The algorithm sets the price: how major U.S. newspapers personalize subscription offers — and what the Washington Post lawsuit means for the industry

A class-action lawsuit filed this month in District of Columbia Superior Court accuses The Washington Post of using subscribers' personal data — location, browsing behavior, demographics — to set individualized subscription prices without disclosure.

The Post is not alone, however. A review of public disclosures, earnings calls, industry research and privacy policies shows that at least three of the four major newspapers named for comparison — The New York Times, The Wall Street Journal and Gannett's USA Today network — operate dynamic, machine-learning-driven paywalls that influence what readers see, when they hit a paywall and, in at least two documented cases, what they are asked to pay. Bloomberg, by contrast, appears to maintain a flat-rate hard paywall.

With New York's Algorithmic Pricing Disclosure Act now in force, Maryland's Predatory Pricing Act on the governor's desk, a California Attorney General sweep under way and a Federal Trade Commission inquiry active, the legal ground beneath personalized news pricing is shifting fast — and the Post case may be only the opening salvo.

How each outlet's paywall actually works

The New York Times — the most documented dynamic paywall in U.S. news

The Times has been the most transparent — at least in industry forums — about running a machine-learning paywall. Engineering posts on the company's NYT Open blog describe a "Dynamic Meter" that uses machine learning to set personalized article limits for each reader before a subscription prompt appears. Hannah Yang, then senior vice president of consumer revenue, told industry audiences the dynamic paywall lets the company "tailor digital subscription offers to different subsets" of readers.

The financial impact is meaningful. On the company's fourth-quarter 2020 earnings call, executives told analysts the ML model directs roughly 80 percent of new subscribers through a step-up pricing path that graduates them from introductory rates ($1 a week, $4 a month) to a standard $25 every four weeks, outperforming a random-sample control. Press Gazette later reported that 1.6 million subscribers graduated to full price in 2021, with roughly one million continuing to graduate each subsequent quarter.

The pricing language in the Times' own privacy policy is unusually direct. The company's privacy policy, effective May 27, 2026, tells readers:

"We also show you prices, promotions, products or services we believe you'll find interesting, based on demographic and usage data, including but not limited to location data such as IP address and zip code and, if you share it with us, your precise geolocation."

And, separately:

"For example, our analysis, which includes the use of technology like machine learning and large language models (proprietary or third party), lets us predict preferences and price points for our products and services."

That is the closest thing to a plain-English personalized-pricing disclosure currently published by a major U.S. newspaper. It does not, however, satisfy the on-screen point-of-sale disclosure label that New York's new algorithmic pricing law requires when an individualized price is presented.

The Wall Street Journal — propensity scoring at 15 million predictions a week

The Journal's parent, Dow Jones, has built what trade publications describe as one of the most aggressive propensity-pricing systems in news. A Press Gazette interview with Karl Wells, then general manager of membership, laid the model out plainly. Wells told the publication: "Our paywall is a propensity-led model, so [the] level of access is determined by an algorithm that understands your likelihood to buy."

He compared reader segments to paint cans — "cold, warm and hot, like Ronseal" — and confirmed the Journal uses "propensity to pay" scores to decide which subscription offer a reader sees, with "multiple variations out in market" at any time.

A separate Media Makers Meet account detailed the inputs: roughly 60 variables, including visit frequency, recency, content depth, device type and content category, feeding an ML model that runs more than 15 million predictions a week.

Real-world pricing variance bears this out. A pricetimeline.com analysis documented Journal annual rates ranging from $52 to $238 a year for ostensibly identical digital access, with introductory $1-a-week offers stepping up to $38.99 a month after 12 months — a 224 percent jump.

Dow Jones' privacy notice, effective Jan. 30, 2026, authorizes the company to: "Monitor and analyze the extent of your use of the Dow Jones Services, to inform modifications, enhancements, or updates to the Dow Jones Services, or to make marketing or pricing decisions." That phrase — "marketing or pricing decisions" — is, alongside the Times' language, the most explicit pricing-personalization disclosure currently in a U.S. newspaper privacy policy.

USA Today / Gannett — AI decisioning, surging ARPU, vague disclosure

Gannett executives told investors in late April that the company is now leveraging AI-driven personalization at the paywall layer. On the company's first-quarter 2026 earnings call, reported by MarketBeat, management said: "We are increasingly leveraging AI-driven personalization, combining dynamic paywall decisioning with personalized for you placements."

The financial signal is sharp. Digital subscription revenue rose to $45.9 million, up 6.2 percent year over year, while average revenue per user hit a company record of $10.30, a 42.7 percent annual jump — a figure that suggests far more than across-the-board price increases. Kristin Roberts, president at USA Today, told the call the company is testing "multiple pathways," including free access, registration walls, article-level and topic- or season-based offers.

Gannett's USA Today privacy policy, as fetched, does not contain explicit personalized-pricing disclosure language analogous to what the Times and Dow Jones now publish. Public subscription pages show standard tiered offers — $1 for the first year as an introductory rate, $9.99 a month for Daily Plus — without documented per-reader variation in the public record. That gap between operational reality (dynamic paywall decisioning acknowledged in SEC-reported earnings remarks) and consumer-facing disclosure is the kind of mismatch plaintiffs' firms tend to seize on.

Bloomberg — the outlier

Bloomberg, alone among the four named outlets, appears to operate on a flat-rate model. An Innovation Media industry report classifies Bloomberg as a "hard paywall" publisher, and the company's own media press release describes a tiered pricing structure — Digital at $34.99 a month, All Access at $39.99 a month, with a $9.99 introductory rate for the first six months — without language suggesting per-reader variation. Annual pricing is similarly fixed, with promotional rates around $129 and standard rates of $299 for Digital and $399 for All Access.

Bloomberg's terms of service reserve eligibility for introductory offers "without prior notice" but contain no algorithmic, personalized or dynamic pricing language. The company's privacy policy does acknowledge that Bloomberg "infer[s] and create[s] a profile about a consumer reflecting the consumer's preferences and behavior," but the inferred-profile language is paired with marketing and product personalization uses, not pricing.

In short, of the four outlets named for this comparison, Bloomberg is currently the only one without public evidence of pricing personalization.

The Atlantic — useful industry reference point

Outside the four-outlet comparison, The Atlantic provides a useful benchmark for how far the industry's most aggressive adopters have gone. An analysis from Poool, a paywall technology vendor, describes The Atlantic's dynamic paywall presenting subscription prices ranging from $60 to $100 a year based on propensity scoring, content type and available ad inventory. Piano, the largest paywall vendor in news, has confirmed it is beta-testing dynamic pricing with five client publishers.

Privacy policy language audit

For editorial purposes, the language to compare side by side:

• New York Times:"show you prices, promotions, products or services … based on demographic and usage data, including but not limited to location data such as IP address and zip code and, if you share it with us, your precise geolocation"; uses "machine learning and large language models … [to] predict preferences and price points." (NYT Privacy Policy)
• Dow Jones / WSJ:uses service data "to make marketing or pricing decisions." (Dow Jones Privacy Notice)
• Gannett / USA Today:no explicit pricing-personalization clause identified. (USA Today Privacy Policy)
• Bloomberg:profiles preferences and behavior, but no pricing-decision language. (Bloomberg Privacy Policy)

The pattern is striking. The two outlets with the most sophisticated machine-learning pricing operations (the Times and the Journal) have published the most explicit disclosures. Gannett, which acknowledges dynamic decisioning to investors, has not extended that acknowledgment to consumers. Bloomberg, which has the least personalized model, also has the least pricing language — appropriately so.

Regulatory exposure analysis

The Washington Post case as precedent

The new D.C. class action,Chelsea Blink v. Washington Post Co., was filed June 11, 2026, in District of Columbia Superior Court by the Clarkson Law Firm. It alleges that the Post used readers' personal data — including geolocation, demographic inferences and browsing patterns — to set individualized subscription prices without disclosure, in violation of the D.C. Consumer Protection Procedures Act and unjust-enrichment doctrine. The complaint seeks treble damages or $1,500 per violation, exposing the Post to potential class damages estimated at up to $1.5 billion.

The Post case matters across the industry for three reasons:

1. It is the first U.S. newspaper-specific surveillance pricing suit to clear a court filing threshold.
2. It applies a general consumer-protection statute — not a pricing-specific one — meaning every state with a similar CPPA-style act becomes a potential forum.
3. It uses unjust enrichment as a fallback theory, which lowers the disclosure-violation bar plaintiffs must meet.

New York's Algorithmic Pricing Disclosure Act

New York's Algorithmic Pricing Disclosure Act took effect July 8, 2025. It requires any merchant offering a price set "by an algorithm that uses consumer personal data" to display the disclosure: "THIS PRICE WAS SET BY AN ALGORITHM USING YOUR PERSONAL DATA." It bans the use of protected-class data for discriminatory pricing and authorizes the attorney general to seek civil penalties of up to $1,000 per violation, per a JD Supra analysis.

Of the outlets in this comparison, three (NYT, WSJ, USA Today) plausibly trigger the statute when serving New York readers. None currently displays the mandated disclosure label at the point of subscription offer, based on publicly visible subscription flows.

Maryland, Connecticut and the state pipeline

Maryland's Predatory Pricing Act, which Gov. Wes Moore is signing, would be the first U.S. law to explicitly ban grocery surveillance pricing — but its definitional language on personal-data-driven pricing extends to other consumer transactions. Connecticut has enacted a similar measure. Bloomberg Law reports more than 70 bills pending across Hawaii, California, Illinois, Kentucky, Connecticut and other states.

California is the most active jurisdiction. Two bills are advancing: AB 446, which would ban surveillance pricing using geolocation, browsing history and behavioral inferences, and AB 325, which would prohibit "common pricing algorithm" agreements between competitors. California Attorney General Rob Bonta opened an investigative sweep in January 2026 targeting travel and retail surveillance pricing under the California Consumer Privacy Act, and news subscription pricing is a logical next vertical.

Federal exposure

FTC Chairman Andrew Ferguson confirmed in April 2026 Senate testimony that the commission's surveillance pricing inquiry remains active and that staff are weighing a policy statement on personalized pricing disclosures. A sub-inquiry into Instacart's pricing is open. House Oversight Chairman James Comer (R-Ky.) opened a parallel investigation in March 2026.

For news publishers, federal exposure is mostly downstream — an FTC policy statement or enforcement action against a non-news pricing personalizer would set persuasive precedent that plaintiffs' firms would cite immediately against publishers.

Exposure ranking

• Tier 1 (highest): The Wall Street Journal, owing to the explicitness of the "propensity to pay" model in trade reporting and the breadth of pricing variance documented in the field. Strong privacy-policy disclosure language reduces but does not eliminate state-law exposure.
• Tier 1 (highest): The New York Times, owing to its scale, the documented 80-percent step-up ratio, and the volume of New York-state readers subject to the algorithmic pricing disclosure law. Its explicit policy language is a partial defense but does not satisfy point-of-sale labeling.
• Tier 2 (high): USA Today / Gannett, owing to the gap between operational reality (acknowledged on earnings calls) and consumer-facing disclosure. Gannett is also the most geographically dispersed, exposing it to the widest set of state laws.
• Tier 2 (medium-high): The Atlantic and other Piano-platform publishers, by association.
• Tier 3 (low): Bloomberg, based on currently available evidence.

Bigger picture

The Washington Post lawsuit is being framed in early coverage as a privacy story. It is more usefully read as a disclosure story — a test of whether news publishers can keep operating sophisticated machine-learning pricing systems while telling readers, on the subscribe page, almost nothing about it.

Three of the four major newspapers named in this review are doing exactly that, with widely varying degrees of candor in their privacy policies and essentially no point-of-sale transparency.

The legal infrastructure to compel that transparency — New York's algorithmic pricing law, Maryland's predatory pricing act, California's pending AB 446 and AB 325, the FTC inquiry, and now the D.C. CPPA class action — is being built in parallel, in real time. The next 12 months will determine whether the industry leads with a voluntary disclosure standard or has one imposed on it case by case.

The more interesting question is no longer whether dynamic pricing works. It clearly does — Gannett's 42.7 percent ARPU jump and the Times' 80-percent step-up rate make that plain. The question is what publishers owe readers when the price on the screen was chosen for them by an algorithm trained on their behavior.

--

Perplexity.ai provided research for this story