Let AI help you draft. Then verify every citation before it goes out the door.
ChatGPT, Claude, Gemini - they're all remarkably good at drafting legal documents. The structure is right. The language is professional. The arguments flow logically. But when it comes to citations, they make things up.
In AI terminology, this is called a hallucination. When an AI model generates information that sounds authoritative but is fabricated - a case that never existed, a statute number that's close but wrong - that's a hallucination. AI hallucinations are especially dangerous in legal work because they look indistinguishable from real citations. The formatting is correct, the reporter abbreviations are right, and the case names sound plausible. Unlike a typo or a formatting error, hallucinated citations are designed by the model to be convincing. That's what makes them so hard to catch without a dedicated verification tool.
Not maliciously. AI models don't have access to legal databases. They generate text that looks like a citation based on patterns in their training data. The result: case names that sound real but reference non-existent opinions, statute numbers that are close but wrong, and citations to real authorities that have nothing to do with your legal issue.
The trap: AI-generated citations are harder to catch than random errors because they're plausible. Real reporter abbreviations. Correct formatting. Sometimes even real case names paired with wrong page numbers. They pass a quick glance. They don't pass verification.
The answer isn't to stop using AI for legal drafting - it's too useful for that. The answer is to add a verification step between "AI generated it" and "I filed it."
1. Draft with AI. Use whatever tool you prefer. Let it structure your arguments, suggest relevant authorities, and format the document.
2. Verify with ADP. Upload the draft. Every citation gets checked against authoritative Florida sources. You'll know in minutes which citations are real, which are hallucinated, and which exist but don't apply.
3. Fix and file. Replace bad citations with real ones. ADP's Deep Analysis can even suggest alternative authorities that actually support your argument.
Judges are already sanctioning attorneys for filing AI-generated documents with fake citations. In Mata v. Avianca, a federal judge sanctioned lawyers who submitted a brief full of non-existent cases generated by ChatGPT. Courts across the country are now requiring attorneys to certify that AI-generated content has been verified.
If courts are holding licensed attorneys to this standard, imagine the scrutiny a pro se filing gets. It takes minutes and it could save your case. Trust but verify.
