeps the photo address that AliExpress gave, which can stop working later.
• Skipped: photos shorter than 280 pixels on the short side, and photos that cannot be downloaded. The review itself is still collected and classified.

API, MCP and integrations

Works with

• Apify API, JavaScript and Python clients — see Call it from code below.
• Apify Schedules — a weekly schedule with Only reviews newer than 7 days keeps watching your own and competitors' listings; our own monitoring runs this Actor every morning.
• AliExpress Search Scraper — find products with leoworks/aliexpress-search-scraper and paste their productUrls here (or pass them from code).
• Claude, Cursor and Claude Code through the Apify MCP server — see below (verified 2026-10-06).
• Other review sources — the same labels for any review dataset: Amazon Review Classifier, Korean Review Classifier.

Call it from code

curl -X POST "https://api.apify.com/v2/acts/leoworks~aliexpress-reviews-classifier/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"productUrls": ["1005008549911526"], "reviewsPerProduct": 50, "maxStars": 3}'

JavaScript (npm install apify-client):

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });
const run = await client.actor('leoworks/aliexpress-reviews-classifier').call({ productUrls: ['1005008549911526'], reviewsPerProduct: 100 });
const report = await client.keyValueStore(run.defaultKeyValueStoreId).getRecord('REPORT');
console.log(report.value.groups.map((g) => g.summary));

Python (pip install apify-client):

from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("leoworks/aliexpress-reviews-classifier").call(run_input={"productUrls": ["1005008549911526"], "maxStars": 3})
defects = [r for r in client.dataset(run["defaultDatasetId"]).list_items().items if "quality_defect" in (r.get("complaintTypes") or "")]

Use with Claude, Cursor or Claude Code (MCP)

Ask your AI agent. Add the Apify MCP server with this Actor as a tool, then say, for example:

• "Get the latest 50 reviews of https://www.aliexpress.com/item/1005008549911526.html and tell me the main complaints."
• "From https://www.aliexpress.com/item/1005008549911526.html fetch 30 reviews with 3 stars or less and list what buyers complain about, including the worst-rated option."
• "Get the 40 newest reviews of https://www.aliexpress.com/item/1005008549911526.html that mention “battery” and summarize what buyers say about it."

The agent calls the tool leoworks--aliexpress-reviews-classifier and reads the rows and the per-product summary with get-dataset-items and get-key-value-store-record.

Claude Desktop or Cursor (mcp.json):

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=leoworks/aliexpress-reviews-classifier",
      "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" }
    }
  }
}

Claude Code: claude mcp add --transport http apify "https://mcp.apify.com?tools=leoworks/aliexpress-reviews-classifier" --header "Authorization: Bearer YOUR_APIFY_TOKEN". Leave out the header to sign in with OAuth in the browser instead. Your Apify token is in Console → Settings → API & Integrations. We verified this setup with the Apify MCP server (v0.17.2) on 2026-10-06.

Accuracy and limits

Accuracy

Measured on hand-labelled AliExpress reviews (earbuds, phone cases, dresses, kitchen tools, LED strips, car holders; about half with 1–3 stars): complaint type 93% and sentiment 87% on 30 reviews it had not seen before. Labels were set by us; automated labels can be wrong, so check samples before making big decisions.

Limits

Item	Limit
Reviews per product	Up to 5,000 per run
Order	AliExpress' default order, which leans to 5 stars; the source cannot sort by date or stars
Filters	Source: all · with photos · with a follow-up · local buyers. Free filters: stars, date, keywords, buyer countries. With free filters on, up to 10,000 reviews per product are read to find matches
Repeated reviews	AliExpress sometimes returns the same review twice (5 repeats while collecting 200 reviews in one of our runs, none in a run of 1,000); repeats are skipped and not charged
Classification	English, Korean, Japanese text; other languages via AliExpress' English translation
Product details	Not included — this Actor collects reviews only. Use the AliExpress Search Scraper for titles and prices
Speed	1,000 reviews (893 classified) in about 8 min on the default 256 MB (measured 2026-10-05)

Run it on a schedule

Weekly is the usual rhythm: set Only reviews newer than to 7 days and a Monitor key, and each run reports new reviews and new serious issues since the last one. Create an Apify Schedule for a saved task with your product URLs — your own listings and competitors'. To start it from other tools, use the Apify n8n or Make integration, or the Apify API.

More Actors from Leoworks

Need	Actor
AliExpress product search, prices and rank tracking	AliExpress Scraper — search results & rank
Korean reviews (Coupang, Olive Young, Naver)	Korean Review Classifier
Amazon reviews	Amazon Review Classifier
Google Maps reviews	Google Maps Review Classifier
Japanese / Spanish reviews	Japanese Review Classifier · Spanish Review Classifier
Social media comments	Social Comment Classifier

FAQ and support

Why not use the official AliExpress API?

AliExpress' Open Platform APIs need an approved app key and are built for affiliate links, product data and orders; we found no endpoint that exports a product's reviews with text, photos and buyer country (checked October 2026). This Actor reads the public review pages without logging in.

What are the alternatives for AliExpress review data?

Reading reviews on the product page (AliExpress shows the best ones first, see Why are AliExpress reviews mostly 5 stars?), building your own scraper, or pasting reviews into a chatbot. This Actor collects the reviews, translates them and, if you want, labels complaint types and sentiment in the same run.

Can I get only negative AliExpress reviews?

Yes. Set Maximum stars to 3 and the Actor reads on until it has the critical reviews — reading is free; you pay only for the reviews that pass your filters.

How do I scrape AliExpress reviews?

Paste one or more AliExpress product URLs or IDs under Product URLs, choose how many reviews per product, optionally turn on AI classification, and click Start. No login or cookies are needed. Step by step: How to scrape AliExpress reviews above.

Which AI makes the judgments?

Jev, TypeSafe's decision model (version jev-1.13.0, pinned). Jev answers each label with a calibrated probability instead of generated text, so a 0.9 means the same thing for every item. Answers are stable but not fixed: re-running the same input can change a few borderline answers (under 1% of answers in our repeat tests) — filter on probability when an item must never flip. Only the review text (or its English translation) and the star rating are sent to Jev.

Why are AliExpress reviews mostly 5 stars?

AliExpress returns reviews in its own order, which puts satisfied buyers first. Use Maximum stars to read the critical ones, and compare with allReviews in the summary — AliExpress' own figures for every review of the product.

AliExpress already shows topic tags like "poor sound quality". What does this add?

Those tags count phrases across all reviews; they do not tell you which review says what. Here every review gets its complaint types with a probability, so you can filter, read the examples, see which option or country they come from and follow them month by month. Both are in the output: the tags as allReviews.topics, our labels per review.

How is this different from scrapers with built-in sentiment?

Some use a word list (a review with "not bad" or "broken box, but works" trips them up). Jev reads the whole review, separates a defect from a size or delivery problem, gives a probability, and its accuracy is measured (see Accuracy).

Why not paste the reviews into ChatGPT?

For a few you can. For hundreds per product you would write prompts, handle limits and parse free text; here you get a fixed set of labels with probabilities, summaries and CSVs, for $0.50 per 1,000 classifications.

How many reviews should I collect?

50–100 critical reviews (Maximum stars 3) usually show a product's real problems; 200+ for a stable complaint rate. Below 20 classified reviews the summary is marked lowSample.

Can I get only new reviews each week?

Set Only reviews newer than to 7 days and schedule the run. AliExpress cannot sort by date, so the Actor reads through the reviews to find them — reading is not charged.

To see what changed since the last run, set Monitor key (any name, e.g. my-suppliers) and run again with the same key: each product row then shows ratingDelta, newReviews and newCriticalIssues (new reviews that need action), with previousRunAt. The first run with a key only saves the baseline.

Why are some reviews missing labels?

Reviews with photos but no text, or in a language without an English translation, are returned without labels and charged only as review.

What happens with a wrong product ID?

It is reported as not_found in an error row and in SUMMARY, and nothing is charged.

Is it legal to scrape AliExpress reviews?

It reads publicly visible review pages without logging in. You are responsible for how you use the data; follow AliExpress' terms and local laws, and do not use personal data.

Reviews and support

If this Actor saved you time, a short review on Apify Store helps others find it. Questions or a product that does not work? Open an issue in the Issues tab — we answer within a day.

Need a custom label set, a different product source or a feature? Say so in the issue — we are open to custom solutions.

Changelog

See the Changelog tab.