eCommerce checkout: The most litigated surface on the internet

A dataset of 8,788 federal web-accessibility lawsuits reveals an unflattering pattern. The same handful of bugs ship to production at the same handful of stages – and the same handful of plaintiffs find them. Of 81,509 issues attorneys catalogued in court filings, more than 4,050 are concentrated in the five surfaces that turn a browser into a buyer: cart, address, payment, error states, and the checkout button itself.

Why checkout is the violation that costs you

Web-accessibility lawsuits are filed at every layer of a website – homepage, navigation, footer, search. But on the rare occasions a complaint reaches dispositive motion, courts are unmoved by aesthetic barriers. They are moved by transactional ones. A blind shopper who cannot find a hero banner has, in the bluntest legal terms, lost nothing. A blind shopper who cannot complete a purchase has been denied a service the seller offered to every other visitor. That is the textbook formulation of a Title III violation, and it is the same fact pattern that recurs in every demand letter the plaintiff bar drafts.

This is why the dataset’s most-cited surfaces are not the surfaces that produce the most complaints overall. “General / Uncategorized” is by far the largest issue bucket (38,671 issues). But on the basis of dollars at stake per failure, the checkout funnel out-litigates every other category combined. Each unlabeled CVC field is a denial of service. Each unannounced “Order Confirmed” is a transaction the user cannot prove they made. Each modal that traps focus on the “Apply Coupon” step is a discount the screen-reader user paid full price to skip.

The dataset

The analysis behind this article draws on two independent corpora. The first is a structured PACER snapshot of 8,788 federal civil cases filed under the Americans with Disabilities Act and parallel state statutes between January 2007 and April 2026, with each case linked to its plaintiff(s), defendant(s), filing date, and disposition date. The second is a parallel extraction of 81,509 distinct accessibility issues pulled from the text of those filings – the actual sentences attorneys wrote describing what their clients could not do.

The two together let us answer questions a single source cannot. We know not just which sites were sued, but what attorneys told the court was broken – and at which point in the user journey it broke. For this article we filtered the issue corpus to five purchase-funnel categories: CHECKOUT FLOW SHOPPING CART ADD TO CART PAYMENT ADDRESS MANAGEMENT. Together these five surfaces account for 4,050 issues – about 5% of the corpus, but the slice with the strongest causal link to dispositive harm.

The curve nobody priced into the Q4 plan

Before discussing the inside of the funnel, it is worth establishing how steep the outside has become. Annual filings have grown by an order of magnitude since 2020 and continue to accelerate. The 2026 figure below covers only Q1; the full-year extrapolation places it on track to exceed 4,500 – almost ten times the 2021 baseline.

▱ 2026 partial: filings through April 2026. At the Q1 pace the year is on track for ~3,500 filings, in line with the 2025 trend.

Two points worth absorbing. First, the inflection year is 2021 – coinciding with the post-pandemic e-commerce boom and a wave of state-court precedent (notably Robles v. Domino’s, denied cert by the Supreme Court in 2019) that left federal courts in the Ninth Circuit reading websites into Title III’s “places of public accommodation.” Second, growth has not plateaued. Each year since 2021 has set a new high, and each new annual high has been hit while Title III has remained, formally, unregulated for private commercial sites.

The funnel within the funnel

Of the 81,509 issues catalogued from filings, 4,050 fall into one of the five purchase-funnel categories below. The proportional distribution maps almost perfectly onto the canonical conversion funnel: more issues are reported earlier in the journey, with each step shedding both legitimate users and litigated barriers.

The “Checkout Flow” bucket is the largest because most filings narrate the experience as a single journey. But the breakdown is instructive: the cart and add-to-cart together (1,740 issues) generate slightly more litigated complaints than the checkout page itself. That is partly because the cart is where the first irreversible action happens – once the screen-reader user cannot confirm an item was added, every step that follows is performed in the dark. Issues that begin upstream compound downstream.

The five failures, in order of frequency

When the 4,050 checkout-stage issues are pattern-matched against their plain-text descriptions, five failure modes account for the majority of plaintiff complaints. We surface them below in order of how often they appear in filings, with the WCAG criteria attorneys most commonly cite alongside each.

The criteria attorneys actually cite

Most filings do not cite WCAG criteria explicitly – they describe symptoms. Where attorneys do cite specific criteria in checkout-stage allegations, however, the distribution is heavily skewed. Six criteria account for 89% of all explicit citations; the remaining sixty-plus criteria of WCAG 2.2 are mentioned in long-tail noise.

What plaintiffs actually wrote

The dataset is built from court-filed pleadings. The sentences below are reproduced from those filings – anonymized to remove site references but otherwise unedited. They are useful because they communicate to engineering teams what users describe, in their own words, the way attorneys translated those descriptions into legal exhibits.

What unifies the dataset is not technical sophistication. The barriers attorneys cite are not novel or obscure – they are the same dozen patterns, repeated across thousands of filings, against thousands of defendants. The plaintiff bar has effectively industrialized the discovery of bugs that engineering teams have not noticed because the engineering teams have not run their own checkouts with a screen reader.

The 97-day phenomenon

Of the 8,788 cases in the dataset, 6,945 had closed by April 2026 with valid filing and disposition dates. Their durations form one of the more revealing distributions in the corpus. The median time from complaint to closure is 97 days; 83.4% of all closures occur in under six months.

Cases that close in under 180 days almost without exception do not produce a published merits ruling. They are settled. The shape of this distribution tells the strategic story of the entire plaintiff bar: file in volume, settle quickly, and avoid the small handful of dispositive opinions that would let defendants distinguish their checkouts from another company’s. The median settlement happens before either party files a motion to dismiss.

The volume model: ten plaintiffs, 1,289 cases

The headline number – 8,788 cases – is misleading without one further breakdown. The plaintiff side of the docket is heavily concentrated. The ten most-active named plaintiffs in the dataset, taken together, account for 14.7% of all filings. The most prolific single plaintiff filed 256 separate cases.

Plaintiff names redacted; counts derived from PACER metadata. Of 8,788 filings the top ten plaintiffs filed 1,289 cases combined (14.66%). The top single plaintiff appears as named party in nearly 1 of every 35 cases in the dataset.

This concentration is not a sign of bad faith – many of these plaintiffs have legitimate, documented disabilities and have personally encountered the barriers their filings describe. But it is a sign that defendants who lose in court rarely lose to a stranger. The same names recur, often represented by the same firms, often filing on near-identical templates. The structure of the litigation rewards efficiency on the plaintiff side and rewards capitulation on the defendant side. It does not reward novel arguments from either.

The triage matrix

If the question is “what should I fix first,” the answer is a function of two things: how often a given barrier is cited (frequency), and how directly it blocks a transaction (severity). The matrix below combines both. Cells are coloured by relative volume of citations in our 4,050-issue checkout corpus, with the darkest cells representing the failure modes most worth a sprint.

Two cells worth singling out. “Status not announced × Add-to-Cart” is the single most-cited barrier in the entire dataset; if you fix nothing else, fix this. “Unlabeled fields × Payment” is the textbook compliance failure – a credit-card form with announced-as-blank fields is the violation that cannot be argued.

What does this cost you, today, before any lawsuit?

Litigation exposure is one cost of an inaccessible checkout. The other – usually larger and always present – is the orders that never close because the buyer could not complete the form. The calculator below uses three inputs you can change to estimate that cost for your own funnel. Defaults are conservative, drawn from cross-industry baselines; adjust them to fit your business.

What an accessible checkout looks like, by contrast

The compare grid below pairs the failure modes most-cited in the dataset against the engineering pattern that resolves each. None of the right-hand patterns are speculative – every one of them is a documented WCAG 2.2 AA technique.

The thirteen-step remediation checklist

Drawn from the citation-density patterns in the dataset, the items below are sequenced in descending order of how often each appears in checkout-stage filings. Working top to bottom of this list addresses the great majority of allegations a serial-plaintiff complaint will plead.

Closing – the engineering question, restated

The thesis of this article – that checkout is the most-litigated surface on the internet – does not depend on the volume of filings, although the volume is striking. It depends on the structure of the user journey. Of all the surfaces on a commercial website, checkout is the one whose failure most precisely matches the legal definition of a denial of service, the one whose damages most cleanly translate into a complaint, and the one whose bugs are most consistent across defendants. None of those facts will change in 2027. What can change is the engineering decision that produces them.

The 4,050 issues this article catalogues are not, on inspection, mysterious. They are the same five or six failures shipping out of the same five or six components. The most efficient remediation a retailer can perform is not on a checkout page; it is on the design system that builds the checkout page. Fix the toast, the modal, the input, the error message, and the submit button – once, in the component library – and the dataset behind this article slims by something like an order of magnitude. The plaintiff bar will move on to a different surface. Your customers will not.