AI is writing ADA lawsuits now. Here's what that changes
The barrier to filing a web accessibility complaint used to be hiring a lawyer. Now a plaintiff with a chatbot can surface violations and draft a complaint in an afternoon. We work these cases as expert witnesses, and from that seat the change is bigger than it looks.
September 23, 2026
2 minutes
The economics of filing just changed
For years the friction that kept accessibility lawsuits in check was cost: someone had to notice the barriers, understand WCAG, and pay a lawyer to write it up. Generative AI erodes most of that friction. Pro se plaintiffs now lean on chatbots to spot likely violations on a page and to draft the complaint itself, and the volume keeps climbing — federal web accessibility filings ran near 4,000 in 2025, and once you count state courts the total cleared 5,000.
When the cost of filing drops, two things follow: more filings overall, and smaller organizations drawn into a fight that used to target big brands. A local business that assumed it was too small to notice is exactly the kind of defendant this makes viable.
AI finds the easy stuff — which is what you are missing
And notice the uncomfortable symmetry. The violations an AI tool can surface from the outside are the automatable ones: missing alt text, low contrast, unlabeled form fields, buttons with no accessible name. Those are the exact failures that show up in almost every complaint — the same ones a basic audit would have caught years earlier. If a chatbot can find them from the curb, so can a plaintiff.
The lesson is not that AI is a sophisticated adversary. It is that the cheap, obvious problems are now cheap and obvious to everyone, and leaving them in place is a standing invitation.
What AI still cannot do — and where cases are decided
Automated tools, AI included, catch a fraction of what actually matters. Whether a blind customer can complete a real task — book the appointment, finish the checkout, recover from a form error — is a question of judgment that still requires a human using the site the way the plaintiff does. From the expert-witness seat, that judgment, and the documented history behind it, is where cases are genuinely won and lost, not in a scan-tool printout.
Which cuts both ways: an AI-drafted complaint that leans on automated findings is often thinner than it looks, and a defendant with real testing and real remediation records is in a far stronger position than one with neither.
The defense is the same as ever, just less optional
Nothing about the remedy has changed; the stakes for skipping it have. Build with semantic structure, test with actual assistive technology rather than a checker alone, and document the work in audits, VPATs, and remediation logs. That last part matters more than teams expect. Roughly 46% of 2025’s federal filings — about 1,400 cases — hit organizations that had already been sued once and left the underlying code broken. Getting named, settling, and never touching the markup is close to a standing invitation to the next plaintiff. We wrote about how these cases look from the inside in what ADA website lawsuits look like from the inside.