How to Survive the Pending AI Litigation Apocalypse

​Two researchers have just published a study warning that AI is about to unleash a wave of pro se lawsuits by self‑represented individuals, with particularly serious implications for businesses already targeted by predatory TCPA claims and for an already strained court system.

The paper, “Access to Justice in the Age of AI: Evidence from U.S. Federal Courts,” looks at roughly 4.5 million non‑prisoner federal civil cases from fiscal years 2005 to 2026 and about 46 million associated PACER docket entries. The following four headline findings matter the most for thinking about the TCPA:

• A Dramatic Rise in Pro Se Filings: For about two decades, pro se civil filings sat at a remarkably stable 11% of non‑prisoner federal civil cases, essentially unaffected by major shocks like the Great Recession or ACA litigation waves. After capable LLMs started to appear (ChatGPT in late 2022 and GPT‑4 in 2023), the nationwide non‑prisoner pro se share jumps to 16.8% in FY2025—an increase with no precedent in 25 years of administrative records.
  
• Concentration in Template-Friendly Cases: The pro se surge is concentrated in categories where the main task is drafting a structured factual narrative and plugging those facts into relatively standard legal frames: civil rights complaints, consumer credit disputes, and foreclosure actions. It does not show up in highly specialized, expert‑heavy domains like patent or securities cases. In other words, AI is expanding access exactly where a plausible “fill‑in‑the‑blanks plus story” complaint can get you into court.
  
• More Activity And a Higher Burden on Courts: Pro se cases are not resolving faster post‑AI, and their disposition mix (wins, losses, dismissals, settlements) looks surprisingly similar to the pre‑AI era. But inside each case, there is much more going on:
  
• Total pro se docket entries per court in the first 180 days of a case rise about 64% on average after the advent of capable LLMs (Q4 2022–Q2 2025).
  
• By 2025 Q2, the volume of these pro se entries in the first 180 days is 158% higher than the pre‑AI mean.

Every docket entry is a “claim on the court’s time,” so this is effectively an administrative burden issue, as well as a legal one.

• Direct Evidence of AI in Complaints: The authors apply an AI‑content detector to 1,600 federal civil complaints from 2019–2026 drawn from CourtListener’s RECAP archive. With almost no false positives in the pre‑AI period, the detector flagged about 1% of complaints as AI‑tainted in 2023, and about 18% of complaints containing AI‑generated text by early 2026. In other words, nearly a fifth of recent complaints show direct signs of AI assistance. Note that the study authors are careful on causation: they don’t claim to “prove” GPT‑4 caused the rise, but they show a long, flat pre‑AI time series followed by a sharp post‑LLM break that is hard to rationalize without AI.

How These Findings Apply to TCPA Litigation

TCPA practice—especially at the consumer plaintiff side—looks very much like the categories in which the pro se surge is already happening.

Template‑Driven, Fact‑Intensive Complaints: TCPA complaints usually follow a familiar pattern: venue and jurisdictional boilerplate, class allegations if any, a handful of statutory hooks, and a detailed but repetitive recitation of call or text logs. That structure is similar to the “formulaic document production” categories (consumer credit, foreclosure, many civil rights cases) where AI‑enabled pro se growth is concentrated in the data.

Low Marginal Cost Per Additional Case: Once a consumer has their call or text log, the marginal cost of turning those facts into a complaint is largely comprised of time and drafting sophistication. The paper’s evidence indicates that generative AI has sharply lowered that cost system‑wide for self‑represented litigants by making it easy to obtain “interactive, case‑specific legal guidance” and passable pleadings without a law degree.

Individualized Harms With Statutory Damages: The TCPA offers statutory damages per call or text, which naturally encourages atomized enforcement. AI’s contribution is not changing the incentive structure; it is lowering the procedural and drafting friction that previously kept many disgruntled consumers out of federal court. The pro se surge in other consumer‑facing domains is strong evidence that this friction is, empirically, what AI is eroding.

You can almost think of TCPA as a “next logical stop” for the pattern the paper documents: where an individual can tell a factual story, attach records, and plug that into a relatively stable statutory framework, AI lowers the barrier to entry.

How AI is Likely to Supercharge Pro Se TCPA Filings

Putting the paper’s numbers together with TCPA’s structure, several concrete pathways emerge.

1. Complaint Generation at Scale. The paper shows that AI assistance has taken the pro se share nationwide from about 11% to 16.8% in just a couple of years, with 18% of recent complaints exhibiting AI‑generated text. In the context of TCPA litigation, AI can:

• Draft entire complaints from a narrative: Consumers can describe who called, how often, what was said, and whether they revoked consent. A competent LLM can convert that into a complaint that, at least facially, checks basic pleading boxes—just as the paper observes happening in civil rights and consumer credit cases.
  
• Clone structure across multiple defendants: Once a consumer understands the “shape” of a TCPA complaint, AI can replicate that format for different callers or campaigns, only swapping out factual particulars.
  
• Help navigate procedure: The paper emphasizes that historically pro se litigation was structurally difficult for plaintiffs because it demanded identifying the right jurisdictional basis, pleading sufficient facts, and navigating variable procedural requirements. Modern LLMs can walk a consumer step‑by‑step through venue, amount‑in‑controversy issues, and even how to caption a complaint or respond to a 12(b)(6) motion. For TCPA defendants, that means more complaints that clear the minimal “pro se liberal construction” bar and require at least some motion practice or judicial screening.

2. AI‑Assisted Motion and Letter‑Writing Campaigns. The paper shows not just more cases, but more intra‑case activity: pro se docket entries per court in the first 180 days more than double relative to the pre‑AI baseline by 2025 Q2. Applied to TCPA, AI can:

• Generate serial filings within a single case: Pro se plaintiffs can easily produce multiple amended complaints, responses to motions, requests for extension, discovery letters, and sanctions threats, mirroring the observed growth in pro se docket activity.
  
• Support parallel complaints and agency submissions: An individual who previously might have filed a CFPB complaint or a demand letter instead of a lawsuit can now use AI to spin those into full‑blown federal complaints, while still generating regulator complaints or FCC submissions from the same factual template.
  
• Lower the literacy and sophistication threshold: The paper notes that LLMs have generated some of the fastest adoption curves in tech history and that experimental literature shows the largest productivity gains often accrue to lower‑ability writers. That suggests AI will enable would‑be TCPA plaintiffs with weaker writing skills or less formal education to produce filings that look polished enough to survive initial review.

The net effect of this is simple to imagine: more paperwork pressure on courts and defendants, even if ultimate win rates for pro se litigants remain low.

3. Strategic Learning Loops for Sophisticated Litigants. Because the paper is built on massive longitudinal court data, it implicitly captures something important: once a few AI‑assisted pro se litigants succeed in getting past dismissals, their templates and language can propagate quickly. This means that a motivated consumer can feed into an LLM excerpts from publicly available successful TCPA complaints and orders denying motions to dismiss, then ask the model to rewrite their complaint to track the language and theories that survived.

AI also allows sophisticated litigants to rapidly “A/B test” legal theories. The paper finds no detectable change in the overall distribution of case outcomes for pro se litigants—judicial dismissals stay around 60–63%, settlements around 20–23%, with wins under 1%. But within the TCPA, AI may accelerate the process by which plaintiffs converge on theories that are harder to knock out at the pleading stage, simply by iteratively adjusting complaints with model feedback based on recent rulings.

Finally, because LLMs can be prompted with local precedents and rules, a pro se litigant can tailor a TCPA complaint to Ninth Circuit case law vs., say, the Eleventh Circuit, in ways that were previously the domain of specialized litigation attorneys. That’s analogous to the paper’s broader point that AI makes it possible to “obtain interactive, case‑specific legal guidance” without a law degree.

What This Means for Courts

The paper’s central systemic concern is that you cannot “buy more judges,” yet AI is flooding the pipeline with more pro se cases and far more pro se filings within each case. For already overburdened court systems, that translates into several implications.

Growing Screening Burden: The data already show that pro se docket activity per court in the first 180 days is up 158% over pre‑AI levels, and there is no sign that the new pro se cases are resolving faster. If TCPA litigation follows the same pattern as other consumer‑facing categories, magistrate judges and district judges will face a rising wave of AI‑polished but legally thin complaints, motions, and letters, all of which need to be reviewed and analyzed.

Pressure to Adapt Gatekeeping Tools: While magazines can hire more editors to deal with a flood of AI-generated slop, courts cannot simply decline cases or materially increase their Article III headcount without Congress. That dynamic increases the appeal of stricter enforcement of Rule 8 and Rule 11 in pro se TCPA cases that are obviously AI‑generated and frivolous.

In the end, the leeway judges typically grant to pro se litigants attempting to navigate the court system will become a thing of the past. To help overcome the inevitable backlog, courts will be far less likely to overlook minor procedural errors, especially when they are committed by pro se litigants.

What This Means for Defendants

For companies engaged in phone or SMS marketing practices, the pending AI-driven litigation surge means they can expect:

More Demand Letters and Lawsuits: AI makes the drafting of a compelling TCPA demand letter or complaint a trivial enterprise. The empirical evidence supports a world where there are simply more demand letters, more pro se cases, more docket entries, and more AI‑assisted pleadings across the board. For TCPA defendants that already receive a steady trickle of individual suits, that will likely become a steady stream.

Evidence and Records Gain Greater Value: AI can manufacture plausible narratives, but it cannot alter call logs, consent records, and opt‑out timestamps. In a world of more AI‑assisted pro se TCPA cases, robust, well‑structured evidentiary systems become the anchor that separates meritorious suits from noise.

How to Cope With AI-Driven TCPA Litigation

For business owners, the wave of AI‑generated TCPA demand letters and small‑claims lawsuits is mostly about managing noise without losing sight of real risk. The letters may look more polished and legalistic than what you saw a few years ago, but that does not mean every claim is valid or worth paying quickly just to “make it go away.” Your first line of defense is a simple, consistent process: log every demand, keep copies of all letters and emails, and match each claim against your own records of calls, texts, consent forms, and opt‑out requests. Often, clean internal records will show that the allegations are exaggerated, incomplete, or flat‑out wrong.

Because you cannot stop people from using AI, focus on what you can control. Make sure your marketing and customer‑contact practices are documented; that you know exactly how you obtain consent, and that you have a reliable way to honor “stop” or “do not call” requests.

When a demand arrives, avoid emotional responses and do not admit fault in writing; instead, give your records a quick check and, where the dollar amount or facts justify it, get a short consultation from experienced counsel so you don’t accidentally create bigger problems. Over time, you can work with counsel to develop a handful of standard response approaches—for example, one for clearly baseless letters, one for good‑faith but mistaken claims, and one for situations where a small, structured settlement really is the cheapest exit.

Finally, remember that you can use technology, too. Basic tools can help you organize call and text logs, track prior complaints from the same person, and even generate draft responses that your team or lawyer can review before sending. The goal is not to turn you into a full‑time litigator, but to keep AI‑driven volume from overwhelming your day‑to‑day operations. With a few routines and better record‑keeping, most businesses can handle an increased trickle of TCPA demands calmly and cheaply, while reserving time and money for the rare cases that truly deserve a serious legal response.