When rent payments at her Long Beach trailer park fell behind and an eviction notice arrived, White sought help from a local tenant advocacy lawyer — and still lost her jury trial.
Determined to appeal, White believed Los Angeles’ pandemic-era protections, which allowed tenants a 12-month rent deferral, applied to her situation. Having previously used AI to create videos for her music production side business, she wondered: Could AI help her fight an eviction, too?
“It was like having God answering my questions,” White said of turning to ChatGPT and the AI search engine Perplexity for guidance.
White fed ChatGPT court documents and detailed context about her case. The chatbot parsed possible judicial errors, identified procedural missteps, mapped out strategies, researched statutes and drafted filings. After months of litigation, White overturned her eviction and dodged roughly $55,000 in penalties plus more than $18,000 in back rent.
“I can’t overstate it: I would never, ever have won this appeal without AI,” she said.
Initially using ChatGPT’s free version, White eventually subscribed to the $20-a-month premium plan and also paid for Perplexity Pro.
She is far from alone. With generative AI tools now widely accessible, a growing number of people are walking into court without lawyers and relying on AI as their primary legal adviser.
To explore the trend, NBC News interviewed more than 15 self-represented litigants, attorneys, legal aid leaders, and AI companies. Their experiences paint a complicated picture: AI can help demystify legal processes — or ruin a case with confidently delivered falsehoods.
A surge in AI-guided self-representation
“I’ve seen more pro se litigants this past year than in my entire career,” said Meagan Holmes, a paralegal at Phoenix law firm Thorpe Shwer. She attributes the spike largely to AI.
Major AI companies caution users not to rely on their tools for legal advice. Google’s terms of service explicitly warn against it. xAI prohibits using its models in regulated industries without meeting compliance rules. But these warnings are buried in fine print — and when asked legal questions, most chatbots answer anyway.
Perplexity spokesperson Jesse Dwyer acknowledged that the company isn’t perfectly accurate but said reliability is their top priority.
OpenAI, Anthropic, Google DeepMind and xAI did not comment.
Among those who turned to AI was New Mexico fitness business owner Staci Dennett. When she was sued over an unpaid debt, she fed ChatGPT her drafts and told it to “pretend to be a Harvard Law professor and tear my arguments apart.”
Ultimately she settled the case, saving more than $2,000 — crediting the win largely to AI.
Even the opposing lawyers, she said, complimented her legal fluency.
AI’s darker side: hallucinations, fabrications and costly mistakes
Legal professionals say that, alongside success stories, AI-assisted litigants frequently file documents riddled with errors.
Holmes said many pro se litigants cite cases that simply don’t exist. Courts often dismiss such filings for failing to state a legitimate claim.
AI hallucinations — fabricated facts, cases, or quotes — are the biggest hazard. Because chatbots predict likely-sounding text, they sometimes invent legal citations entirely.
Recent examples include:
A $311M energy drink mogul
Jack Owoc, the Florida entrepreneur behind a famous energy drink brand, filed a motion citing 11 nonexistent cases. He admitted he used generative AI to save money. The court sanctioned him with 10 hours of community service and now requires him to disclose AI use in all filings.
A pickleball businessman
Florida litigant Earl Takefman repeatedly cited a fictional case — Hernandez v. Gilbert (1995) — even after a judge warned him. When he asked ChatGPT to provide “actual quotations,” the bot manufactured those too.
“I told ChatGPT it had really screwed me over,” he said. “It apologized.”
Though the court ultimately spared him sanctions, he now triple-checks everything: he cross-references responses between Grok and ChatGPT, then verifies all cases manually in Google.
A growing global database
Legal researcher Damien Charlotin tracks AI-related judicial incidents. His database lists 282 U.S. cases and more than 130 international ones since 2023 — likely just the tip of the iceberg.
He says hallucinations fall into three categories:
- Fabricated cases
- Fake quotations of real cases
- Misrepresentations of real case law
The first two are relatively easy to detect. The third is not.
Even licensed attorneys have been caught. A California lawyer was fined $10,000 after 21 of his 23 cited case quotations turned out to be hallucinated by ChatGPT — the largest known penalty of its kind.
When AI overwhelms the courts
Some self-represented litigants now flood courts with AI-generated filings.
Registered nurse Matthew Garces of New Mexico, a vocal pro-AI litigant, currently has 28 federal civil cases in progress. His filings often reference nonexistent cases. The 5th U.S. Circuit Court of Appeals warned him — for a second time — to stop submitting frivolous or abusive filings. Another judge recommended he be restricted from filing new cases without approval.
Still, Garces defends his approach: “AI provides access to the courthouse doors that money often keeps closed.”
The legal profession tries to adapt
Despite the pitfalls, some lawyers and public-interest groups see promise in teaching people how to use AI safely.
“This is the most exciting time to be a lawyer,” said Zoe Dolan, supervising attorney at Los Angeles–based Public Counsel. “One advocate can now have exponentially more impact.”
Dolan helped launch a class for self-represented litigants in L.A. County, teaching them how to:
- craft effective prompts
- draft documents
- fact-check AI output
- use multiple chatbots to cross-verify results
Graduates — including White — have since won cases using AI tools.
Many lawyers say they personally use AI for brainstorming and initial research, but never trust citations without verifying them.
“Going forward, every attorney will have to use AI. Otherwise they’ll be outgunned,” said Southern California attorney Andrew Montez. “AI is the great equalizer.”
Still, Montez doubts AI can grasp enough context to reliably win most complex cases for untrained users — though he does think success in small claims court will become increasingly common.
“My slingshot was AI”
White, the tenant who successfully fought off eviction, sees AI as a powerful equalizer, especially for those who can’t afford legal representation.
Asked what advice she’d give other self-represented litigants, she drafted her answer with ChatGPT:
“AI gave me research support, drafting help, and organizational skills I simply couldn’t access on my own. It was like having a virtual law clerk.”
Then she added her own line — the one she believed captured her experience perfectly:
“It felt like David and Goliath, except my slingshot was AI.”
