Amazon Customer Insights & Aspect Sentiment

Give it ASINs, get back what customers actually talk about — every aspect Amazon publishes, with sentiment and mention counts, plus the AI review summary.

The negative and mixed aspects are the useful part: they are the questions a listing has failed to answer. That is your FAQ section and your SEO copy, taken straight from a competitor's own review base.

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

Note on Amazon Q&A

If you came looking for a Customer Questions & Answers scraper: that feature no longer exists on Amazon. /ask/questions/… now redirects to the sign-in page, and product pages carry zero Q&A content — verified across several ASINs, the mobile site and every alternate endpoint.

Amazon replaced it with the "Customers say" insight widget, which is what this actor reads. For gap analysis it is arguably better: the opinions arrive already clustered by topic, with sentiment and volume attached.

What you get

One row per aspect:

{
  "recordType": "ASPECT",
  "asin": "B087F16GKW",
  "title": "Gaiam Yoga Mat - Premium 5mm Solid Thick Non Slip",
  "aspect": "grip",
  "sentiment": "MIXED",
  "mentions": 203,
  "mentionRank": 2,
  "mentionSharePercent": 21.4,
  "isContentGap": true
}

Plus one summary row per product, with Amazon's generated paragraph and the gap shortlist:

NAKED Whey Protein          8 aspects, 3 gaps   [mixability, digestibility, texture]
Gaiam Yoga Mat              8 aspects, 4 gaps   [grip, thickness, durability, flatness]
Apple iPhone 11 Renewed     8 aspects, 2 gaps   [battery life, condition]

Read that yoga-mat row as a brief: customers are split on grip, thickness, durability and flatness. Those four headings answer themselves.

How to use it for copy

Goal	Setting
Full picture of one catalogue	defaults
Just the complaints, ranked	contentGapsOnly: true
Ignore long-tail noise	minMentions: 50
Feed a spreadsheet	emitAspectRows: true (default) — one row per aspect sorts and pivots directly

Input

{
  "asins": ["B00XIKPUK4", "B087F16GKW", "https://www.amazon.com/dp/B07ZPKN6YR"],
  "marketplace": "com",
  "contentGapsOnly": false,
  "minMentions": 0
}

Field	Description
asins	ASINs to analyse. Product URLs work too
marketplace	com and ca verified; others see the caveat below
emitAspectRows	One row per aspect alongside the product row (default on)
contentGapsOnly	Only negative and mixed aspects
minMentions	Drop thinly-mentioned aspects
delaySeconds	Pacing between products
proxyConfiguration	Residential strongly recommended

Coverage

Measured over 10 ASINs spanning supplements, electronics, books and sporting goods:

	Coverage
Aspect list (with sentiment + mentions)	10/10
AI review summary	10/10
Aspects per product	8 in every case

That is markedly more reliable than Amazon's Buy Box data, which is withheld from roughly half of anonymous requests. The insight widget was served every time.

Things worth knowing

Products with few reviews carry no insight widget, and that is a real answer rather than a failure. hasInsights: false marks them, so an empty aspect list is never ambiguous.

Amazon caps the widget at 8 aspects. Every product tested returned exactly 8, so treat this as the top 8 by volume, not an exhaustive list.

The summary is Amazon's text, not ours. aiSummary is Amazon's own generated paragraph, extracted verbatim. This actor does not write or paraphrase it.

Only amazon.com and amazon.ca were verified. The development IP was bot-walled on co.uk and de across all 10 TLS fingerprints tried, so their behaviour is genuinely unknown rather than known-bad. Those marketplaces are offered, but use a residential proxy in the matching country, and expect aspect names to be localized — mention counts may then be null while the aspect and sentiment still come through.

Blocked ASINs are reported, never silently dropped. They emit a CAPTCHA_ESCALATION row so a short run is never mistaken for a set of products nobody has opinions about. CAPTCHAs are never solved.

Known limits

• 8 aspects per product, Amazon's cap.
• No individual review text. /product-reviews/ is behind the sign-in wall for anonymous clients, so per-aspect review quotes are not obtainable — the aspect counts and Amazon's summary are what is public.
• Sentiment is Amazon's classification, not an independent model.

See CRAWLING_METHOD.md for the Q&A retirement evidence, the base64 fragment format, and the fingerprint matrix.