We Asked ChatGPT to Help a Borrower Sue. It Found 27 Claims.

Free AI tools now give self-represented borrowers polished, procedurally sound complaints in minutes, and that is moving cases past the stage where pro se litigation used to end.

Not long ago, a pro se filing in a foreclosure case might arrive handwritten on a single sheet of legal paper. Defense counsel could usually resolve it at the motion-to-dismiss stage on service, pleading, or procedural defects, often in under ten attorney hours.

That is changing. Padgett Law Group litigators across Florida, Tennessee and Arkansas report a steady rise over the past year in pro se complaints, answers and discovery requests that are structured, cite real statutes, and allege specific facts. To see what borrowers are working with, Padgett attorney Zachary Johnson ran a simple test with ChatGPT.

The experiment

One plain-language prompt and a standard FHA deed of trust produced 27 potential claims against the lender and servicer.

Zach started with a boilerplate FHA deed of trust, with every piece of identifying information removed. He gave it to ChatGPT with the kind of message a distressed borrower might write: the bank is trying to foreclose, it isn't answering calls or emails, I want to sue, what can I do to save my home?

ChatGPT returned 27 possible claims. Asked to narrow them to the strongest five, it offered:

1. Failure to follow FHA and HUD loss mitigation requirements, including whether a loss mitigation review and a face-to-face meeting attempt took place before acceleration
2. A defective acceleration notice for lack of proper notice
3. RESPA violations for failing to respond to written inquiries
4. Breach of contract and breach of the duty of good faith and fair dealing
5. A defective substitution of trustee or a broken chain of assignments, depending on the jurisdiction

It then offered to draft a complaint for Chancery Court in Shelby County, Tennessee, and an affidavit to go with it. When Zach fed it a follow-up fact (no face-to-face meeting was offered), it asked a series of questions about notices received, explained which notices were due and when, and offered to draft a response.

"Something that takes us three, five, ten hours to draft and do properly, they're getting done in ten minutes," Zach said. For the borrower, the only real cost is the filing fee.

What the AI got right, and what it got wrong

Free AI tools are fairly reliable on statutes and unreliable on case law, and even their statutory citations can point to the wrong source.

Zach's assessment: a B-minus on statutory citations and an F-minus on case law. The tools will name specific sections of the U.S. Code with reasonable accuracy. Case citations are another matter. Padgett's supervising litigation attorney for Florida says a foreclosure case citation with no bank or trust in the caption is an immediate red flag that the case may not exist, or may not say what the filing claims.

The experiment itself showed the problem with statutes. ChatGPT told the borrower that 12 U.S.C. § 2605 required the servicer to offer a face-to-face meeting. Section 2605 is RESPA's servicing provision. The FHA face-to-face meeting requirement comes from HUD's regulations at 24 CFR 203.604. The claim sounded right, cited a real law, and pointed to the wrong one.

That is why AI-drafted filings take real time to answer. Even with legal research tools that check an opponent's citations, a licensed attorney still has to open every case and every statute and confirm what it actually says.

Why it matters: the cost moves to discovery

The risk isn't that AI-assisted borrowers win; it's that their cases survive long enough to reach discovery, where defense hours multiply.

Civil cases move through a series of stages: the initial filing, a motion to dismiss, discovery, then summary judgment. Pro se cases traditionally ended at the second stage. An AI-drafted complaint that alleges specific facts, rather than bare conclusions, is much more likely to survive motion to dismiss.

Michael Burns, who moderated the webinar, described a current pro se lawsuit against a consumer finance client. Before AI, he estimates he would have resolved it on an early motion to dismiss in under ten hours. Instead, the case passed the pleading stage and he is more than 60 hours into discovery, including disputes over producing internal policies and procedures.

Tennessee gives defendants 30 days to respond to a borrower complaint or counterclaim, and some defenses are waived if not raised early. A polished filing that sits unreviewed can quickly become a default or a lost defense.

The outcome in most of these cases is unlikely to change. FHA deeds of trust and standard mortgage instruments have been tested for decades. What changes is the cost to get there: in Zach's words, whether an attorney is on a matter for 10 hours or 80.

Where AI-assisted litigants still fall short

AI can write a borrower's filings, but it can't stand beside them in a courtroom or a deposition.

The Padgett team is seeing pro se litigants ask sharper, more lawyerly deposition questions than before. We also note a shift in Florida: litigants who once filed "sovereign citizen" arguments are being steered by AI toward real procedural claims instead.

The weakness shows up live. When a pro se litigant has to introduce evidence, respond to an objection, or follow up on an answer they didn't expect, there is no script to read from. Experienced trial counsel know the rules of evidence without looking them up, and that is where AI-built cases tend to come apart.

Pointing out AI use alone rarely wins a motion. Courts still expect a procedural or merits argument, and at least one federal appellate court, the Seventh Circuit in January 2026, has suggested pro se litigants may deserve some leniency for flawed AI citations. One response is to raise AI misuse alongside sanctionable conduct, not in place of it.

Considerations for servicers

With fewer cases ending on technicalities, the strongest defense is a clean file before referral and fast attorney review after filing.

• Pre-referral compliance. Loss mitigation reviews, face-to-face meeting attempts, required notices and their timing, and the chain of assignments are the claims AI surfaces first. Gaps that once went unexamined are more likely to come out in discovery.
• Early attorney review. Contested filings benefit from reaching counsel immediately, given response deadlines and the risk of waiving defenses.
• Early access to internal documents. Counsel can object more effectively to broad or unnecessary discovery requests when they have already seen communication policies and procedures, even if those documents are never produced.
• Cost planning. Contested pro se matters may cost more to carry than they did a few years ago, even where fees are recoverable under the loan documents.
• AI use by your own staff. Borrower names, loan numbers, FHA case numbers and other personal information should stay out of public AI tools that don't state how inputs are stored or deleted.

This article is general information, not legal advice. Specific situations should be reviewed with counsel.

Frequently asked questions

Are AI-drafted pro se complaints surviving motions to dismiss?
More often than before. AI tools help self-represented borrowers allege specific facts and follow procedural rules, so their complaints are more likely to pass the pleading stage and move into discovery.

How much does AI-assisted pro se litigation add to defense costs?
It varies by case. Padgett litigators report matters that once resolved in under 10 attorney hours now running 60 hours or more once they reach discovery.

Do free AI tools cite the law accurately?
Not reliably. They are more accurate with statutes than with case law, but even statutory citations can point to the wrong law. Every citation in an AI-drafted filing needs to be checked against the source.

Can a servicer win a case by showing the borrower used AI?
Usually not on that basis alone. Courts still expect a procedural or merits argument. AI misuse is more effective when raised alongside other sanctionable conduct.

Do courts require parties to disclose AI use in filings?
Some do. In Florida, the 11th Judicial Circuit (Miami-Dade) and the 17th Judicial Circuit (Broward) require a certification when AI was used to prepare a filing, and Broward requires naming the tool. Rules vary by court and are changing quickly.

Should servicer staff put loan documents into ChatGPT or other public AI tools?
Not with identifying information. Borrower names, loan numbers, FHA case numbers and other personal information should be removed before any document goes into a public AI tool.

About the authors

Zachary Johnson, Esq. has nearly eight years of litigation experience, about five of them in foreclosure and default services. He handles contested litigation from Padgett's Tennessee and Arkansas office, from foreclosures through evictions and property preservation.

Michael Burns, Esq. moderated the webinar and litigates consumer finance and default matters for Padgett. He is an Assistant Vice President of Foreclosure Operations for the firm and is based out of the Dallas, Texas office.

Watch the full webinar

This article is drawn from "Navigating AI-Enabled Borrower Defense Tactics," a Padgett Litigation Track webinar. The recording and companion guidance deck are in the Padgett webinar archive (link to update).