Amazon Sponsored Ads Intelligence

Extract every Sponsored placement on page one of Amazon search for the keywords you care about — with slot positions and a per-keyword advertiser roll-up showing who is bidding hardest on what.

For brands monitoring competitors' ad strategy and share of voice.

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

What you get

Per sponsored placement:

{
  "recordType": "SPONSORED_AD",
  "keyword": "yoga mat",
  "sampleIndex": 1,
  "position": 1,
  "slotBand": "TOP",
  "asin": "B09M8D4JMD",
  "title": "COOLMOON Yoga Mat Non Slip, Anti-Tear 1/4 Thick TPE...",
  "price": 36.99,
  "currency": "$",
  "rating": 4.6,
  "reviewCount": 1284,
  "isPrime": true,
  "productUrl": "https://www.amazon.com/dp/B09M8D4JMD"
}

Plus one summary per keyword — the competitive picture:

yoga mat           27 ads / 168 slots (16.1%)  topSlot=9  advertisers=16  avgPrice=$61.26
   positions=[1, 1, 2, 2, 3, 3, 3, 4, 4, 11, 11, 12]
   mostPersistentAsins=[{B09M8D4JMD: 2}, {B0F5VW37M7: 2}, {B0HGL1PBQ6: 2}]

wireless earbuds    2 ads /  48 slots ( 4.2%)  topSlot=1  advertisers=2   avgPrice=$23.49
protein powder     12 ads / 156 slots ( 7.7%)  topSlot=4  advertisers=12  avgPrice=$79.06

mostPersistentAsins is the core signal: an ASIN appearing in every sample is bidding continuously, not just winning one impression.

⚠️ Why it samples each keyword more than once

Amazon auctions ad slots per impression, so a single fetch is one draw, not the truth. Measured in one run:

Keyword	sample 1	sample 2	sample 3
yoga mat	13 ads	1 ad	13 ads
wireless earbuds	1	0	1
protein powder	0	0	12

A single fetch of "protein powder" would have told you nobody advertises on it — on a keyword with twelve advertisers. samplesPerKeyword defaults to 3 for exactly this reason. Raise it to 5 for keywords you care about.

It also means zero ads is a real answer, not a bug: measured density across seven keywords ranged 0%–21.7%. A summary row is emitted either way.

Input

{
  "keywords": ["yoga mat", "wireless earbuds", "protein powder"],
  "marketplace": "com",
  "samplesPerKeyword": 3,
  "includeOrganic": false
}

Field	Description
keywords	Search terms to monitor
marketplace	com, co.uk, de, fr, it, es, ca, com.au, co.jp, in
samplesPerKeyword	1–5 impressions per keyword (default 3)
includeOrganic	Also emit non-sponsored results, to compare paid vs organic placement
includeSummary	Per-keyword advertiser roll-up (default on)
maxResultsPerKeyword	Cap results parsed per sample
delaySeconds	Pacing — Amazon serves CAPTCHAs to traffic it dislikes
proxyConfiguration	Residential strongly recommended

Things worth knowing

Prices are forced into the marketplace's currency. Amazon prices pages from the request's geo-IP — from an Indonesian IP, amazon.com returned IDR 336,939.57 instead of $19.99. The actor sends the cookie that pins the currency, so figures stay comparable wherever the run exits from. The currency field is on every row regardless, so never assume USD.

Amazon search is fussy about TLS fingerprints. Most are rejected outright with a 503 block page; the actor ships only the five profiles verified to work and rotates through them. This is a different ladder from the Best Sellers actor — same domain, different defences.

CAPTCHAs are reported, never solved. A challenged keyword emits CAPTCHA_ESCALATION and the run continues with the rest.

Known limits

• Page one only — that is where the bidding competition is.
• This measures ad presence and position, not spend. No public surface exposes bids or CPC; slot position and persistence across samples are the observable proxies for aggression.
• Ad density may be lower for anonymous traffic than for a logged-in shopper, so counts are a lower bound.
• Sponsored Brand banners and video ad units were not present on the pages tested; the actor captures in-line Sponsored Product placements.

See CRAWLING_METHOD.md for the fingerprint matrix, the currency-forcing tests, the parsing pitfall that inflated ad counts 7×, and the full density measurements.