Build1 distinct publisher3 min readUpdated
Seth Godin called Amazon's search ads "legal theft" and got the arithmetic wrong. Amazon's filings show a bigger business than he described, and a discovery layer that brands now rent back.
The Engineer · Build desk

Compiled by The EngineerSomething wrong?How this is made
Seth Godin attacked Amazon's search-ad model on August 18th, after his publisher began paying roughly $1 per click to bid on the query "Seth Godin The Knot" - the words a customer types when they already know what they want [1]. The profit arithmetic in his post does not survive contact with Amazon's disclosures, but the structural complaint underneath it does, and Amazon's own numbers make the case better than his did [3][4][7].
Godin, who founded Yoyodyne and Squidoo, called the system "legal theft" on his blog, arguing that Amazon controls product discovery, moves paid listings into that discovery layer, and leaves merchants buying ads to defend demand their brands created elsewhere [2]. His publisher's campaign supplied the example: the highest-yielding search term was the exact combination of his name and his book title [c2b].
He also wrote that Amazon makes nearly $1 billion in weekly profit from search ads and more than $50 billion a year [3]. Amazon reported $19.81 billion in advertising-services revenue for the quarter ended June 30th, up 26% year on year [4], about $1.52 billion per week [5], and roughly $76.1 billion across the trailing four quarters [6]. Annualising the June quarter gives a run rate near $79.2 billion, some $29 billion above Godin's annual figure [1][2]. The 26% growth implies a year-ago quarter of about $15.7 billion, an increase of roughly $4.1 billion in twelve months [4]. Those totals cover sponsored ads, display and video; Amazon breaks out neither search-ad revenue nor advertising profit, so the weekly profit number cannot be checked against the filings at all [7].
The employee-bonus flourish fails the same way. Godin said one year of ad proceeds could fund a $35,000 bonus for every employee [c3b]; Amazon reported 1.595 million full- and part-time employees at the end of the second quarter [8], which puts trailing ad revenue at about $47,700 per head [9]. The bonus would cost roughly $55.8 billion, about 73% of gross ad revenue, before anyone pays for the infrastructure that serves the ads [3].
Strip the profit claims out and the mechanism remains. Amazon describes Sponsored Products as cost-per-click listings that appear at the top of, alongside or within shopping results, on product pages, and in some cases off Amazon entirely; advertisers pick keywords, set bids and pay on the click [10]. That is a defensive auction: a brand builds demand through product, reviews and outside marketing, then either bids on its own name or cedes the most visible slot to a competitor who will [11]. The Federal Trade Commission alleges in its ongoing case that sponsored listings took over the most visible search positions and pushed organic results down the page, and a later complaint cited an internal Amazon study finding sponsored products frequently priced above neighbouring organic results - allegations, not findings [12][13]. A 2024 study of more than 2,000 Marketplace searches found the three most-clicked ads were on average 17% more expensive and a third less relevant than comparable organic listings [14]. In a Management Science field experiment on an unnamed platform, Sarah Moshary removed sponsored search for 3% of visitors and found ads cannibalised organic transactions and cut total volume, while still paying off because ad revenue exceeded lost commissions [15][16][17].
Watch the disclosure line, not the rhetoric: Amazon said on July 30th that it had expanded Ads Agent, an AI planning system, which lowers the cost of participating in the auction without lowering the toll [18]. If the next quarter posts another 26% and search revenue still is not broken out, both Godin's critics and his sympathisers are arguing from the same missing number [4][7].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Seth Godin attacked Amazon's search-ad model on August 18th after his publisher began paying roughly $1 per click to advertise his new book against the query "Seth Godin The Knot" - the words a customer would type when already looking for it.
Godin, the author and entrepreneur who founded Yoyodyne and Squidoo, called the system "legal theft" in a post on his blog, arguing that Amazon controls product discovery, moves paid listings into that discovery layer, and leaves merchants buying ads to protect demand their brands created elsewhere.
Godin's publisher found that the highest-yielding search term was the exact combination of his name and book title.
Godin wrote that Amazon makes nearly $1 billion in profit from search ads each week and more than $50 billion annually.
Godin argued that Amazon could give each employee a $35,000 bonus with one year of ad proceeds.
Amazon reported $19.81 billion in advertising-services revenue for the quarter ended June 30th, up 26% from a year earlier.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Primary filings plus peer-reviewed research, with the key profit number unverifiable
The financial spine rests on Amazon's own investor disclosures (quarterly and trailing advertising revenue, headcount) and the auction mechanics on Amazon's advertising documentation. The market-effect claims draw on a 2024 study of more than 2,000 Marketplace searches and a Management Science field experiment, plus FTC complaint material. Evidence is short of complete on the two points that matter most to the polemic: Amazon does not break out search-ad revenue or ad profit, and the field experiment's platform is unnamed.
Advertiser spend is large, growing and being further automated
Adoption of the mechanism under scrutiny is not in doubt: roughly $76.1bn of advertising-services revenue over four quarters, 26% year-on-year growth in the June quarter, and an implied ~$4.1bn quarterly increase all show merchants paying at scale. Amazon is expanding Ads Agent to lower the effort of participating, and the cluster contains one concrete named advertiser instance (Godin's publisher paying ~$1 per branded-query click).
The viral profit math overstates; the underlying structure is understated
Godin's headline numbers run ahead of what can be shown: "nearly $1bn in profit per week" from search ads is unverifiable because Amazon reports no search-ad profit, and the $35,000-per-employee bonus would cost about $55.8bn, roughly 73% of gross trailing ad revenue rather than a share of profit. In the opposite direction his ">$50bn annually" understates a business running near $79.2bn annualised. Net positive because the sensational profit framing exceeds available evidence, moderated by the fact that the structural tollbooth argument is corroborated by Amazon's own product documentation, FTC allegations and independent research.
Interested parties on every side of the numbers
The critique originates from an author actively marketing a new book, whose campaign is itself the example; the efficiency figures for Ads Agent are company-supplied by Amazon without a published sample; and the sharpest allegations about sponsored listings come from a regulator litigating against Amazon. Peer-reviewed research provides the least interested evidence in the cluster, and the reporting publisher discloses these positions rather than obscuring them.
Well-anchored numbers, single-publisher cluster
Confidence is lifted by the quantitative claims tracing to Amazon's own disclosures and to named peer-reviewed studies, and by the article's explicit caveats. It is capped by there being only one publisher in the cluster with no independent corroboration or Amazon response, and by the fact that the core dispute — search-ad profitability — sits behind a disclosure gap no source in the cluster closes.
invest
Behind-the-meter gas is the data center buildout's real cost: 318 Mt a year1 distinct publisher
leadership
Sponsorship Is Not Retention: What An Amazon H-1B Win Still Does Not Buy1 distinct publisher
product
Also raised $150M for a parts bin, not a bicycle3 distinct publishers
invest
PublicSquare's 99% collapse shows political affinity is a marketing channel, not a moat1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 18, 2026