Pluralistic: Amazon Achieves Enshittification Inception  🔗 **Source:** [Pluralistic: Amazon Achieves Enshittification Inception](https://pluralistic.net/2026/09/04/cheating-at-fraud/) > [!note]+ Document Notes > tags: [enshittification, amazon, platform-monopoly, antitrust, auction-design, marketplace-ranking] > author: Cory Doctorow > source: Pluralistic > reference: https://pluralistic.net/2026/09/04/cheating-at-fraud/ > date created: 2026-09-04 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: Amazon achieves enshittification inception > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - Source: https://pluralistic.net/2026/09/04/cheating-at-fraud/ (~3,475 words, published 2026-09-04) - [FTC and states sue Amazon over secret ad surcharge scheme](https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme) — the primary document behind the piece - [Ars Technica: FTC alleges Amazon illegally made $20 billion by rigging billions of ad auctions](https://arstechnica.com/tech-policy/2026/09/ftc-alleges-amazon-illegally-made-20-billion-by-rigging-billions-of-ad-auctions/) — source for Amazon's stated defense - [Gizmodo: FTC sues Amazon for allegedly duping advertisers](https://gizmodo.com/ftc-sues-amazon-for-allegedly-duping-advertisers-2000805199) - [Payola](https://en.wikipedia.org/wiki/Payola) and [Vickrey auction](https://en.wikipedia.org/wiki/Vickrey_auction) and [auction theory](https://en.wikipedia.org/wiki/Auction_theory) — the conceptual scaffolding - Doctorow's own prior posts on [Most Favored Nation price-fixing](https://pluralistic.net/2026/02/25/most-favored-nation/#price-fixing) and [subprime attention rent](https://pluralistic.net/2023/11/03/subprime-attention-rent-crisis/#euthanize-rentiers) - Tim Wu, *The Age of Extraction* — cited for the newspaper-advertising comparison - Doctorow, *Enshittification* (Farrar, Straus and Giroux) — the source of the earlier payola figures ## 📝 Summary - **What happened.** The FTC and 22 state attorneys general sued Amazon in August 2026, alleging that for seven years it secretly overcharged winners of its search-advertising auctions — charging them their own maximum sealed bid rather than the second-highest bid plus a penny, which is what a second-price auction is defined to do. - **The scale.** According to the suit as Doctorow reports it, Amazon did this roughly 80% of the time; Ars Technica's headline figure is $20 billion. - **Why it matters more than a normal ad-fraud story.** Search placement is, on Doctorow's account, Amazon's single largest profit centre — larger than everything it sells directly and larger than AWS — on track to top $80 billion this year, up from the mid-$30 billions when he was researching *Enshittification*. - **The structural argument.** Amazon has converted its two largest cost centres into profit centres funded by its own competitors: fulfillment is more than fully paid for by platform-seller fees, and AWS makes Amazon's own compute free and profitable. Payola is bigger than both. - **The line the piece turns on.** Amazon's defense is that it could secretly charge the maximum when the closing price fell below "the true market value of the ad placement." Doctorow's rebuttal is that the closing price of an auction *is* the definition of true market value — which concedes the mechanism is not price discovery but willingness-to-pay extraction. - **Bottom line.** A tight, well-sourced argument that the most profitable thing Amazon does is a bribery market for search results, and that it was caught cheating at its own bribery market. The economics are sound; the political prediction at the end is speculation and reads as such. ## 📖 Main Content - **The three-business framing does real analytical work.** Doctorow sets up Amazon as fulfillment, AWS, and payola, and the sequencing matters. Fulfillment is described as more than fully subsidized by platform-seller fees, so shipping costs Amazon nothing — with the cost noted in passing as falling on workers' "labor rights, bodies and bathroom breaks." AWS turned Amazon's own compute from a cost centre into a profit centre funded partly by competitors renting the same servers. The rhetorical setup is that both are extraordinary, and then payola is bigger than either. - **The payola analogy is the piece's central move and it is earned rather than decorative.** He spends four paragraphs on the 1950s radio scandal before applying it, and the details he selects are the load-bearing ones: radio stations held a genuinely scarce public resource (spectrum) under an expectation of serving the public, the selection process was opaque, and the opacity plus the stakes made corruption inevitable. The harms he enumerates are notably not limited to listeners — musicians lost royalties because labels charged the bribes to "promotions" and deducted them from artist statements, stations degraded, and the labels themselves bid away their own margins. - **His framing of why it persisted is the transferable part.** "Concentrated gains and diffuse losses" is offered as the reason payola ran for decades before anyone acted, and he applies the same phrase to Amazon later. That is a general mechanism for why extractive arrangements survive: the beneficiary has enormous incentive to defend it and each victim has too little at stake to fight. - **The mechanism, laid out.** Amazon search results are auctioned rather than ranked by match quality. Sellers raise prices to fund the bribes. A Most Favored Nation clause requires them to match or beat their Amazon price everywhere else — so the increase propagates to Target, Walmart, and the corner shop. Doctorow's conclusion is that Amazon therefore levies an economy-wide tax on nearly every product, and he supports the inescapability claim with Prime's capture of most of the richest decile of American households. - **The AI detail is small and genuinely new.** He notes that AI made Most Favored Nation *enforceable* in a way it previously was not — parsing competitors' websites at scale to detect discounting, with instant punishment by burial in search results. That is a concrete example of AI strengthening an existing anticompetitive mechanism rather than creating a new one, and it is the kind of specificity this genre usually lacks. - **The numbers he offers on ranking quality.** The first item on a typical results page is 29% more expensive than the best match; the top row averages 25% more; the best result sits around position 17, on the second screen. These are sourced to his own earlier writing rather than to a study within this piece, which is worth noting before repeating them. - **The per-unit pricing observation is the best small argument in the piece.** Amazon has many comparison tools and conspicuously lacks per-unit pricing. His evidence that this is deliberate rather than an oversight is elegant: Amazon-owned Whole Foods puts per-unit pricing on its shelf tags, so the company demonstrably understands why shoppers want it. And the mechanism is clear — top bidders package in irregular quantities, so sorting by price does not reveal the cheapest per item. A per-unit sort would let searchers route around the auction entirely. - **The revenue trajectory.** Mid-$30 billions during the research for *Enshittification*, more than $50 billion a year later when touring with Tim Wu, and on track to top $80 billion this year. He is explicit that these are figures he learned at different times rather than a single audited series, which is appropriate care. - **The Tim Wu comparison is the most quotable statistic.** Amazon's payola revenue exceeds the advertising revenue of all the newspapers in the world by 300%. Doctorow's use of it is pointed — against that number, Bezos buying the Washington Post and degrading its editorial page "barely registers." The argument is that the structural transfer of advertising revenue away from publications dwarfs any individual proprietor's editorial interference. - **The "you are the product" inversion.** He explicitly rejects the aphorism here, and correctly for this case: the advertisers *are* paying customers, and they are the ones being defrauded. That is a useful corrective to a slogan that has become a reflex. - **The auction explanation is unusually good popular exposition.** He explains the sealed-bid second-price auction from first principles — bidders submit true maxima, the winner pays second-highest plus a penny — and justifies it on two grounds. Open-call auctions are too slow for an eyeblink decision that must render a search page. And they cause winner overpayment: if you jump $10 to $50 you cannot know whether rivals would have stopped at $15, so you may have overpaid by $34.99. The SBSPA lets you name your true price while paying only what an open call would have produced. - **The allegation.** For seven years Amazon charged winners their own sealed bid instead of second-price-plus-a-penny, in roughly 80% of auctions. Ars puts the take at $20 billion; Doctorow characterises it as tens of billions extracted from sellers and passed to consumers, amplified economy-wide through AI-enforced MFN. - **Amazon's defense, and why Doctorow thinks it is self-defeating.** Amazon says sellers misunderstood: it operates a special SBSPA in which it may unilaterally and secretly charge the maximum where, in its judgment, the closing price fell below "the true market value of the ad placement." Doctorow's rebuttal is definitional — the closing price of an auction is what "true market value" means, which is precisely why auctions are treated as the canonical price-discovery mechanism and why auction design wins Nobel prizes. If Amazon can overrule the outcome, the auction is not discovering a price; it is a device for eliciting each bidder's maximum willingness to pay. - **That rebuttal is the strongest thing in the piece** and it is worth separating from the rhetoric around it. The argument does not require you to accept "enshittification" as a frame, or to share his politics. It is a clean internal contradiction in a stated legal defense. - **The political coda is speculation and is flagged loosely.** He suggests Amazon could escape via Trump's FTC — citing the Ticketmaster precedent where, on his account, a podcaster's tweet prompted presidential intervention — but notes the 22 state attorneys general as an independent path to trial. The joke about "Aspiring Governors" and the Nobel gag at the end are entertainment, not argument. - **Uncertainties.** The 80% figure and the seven-year span come from the complaint as reported, which is an allegation and not a finding. The $20 billion is Ars Technica's framing of the FTC's claim. The 29%/25%/position-17 numbers trace to Doctorow's earlier work rather than to a source in this piece. The 51–60% junk-fee range is asserted without a citation here. Amazon's defense is characterised from Ars's reporting rather than from a filing. ### 🗒️ Notes #### The Enshittocene framing > [!quote] > Amazon's own balance sheet presents the most compelling evidence that we are stuck in the Enshittocene, the era in which everything is turning into a pile of shit, because the worst ideas of the worst people now make the most money. - The thesis stated up front: the profitability ranking of Amazon's business lines is itself the evidence. #### Two cost centres turned into profit centres - Fulfillment: more than 100% of operating cost covered by platform-seller fees, so delivery costs Amazon nothing. Sellers compete with Amazon and carry delivery as a cost centre. - Worker cost noted in passing: labor rights, bodies, bathroom breaks. - AWS: built more servers than needed; customers, many of them competitors, fund Amazon's own compute and generate profit on top. > [!quote] > One of the largest corporations in the history of the world has converted its two largest cost centers into profit centers, and those profits are substantially generated by extracting payments from the company's own competitors! #### Payola, the 1950s original - Radio stations held exclusive slices of spectrum — a scarce public resource — under an expectation of serving the public. - Opaque selection plus high stakes made corruption structural. Envelopes of cash and drugs to DJs, managers, owners. - The predictor of a hit was label spend on bribes, not public preference. - Harms enumerated: worse music for the public; musicians' royalties eroded because bribes were charged to "promotions" and deducted from artist statements; degraded stations; labels bidding away their own margins and starving themselves of capital. > [!quote] > As with every instance of corruption, this was a system of concentrated gains and diffuse losses, which is why it continued for so long (decades!) and got so bad before anyone took action to end it. #### Amazon's version > [!quote] > Amazon's search results reflect who paid the biggest bribe, not who has the best product. - Sellers raise prices to fund bribes; Most Favored Nation forces the increase everywhere else. - Junk fees put at 51–60% of purchase price. Prime's capture of most of the richest 10% of US households makes Amazon unavoidable for sellers. - **AI as enforcement:** parses competitors' sites for discounting and triggers instant burial in search results. A concrete case of AI making an existing anticompetitive term enforceable. - Fulfillment fees carry the same coercion — not using Amazon logistics also buries your listings, which is why fulfillment is profitable. #### Ranking quality, quantified - First item on a results page: 29% more expensive than the best match. Top row: 25% more. Best result: around position 17, second screen. - Sourced to Doctorow's own earlier writing rather than to a study in this piece. #### The missing per-unit price - Amazon offers many comparison tools and no per-unit pricing. - Top bidders package in irregular quantities, so sorting by price does not identify the cheapest per item. - Amazon-owned Whole Foods puts per-unit pricing on shelf tags — evidence the company understands the need. - A per-unit sort would let searchers route around the auction, which is the stated reason for its absence. #### Scale - Mid-$30 billions during *Enshittification* research; over $50 billion a year later; on track to top $80 billion this year. - Tim Wu: Amazon's payola take exceeds the advertising revenue of all the world's newspapers by 300%. - Against that, Bezos buying the Washington Post "barely registers." #### The inversion > [!quote] > You know how people like to say, "If you're not paying for the product, you're the product?" It's bullshit. - The advertisers are paying customers, and they are the defrauded party. #### How a sealed-bid second-price auction works - Bidders submit their true maximum privately; the winner pays second-highest plus one cent. - Justified on speed — open-call auctions cannot resolve in the eyeblink a search page requires. - And on overpayment — jumping $10 to $50 risks paying $34.99 more than necessary if rivals would have stopped at $15. - The SBSPA lets you name a true maximum while paying what an open call would have yielded. #### The allegation - FTC and 22 state AGs, filed August 2026. - For seven years Amazon secretly charged winners their own sealed bid rather than second-price plus a penny. - Roughly 80% of auctions, per the suit. Ars Technica's figure: $20 billion. - Costs passed to consumers and propagated economy-wide by AI-enforced Most Favored Nation. #### The defense, and the rebuttal - Amazon: sellers misunderstood; it runs a special SBSPA permitting it to secretly charge the maximum where the closing price fell below "the true market value of the ad placement." > [!quote] > The definition of "true market value" is "the closing price in an auction." Amazon claiming that it secretly jacked people because the auction generated a price that was "below the true market value" of an ad tells you that the whole business is a sham. - The conclusion drawn: the scheme exists to extract each seller's maximum willingness to pay, and the fake SBSPA was the instrument for eliciting it. > [!quote] > Cheating on a bribery scheme is a mood. This isn't just enshittification, it's enshittification inception! #### The political coda - Filed by Trump's FTC; the Ticketmaster precedent offered as a route to escape. - 22 state AGs as an independent path to trial. - The "Aspiring Governors" joke and the Nobel gag close the piece as entertainment rather than argument. ## 📚 References - [FTC press release, August 2026](https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme) — read this before repeating any figure from the coverage - [Ars Technica on the $20 billion allegation and Amazon's defense](https://arstechnica.com/tech-policy/2026/09/ftc-alleges-amazon-illegally-made-20-billion-by-rigging-billions-of-ad-auctions/) - [Gizmodo coverage](https://gizmodo.com/ftc-sues-amazon-for-allegedly-duping-advertisers-2000805199) - [Vickrey auction](https://en.wikipedia.org/wiki/Vickrey_auction) and [auction theory](https://en.wikipedia.org/wiki/Auction_theory) — the mechanism and why economists treat it as price discovery - [Payola](https://en.wikipedia.org/wiki/Payola) — the 1950s analogy - [Doctorow on Most Favored Nation clauses](https://pluralistic.net/2026/02/25/most-favored-nation/#price-fixing) — the price-propagation argument in full - [Doctorow on subprime attention rent](https://pluralistic.net/2023/11/03/subprime-attention-rent-crisis/#euthanize-rentiers) — source of the 29%/25%/position-17 figures - Tim Wu, *The Age of Extraction* — the newspaper-advertising comparison - Related in the current reading queue: the WP Minute "AI Gold Rush" episode makes the same platform-capture argument from the practitioner side, and the Melapress plugin-review interview describes a marketplace where ranking is *not* purchasable and the constraint is reviewer capacity instead ## 🗣️ Paraphrase - **The claim.** Amazon's own profit ranking shows what the economy rewards: the most lucrative thing it does is sell search placement to the highest bidder. - **Two cost centres flipped.** Platform sellers more than fully fund Amazon's delivery network, and cloud customers — many of them competitors — fund its computing. Both became profit sources. - **Payola explained.** In the 1950s, labels bribed radio DJs to play records. The public got worse music, artists lost royalties to bribes booked as promotion costs, and labels destroyed their own margins bidding against each other. It lasted decades because the gains were concentrated and the losses spread thin. - **The modern version.** Amazon auctions the top of its search results. Sellers raise prices to pay, and a Most Favored Nation clause forces those higher prices into every other shop in the country. AI now makes that clause enforceable by detecting discounts and burying the offender. - **What it costs shoppers.** Junk fees run 51–60% of purchase price. The first result costs 29% more than the best match, and the best match is usually around the seventeenth position. - **The deliberate omission.** Amazon offers no per-unit price comparison, though its own Whole Foods puts per-unit prices on shelf tags — because a per-unit sort would let shoppers bypass the auction. - **The size.** Payola revenue has gone from the mid-$30 billions to a projected $80 billion-plus, and now exceeds the world's total newspaper advertising revenue by 300%. - **The auction.** Winners are supposed to pay the second-highest bid plus a penny, a design that lets bidders name a true maximum without overpaying. - **The allegation.** For seven years, in about 80% of auctions, Amazon charged winners their own maximum instead. - **The defense, and its problem.** Amazon says it may override the outcome when the closing price is below the ad's true market value. But an auction's closing price is what true market value means — so the defense concedes the auction was never discovering a price, only extracting one. ## 📰 Newsletter **Amazon got caught cheating at its own bribery market, and the defense is more damning than the allegation.** Cory Doctorow's Thursday piece walks through the FTC and 22-state suit filed in August. The claim: for seven years, in roughly 80% of its search-ad auctions, Amazon charged the winning bidder their own maximum sealed bid rather than the second-highest bid plus a penny — which is what a second-price auction is *defined* to do. Ars Technica puts the take at $20 billion. The setup matters as much as the allegation. Doctorow's framing is that Amazon converted its two largest cost centres into profit centres funded by its own competitors — platform sellers more than fully pay for the fulfillment network, and cloud customers make Amazon's own compute free and then some. And **payola is bigger than either.** Bigger than everything Amazon sells directly, bigger than AWS. Mid-$30 billions when he was researching *Enshittification*, over $50 billion a year later, on track to top $80 billion this year. Tim Wu's comparison is the one I'd keep: **Amazon's payola revenue exceeds the advertising revenue of every newspaper in the world by 300%.** Doctorow's point lands hard — against a structural transfer that size, Bezos buying the Washington Post barely registers. **Two details worth pulling out.** First, an AI angle that's specific rather than hand-wavy. Amazon's Most Favored Nation clause requires sellers to match or beat their Amazon price everywhere else, so bribe-funded price rises propagate to Target, Walmart, and the corner shop. That clause was always hard to *enforce* — and AI made it enforceable, parsing competitor sites at scale for discounts and instantly burying offenders in search. That's AI strengthening an existing anticompetitive mechanism, not inventing a new one. Second, the sharpest small argument in the piece: **Amazon has no per-unit price comparison.** Top bidders package in irregular quantities, so sorting by price doesn't tell you the cheapest per battery. Is that an oversight? Amazon-owned Whole Foods puts per-unit pricing on its shelf tags. They understand exactly why you'd want it. A per-unit sort would let you route around the auction entirely. **But the reason to read it is the defense.** Amazon says sellers misunderstood — it runs a special second-price auction where it may secretly charge the maximum if, in its judgment, the closing price fell below "the true market value of the ad placement." The closing price of an auction *is* the definition of true market value. That's the entire reason economists treat auctions as the canonical price-discovery mechanism and hand out Nobels for auction design. If you can overrule the result whenever you don't like it, you were never running price discovery — you were running a device to make bidders reveal their maximum willingness to pay. That argument doesn't require you to buy "enshittification" as a frame or share Doctorow's politics. It's a clean internal contradiction in a stated legal defense, and it's the part that survives being quoted to a skeptic. Worth reading alongside the WP Minute "gold rush" episode from earlier this week, which makes the same platform-capture argument from the practitioner's chair — delightful, then extractive — and the Melapress plugin-review interview, which describes a marketplace where ranking *isn't* purchasable and the binding constraint is ten volunteers instead. The contrast is instructive. Verify before repeating: the 80% and seven-year figures are allegations from a complaint, not findings. The 29%/25%/position-17 ranking numbers trace to Doctorow's earlier work, not to a study in this piece. Links: [the post](https://pluralistic.net/2026/09/04/cheating-at-fraud/) · [FTC release](https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme) · [Ars](https://arstechnica.com/tech-policy/2026/09/ftc-alleges-amazon-illegally-made-20-billion-by-rigging-billions-of-ad-auctions/) ## ✅ Action Items #### DevRel - Developer advocacy: the per-unit-pricing omission is a superb worked example of a missing feature being a business decision. Reusable in any talk about product incentives. - Education: the SBSPA explanation is genuinely good popular exposition of a mechanism most people encounter daily and cannot describe. Worth borrowing the structure. - Developer experience: not applicable to this source. - Developer marketing: "concentrated gains and diffuse losses" explains why extractive platform terms survive community objection. Useful framing when arguing for change. - Community management: not applicable to this source. - Cross-project: not applicable to this source. #### OSPO - Policy & Governance: any marketplace your organization depends on where ranking is purchasable has this failure mode latent. Worth naming in a dependency risk register. - License Compliance: not applicable to this source. - Security Management: not applicable to this source. - Software Inventory: platform dependency is a supply-chain concern even when no code is involved — pricing power is a dependency. - Developer Support: not applicable to this source. - Community Engagement: not applicable to this source. - Strategic Alignment: if a distribution channel takes 51–60% and can bury you for discounting elsewhere, channel strategy is a board-level question, not a marketing one. #### Community - Maintainers: relevant by analogy — any directory where visibility can be bought changes what gets built, not just what gets found. - Contributors: not applicable to this source. - Sponsors: not applicable to this source. - Ecosystem: the contrast with the WordPress plugin directory is the interesting part. Ranking there is not for sale; the constraint is reviewer capacity. Different failure mode, and worth being explicit that it *is* different. #### WordPress - Core: not applicable to this source. - Core Performance: not applicable to this source. - Core AI: the MFN enforcement example is a concrete case of AI making a previously unenforceable contract term enforceable — a pattern worth watching in any policy context. - **Core Program Team:** not applicable to this source. - Mobile: not applicable to this source. - Playground: not applicable to this source. - Community: the payola history is a good meetup talk on why opaque ranking systems corrupt. - Training: not applicable to this source. - Marketing: not applicable to this source. - Support: not applicable to this source. - Design: the absent per-unit sort is a design-ethics case study — a feature withheld because it would work. - Documentation: not applicable to this source. - Photos: not applicable to this source. - TV: not applicable to this source. - Polyglots: not applicable to this source. - Accessibility: not applicable to this source's subject matter. - Meta: plugin directory search ranking is a trust asset precisely because it is not auctioned. That is worth stating as a deliberate value rather than an accident. - CLI: not applicable to this source. - Hosting: not applicable to this source. - Tide: not applicable to this source. - Openverse: not applicable to this source. - Themes: not applicable to this source. - Plugins: see Meta. The directory's non-purchasable ranking is a differentiator worth defending explicitly. #### The WPOCC - Programs: not applicable to this source. - Projects: not applicable to this source. - Funding experiments: the diffuse-losses mechanism explains why individually-harmed maintainers rarely organize; relevant to any collective-funding design. ### FAIR - Technical independence: the piece is a long argument for why a single controlling intermediary is dangerous even when it started out useful. That is FAIR's premise with a different example. - Policies and covenants: anti-capture provisions exist for exactly the trajectory described here — a mechanism that is genuinely valuable becoming a rent extraction point once the controller has no competition. - Productized governance artifacts: a published commitment that ranking is not for sale is a governance artifact worth formalizing while it is still uncontroversial. ## 🧪 Hypotheses - Marketplaces that auction ranked placement show measurably higher median prices for equivalent goods than marketplaces ranking on relevance, controlling for category. - Adding a per-unit-price sort to a marketplace measurably reduces the revenue from ranked placement, which is why such sorts are rare where placement is sold. - AI-assisted price monitoring materially increased Most Favored Nation compliance rates over the past three years, and that increase is detectable in cross-retailer price dispersion. - The share of a platform's profit derived from ranked placement rises monotonically with its share of high-income households, because seller inescapability is what makes the auction bindable. - In sealed-bid second-price auctions run by a party with an interest in the outcome, deviation from the stated rule is the default rather than the exception, and 80% is not an outlier. - Publications' advertising revenue decline correlates more strongly with the growth of retail-media placement than with general display-ad shifts. ### 📊 Methodology - **Price comparison.** Match identical SKUs across a placement-auctioning marketplace and a relevance-ranked one; compare median and dispersion controlling for category, seller, and fulfillment method. - **Per-unit sort effect.** Natural experiment where a marketplace has added or removed per-unit sorting; measure change in sponsored-placement revenue and in the price-per-unit of top-ranked results. - **MFN enforcement.** Measure cross-retailer price dispersion for multi-channel sellers over time; a tightening distribution would be consistent with improved detection. - **Auction integrity.** For any platform running second-price auctions, compare charged prices against second-highest bids — the FTC's own approach, and it generalizes. - **Revenue substitution.** Time-series comparison of retail-media revenue growth against publisher advertising revenue decline, testing Wu's substitution claim directly. - **Data sources.** The FTC complaint and any discovery released; retailer price scrapes; publisher revenue filings; platform investor disclosures. - **Limits.** This is a blog post arguing from a complaint. The complaint is an allegation. Several figures come from the author's own prior work and one from a book cited in passing. Treat all of it as a hypothesis generator. ### ❓ Questions - What does the FTC complaint actually allege, in its own words, versus how the coverage characterises it? - Is the 80% figure a finding, an estimate, or a range? Doctorow reports it as flat, and complaints rarely are. - Where do the 29%, 25%, and position-17 numbers originate, and what was the sample? - Is the 51–60% junk-fee range Amazon-wide, category-specific, or drawn from a particular seller survey? - Does Amazon's contract language actually contain the "true market value" carve-out, or is that a litigation position constructed afterward? - If the state AGs proceed independently, what remedy is plausible — disgorgement, an injunction on the auction mechanism, or structural relief? - What is the honest disanalogy between this and a directory like WordPress's, where ranking is not sold? Naming it precisely is more useful than the parallel. ### 🔮 Future Research - Auction integrity as an audit discipline: whether platforms running second-price auctions can be independently verified without discovery, and what a verifiable auction protocol would look like. - Retail media's effect on consumer prices, which is a large and under-measured transfer with almost no independent literature. - The enforceability effect of AI on contract terms generally — MFN is one instance, and price-fixing-adjacent terms that were previously unenforceable at scale are a category worth studying. - Whether per-unit price disclosure requirements, standard in physical retail in many jurisdictions, could or should extend to online marketplaces. - Comparative marketplace governance: what distinguishes directories that resist selling ranking from those that do not, and whether the difference is structural or merely historical. ### 💡 Implications - **Practice.** If you sell through a marketplace that auctions placement, the auction rules are a term of your business you cannot verify. That is worth pricing into channel decisions rather than discovering in a complaint. - **Policy.** Per-unit price disclosure is mandatory in physical retail in many places and absent online. That is a small, concrete, achievable regulatory ask with a clear consumer benefit, and this piece makes the case for it almost incidentally. - **Security.** Not applicable to this source, except by analogy: an unverifiable mechanism operated by an interested party is a trust problem regardless of domain. - **Education.** The SBSPA explanation is genuinely useful for anyone who wants to understand how ad markets work, which is most people who use the web and almost none who can describe it. - **Funding.** The publisher-revenue substitution argument is the structural context for why independent media funding is hard, and it is more explanatory than most accounts of the same decline. ### 🧠 Learning Styles Assessment - Reading level estimate: Flesch-Kincaid roughly Grade 11–12. Long sentences with heavy subordination, economics vocabulary (price discovery, most favored nation, cost centre) mostly defined in context, and profanity used deliberately for emphasis. #### 👀 Visual learners - Present: one illustrative header image; no diagrams, charts, or tables. - Improve: the auction mechanism is the clearest diagram opportunity in the piece — three bidders, sealed bids, correct charge versus alleged charge. A single graphic would carry the whole argument. #### 🔊 Aural learners - Present: Doctorow publishes an audio edition of Pluralistic, though it is not linked in the captured content. - Improve: link the audio version from the post body; the prose has a strong spoken cadence and reads well aloud. #### 🗣️ Verbal learners - Present: excellent, and the piece's strength. Terms are defined as introduced, the payola analogy is built before it is used, and the auction explanation proceeds from first principles. - Improve: nothing substantive. #### 🧩 Kinesthetic learners - Present: minimal — there is no action offered to the reader. - Improve: "sort your next Amazon search by unit price and notice you cannot" would turn the strongest observation into something the reader verifies personally in thirty seconds. #### 🧠 Logical learners - Present: strong. The three-business structure, the analogy, the mechanism, the allegation, and the definitional rebuttal build in a defensible order, and the rebuttal is a genuine internal contradiction rather than an assertion. - Improve: several load-bearing figures are unsourced within the piece and trace to the author's own prior writing. Inline citation for the 51–60% and the ranking percentages would close the gap. #### 🧑‍🤝‍🧑 Social learners - Present: the newsletter format has an established readership and the piece links extensively into an ongoing conversation. - Improve: no comments or discussion mechanism in the captured content. #### 🧘‍♀️ Solitary learners - Present: excellent. Self-contained, densely linked for follow-up, and readable in a sitting. - Improve: nothing substantive. #### 🌿 Naturalistic learners - Present: the "Enshittocene" coinage borrows geological-epoch framing deliberately. - Improve: the environmental cost of the logistics network the piece describes goes entirely unmentioned, and the worker-body cost gets half a clause. ### 📊 Accessibility Assessment #### Screen reader semantics - Present: good. The captured markup shows proper paragraph structure, real heading elements, genuine link elements with descriptive surrounding text, and a navigational list at the top with named anchors (Object permanence, Upcoming appearances, Colophon). - Improve: many links are bare URLs used as their own anchor text, which screen readers announce character by character. Descriptive anchor text would be a substantial improvement given the link density. #### Keyboard navigation - Present: the anchor-list navigation at the top is a genuine keyboard convenience for a long single-page post. - Improve: not evaluable further from the captured content. #### Contrast and readability - Present: not evaluable from the capture. Pluralistic's design is conventionally text-first with minimal chrome. - Improve: verify inline link contrast against body text at the live site. #### Text resizing and reflow - Present: not evaluable from the capture; the layout is simple enough that reflow is unlikely to be a problem. - Improve: verify at 200% zoom. #### Media and captions - Present: one header image with an attribution line naming the photographer, the licence, and the fact of modification — better sourcing practice than most publications manage. - Improve: whether the image carries meaningful alt text is not verifiable from the capture. Given it is decorative here, empty alt would be correct. #### Forms and interactions - Present: none in the article body. - Improve: not applicable. #### Standards reference - WCAG 2.2 AA notes: 1.3.1 Info and Relationships well met through genuine heading and paragraph structure. **2.4.4 Link Purpose is the real weakness** — bare-URL anchor text appears dozens of times and conveys nothing out of context. 1.1.1 Non-text Content unverifiable for the single image. 2.4.5 Multiple Ways met via the anchor navigation. ### 🌍 DEIB Assessment #### Cultural breadth - Present: limited by design. The analysis is US-centric — American radio history, American income distribution, the FTC, state attorneys general — which is appropriate to the subject. - Improve: Amazon is a global marketplace and Most Favored Nation propagation would work differently under EU or UK competition law. One sentence acknowledging the jurisdictional boundary would strengthen rather than dilute the argument. #### Inclusive language - Present: no exclusionary language toward any group. - Improve: sustained profanity and personal invective ("sloshing, ambulatory pile of hot liquid garbage," "chud podcaster") will keep this out of professional contexts where the substance would otherwise be welcome. That is a deliberate stylistic choice with a real distribution cost. #### Neurodiversity support - Present: strong. Short paragraphs, one idea at a time, consistent structure, and the anchor navigation at the top. - Improve: the closing sections blend appearance listings and link roundups with the argument, which makes the end of the piece hard to parse as a boundary. #### Representation - Present: the harms described include Amazon warehouse workers, platform sellers, musicians in the historical analogy, and consumers — a genuinely broad account of who bears the cost. - Improve: the worker cost is a half-clause ("labor rights, bodies and bathroom breaks") in a 3,500-word piece, and no affected person is quoted. Sellers are discussed at length and never heard from. #### Psychological safety - Present: not a significant dimension for this genre; the piece is analysis rather than anything requiring pastoral care. - Improve: the register is unrelentingly angry, which is authentic to the author and can be wearing at length. Readers already anxious about economic precarity get a well-argued case that things are bad and no available action. ### ✅ Logical Validity #### Strong reasoning - **The definitional rebuttal.** Amazon's defense asserts a closing price can be below true market value; auction theory defines true market value as the closing price. Therefore the defense concedes the auction was not performing price discovery. This is a genuine internal contradiction and it is the piece's strongest move. - **The analogy is properly constructed.** Payola is not asserted as similar — the structural features are established first (opaque selection, high stakes for bidders, concentrated gains, diffuse losses) and then shown to hold in the new case. That is how an analogy earns its conclusion. - **The per-unit pricing inference.** Premise: Amazon operates Whole Foods, which displays per-unit pricing. Therefore Amazon understands the demand. Therefore its absence online is a choice, and the mechanism explains why. Well-formed abductive reasoning with a genuinely clever piece of evidence. - **The cost-centre-to-profit-centre observation.** Factually grounded and non-obvious: the profits come substantially from the company's own competitors, which is what makes the position defensible against entry. - **The "you are the product" inversion.** Correctly identifies that the slogan does not apply where the advertiser is the defrauded party, which is a real limit of a widely-repeated aphorism. #### Fallacies - **Unsourced statistics within the piece.** *Location:* the 51–60% junk-fee range; the 29%/25%/position-17 ranking figures. *Weakness:* presented flatly, sourced only to the author's own earlier posts, with no methodology stated in either place visible here. *Fix:* cite the underlying study; the numbers are load-bearing for the consumer-harm claim. - **Allegation reported as established fact.** *Location:* "Over the past 7 years, Amazon has been secretly charging..." *Weakness:* stated in the indicative although it is an unproven allegation in a complaint; the qualifier "according to the suit" appears one sentence later, applied only to the 80% figure. *Fix:* attribute consistently. The argument does not weaken if the allegation is framed as one. - **Ad hominem in the coda.** *Location:* "sloshing, ambulatory pile of hot liquid garbage," "chud podcaster." *Weakness:* adds nothing to a case that is already made and narrows the audience able to receive it. *Mitigation:* clearly signalled as invective rather than argument, and confined to the closing paragraphs. - **Speculative political prediction.** *Location:* the suggestion Amazon can escape by having a podcaster tweet at the president. *Weakness:* one prior instance generalized into a mechanism, with no assessment of how the cases differ. *Mitigation:* he immediately names the 22 state AGs as an independent path, which is the substantive point and undercuts his own cynicism. - **Loaded framing as premise.** *Location:* "bribe" used throughout for a legal advertising transaction. *Weakness:* the moral conclusion is embedded in the vocabulary before the argument for it is complete. *Mitigation:* he does argue for the equivalence via the payola analogy rather than merely asserting it, so the framing is earned rather than smuggled — but a reader who rejects the analogy is given no neutral term to think with. - **Selective harm accounting.** *Location:* the case throughout. *Weakness:* no counter-consideration is offered — that ranked placement might fund lower headline prices, or that sellers derive value from the traffic. *Fix:* addressing the strongest defense of retail media would make the piece more persuasive to anyone not already convinced. #### Tone of voice - Angry, erudite, and deliberately profane — the established Pluralistic register. The structure is disciplined even where the language is not: exposition is patient and genuinely instructional for two-thirds of the piece, and the invective is quarantined in the closing paragraphs. Doctorow is careful in places one might not expect, hedging the revenue figures as things he learned at different times and flagging the political prediction as speculation. The cost of the register is distribution — a well-made argument that cannot be forwarded in many professional contexts. ### 🛡️ DARVO Analysis - **Deny:** Not present in the author's voice. Notably, the *subject's* behavior as reported is a denial — Amazon's "sellers misunderstood how our auction works" defense is a textbook minimization, recasting alleged fraud as a comprehension failure on the victims' part. Doctorow identifies and dismantles it rather than reproducing it. - **Attack:** Present in the author's voice, in the closing paragraphs, directed at Amazon's unidentified scheme designer, at a podcaster, and at economics-prize institutions. This is polemic rather than a defensive structure, and it follows the argument rather than substituting for it. - **Reverse victim/offender:** Not present in the author's voice. Worth flagging that Amazon's reported defense contains the seed of it — positioning itself as misunderstood by sellers who failed to read the terms, which relocates fault onto the alleged victims. - **Assessment:** DARVO is not the author's strategy. The more interesting finding is that the piece *documents* a deny-and-reverse move by its subject: sellers misunderstood, and the price they agreed to was below true value, so charging them more was justified. Doctorow's response is definitional rather than rhetorical, which is the right way to answer it. ## 🎯 Conclusion - **Bottom line.** Search placement is Amazon's largest profit centre, on track to exceed $80 billion this year, and the FTC and 22 states allege the company spent seven years cheating the auction that sets its price — charging winners their own maximum instead of the second-highest bid, in roughly 80% of cases. - **The argument that survives quoting to a skeptic.** Amazon's defense says it could override a closing price that fell below "true market value." An auction's closing price is the definition of true market value. The defense therefore concedes the mechanism was never price discovery — it was a device to make bidders reveal their maximum. - **The detail worth stealing.** Amazon has no per-unit price sort, and Amazon-owned Whole Foods puts per-unit pricing on its shelf tags. The company understands the demand; the omission is a business decision, because a per-unit sort would let shoppers route around the auction. - **The AI specific.** Most Favored Nation clauses were always hard to enforce. AI made them enforceable at scale — detect a discount anywhere, bury the listing instantly. That is AI hardening an existing anticompetitive term rather than creating a new one, and it is the sharpest technology observation in the piece. - **Practical next steps for readers.** Read the FTC release before repeating any figure; several numbers here trace to the author's own prior posts rather than to a study. Try sorting your next marketplace search by unit price and notice you cannot. And if you sell through any marketplace that auctions placement, treat the auction rules as an unverifiable term of your business — because that is what this case establishes they are. ## 🔄 Refactor Appendix — Amazon achieves enshittification inception ### 📚 Literature Atoms (H3) #### [H1] An auction's closing price is the definition of market value > [!quote] > The definition of "true market value" is "the closing price in an auction." Amazon claiming that it secretly jacked people because the auction generated a price that was "below the true market value" of an ad tells you that the whole business is a sham. **Context & use**: - The rebuttal to Amazon's stated legal defense, near the end of the piece. - A clean internal contradiction: if the operator may override the outcome, the mechanism is not discovering a price but eliciting a maximum. Survives being quoted to someone who rejects everything else in the piece. **Tags**: #xref #articleseed #aphorisms <!-- atomicity: 1 idea ✓ --> #### [H2] Corruption persists on concentrated gains and diffuse losses > [!quote] > As with every instance of corruption, this was a system of concentrated gains and diffuse losses, which is why it continued for so long (decades!) and got so bad before anyone took action to end it. **Context & use**: - Said of 1950s radio payola, then applied to Amazon's search auction. - The general mechanism for why extractive arrangements survive objection: the beneficiary is highly motivated to defend it and no individual victim is motivated enough to fight. **Tags**: #xref #aphorisms <!-- atomicity: 1 idea ✓ --> #### [H3] A missing comparison feature can be the business model > [!quote] > Amazon clearly understands why shoppers would want to compare unit pricing, but offering a per-unit sort option to its search would make the bribery racket a lot less effective. **Context & use**: - Evidenced by Amazon-owned Whole Foods displaying per-unit pricing on shelf tags. - The cleanest available example of a feature withheld precisely because it would work. Generalizes to any interface where the absent affordance is the one that would let users route around the revenue mechanism. **Tags**: #xref #articleseed <!-- atomicity: 1 idea ✓ --> #### [H4] AI made an unenforceable contract term enforceable > [!quote] > AI has made Amazon much better at enforcing Most Favored Nation terms, because AI is actually pretty good at parsing competitors' websites and finding instances of discounting, which Amazon instantaneously punishes. **Context & use**: - Mid-piece, explaining how bribe-funded price rises propagate to every other retailer. - A specific, checkable instance of AI strengthening an existing anticompetitive mechanism rather than creating a novel one — a more useful class of claim than most AI-harm arguments. **Tags**: #xref #articleseed <!-- atomicity: 1 idea ✓ --> ### 🧠 Permanent Candidates (H4) #### [P1] A mechanism its operator can override is not a mechanism **Claim** (one sentence, Courtney's voice): When the party running a process reserves the right to set aside its output whenever the output disfavors them, the process is theatre and its result carries none of the authority the process was chosen for. **Evidence pointer**: [[#H1]] **Scope & conditions**: Holds wherever a procedure's legitimacy comes from its impartiality — auctions, elections, peer review, automated ranking, dispute resolution. Breaks down where an override is disclosed, bounded, and appealable, which is what distinguishes a discretionary exception from a sham. **Links**: - [[A trust regime that exempts its administrator is not a trust regime]] — because extends the same principle to certification and audit - [[A control that does not match the failure mode buys compliance, not safety]] — because contrasts on procedures adopted for appearance rather than effect - [[Compliance and refusal are both choices, and only refusal reveals which]] — because applies-in the infrastructure-discretion question <!-- atomicity: 1 idea ✓ --> #### [P2] The affordance a platform withholds tells you where its revenue comes from **Claim** (one sentence, Courtney's voice): Look for the obvious comparison or export feature that is conspicuously missing, and you have found the mechanism the platform is protecting. **Evidence pointer**: [[#H3]] **Scope & conditions**: Holds where the platform monetizes friction or placement and the missing feature would dissolve it — per-unit sorting, data export, cross-provider comparison, unranked listing. Weakest where the omission is genuinely a cost or complexity decision, so the test is whether the operator demonstrably provides it elsewhere. **Links**: - [[Silence in a source is filled by the audience's prior, so name what is absent]] — because extends the discipline of reading omissions - [[What a friendly interviewer omits is the finding]] — because contrasts on the same inference in a different domain - [[FAIR technical independence]] — because applies-in the argument for portability <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Read the FTC complaint directly before repeating the 80% or the seven-year span — both are allegations reported in the indicative #fleeting #inbox - [F2] Trace the 29%/25%/position-17 ranking figures to their underlying study; they carry the consumer-harm claim and are sourced only to Doctorow's earlier post #fleeting #inbox - [F3] Write the contrast note: WordPress plugin directory ranking is not purchasable and the constraint is reviewer capacity — name the disanalogy precisely rather than leaving the parallel loose #fleeting #inbox - [F4] Per-unit price disclosure is mandatory in physical retail in many jurisdictions and absent online — small, concrete regulatory ask worth tracking #fleeting #inbox - [F5] Pairs with the WP Minute gold-rush episode on platform capture; both describe delightful-then-extractive, from opposite ends #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Read the FTC press release and note which figures are findings versus allegations before citing this anywhere #things - [ ] Draft the "a mechanism its operator can override is not a mechanism" permanent note and link it to the plugin-audit thread #things <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Pluralistic Amazon Achieves Enshittification Inception.md|Pluralistic: Amazon Achieves Enshittification Inception by Cory Doctorow]] ## Highlights - AI has made Amazon *much better* at enforcing Most Favored Nation terms, because AI is actually pretty good at parsing competitors' websites and finding instances of discounting, which Amazon instantaneously punishes by relegating the sellers' product listings to page umpty-billion of Amazon's search results. ([View Highlight](https://read.readwise.io/read/01m1p9fbt0tb2jbrj2xjb8afnx)) - Note: The sharpest technology observation in the piece: AI hardening an existing anticompetitive term rather than creating a new one. MFN was always hard to enforce; detection at scale is what made it bite — and it's the mechanism that propagates Amazon's bribe-funded price rises to every other retailer. - Tags: atom/literature/h3 - Amazon actively helps its biggest bribers close the sale. Amazon has *lots* of "comparison shopping" systems built into the service, but one comparison tool is conspicuous by its absence: an "apples to apples" tool that lets you compare *unit* prices. Amazon's most prolific bribe-payers package their goods in weird quantities, selling everything from batteries to t-shirts to shampoo in larger or smaller quantities than their competitors. Sorting your search results by price doesn't actually tell you who's got the cheapest price *per item*, because the company with the cheapest AA batteries might be selling a smaller *quantity* of batteries at a higher price *per battery*. ([View Highlight](https://read.readwise.io/read/01m1p9f1qptpgvbv0rjn9dwpjh)) - Note: The affordance a platform withholds tells you where its revenue comes from. Proof it's deliberate rather than an oversight: Amazon-owned Whole Foods puts per-unit pricing on its shelf tags. A per-unit sort would let shoppers route around the auction entirely. - Tags: atom/literature/h2, atom/zettel/z1 - The *definition* of "true market value" is "the closing price in an auction." Amazon claiming that it secretly jacked people because the auction generated a price that was "below the true market value" of an ad tells you that the whole business is a sham. The point of Amazon's payola scheme is only and ever a way to parasitically extract the maximum amount a platform seller is willing to part with, and by running a fake SBSPA, Amazon was able to trick its customers into revealing those maximum prices. ([View Highlight](https://read.readwise.io/read/01m1p9ep7tk10sxn2w13wg70g4)) - Note: The one argument that survives being quoted to a skeptic — it needs no agreement with "enshittification" or the politics. A clean internal contradiction in a stated legal defense: if the operator can override the closing price, the auction was never price discovery, only a device to elicit maximum willingness to pay. - Tags: atom/zettel/z1, atom/literature/h1