AI legal research tools sound safe, but does that just shift who gets caught faking citations, not how often it happens?

This explores whether solo and small-firm lawyers are more exposed to AI-invented case citations than large firms, or whether each faces a different version of the risk. The headline number points one way. In one count of 114 US court cases with suspected AI errors, 90 percent involved solo or small firms and 56 percent involved plaintiff's counsel Do small law firms misuse AI more often than large ones?. That count only covers errors that got caught. It tells you who gets caught, not who misuses AI most often. Small practices usually have no second reader, no cite-checking staff, and often use general chatbots. Their mistakes are more likely to reach a judge unfiltered and in a form that is easy to spot.

Large firms face a quieter risk. They usually pay for legal research tools sold as safe, but a preregistered test found that Lexis+ AI, Westlaw AI-Assisted Research, and Ask Practical Law AI still produced hallucinations 17 to 33 percent of the time How often do legal AI tools actually hallucinate citations?. Because these are closed systems, outsiders can't check them independently. The danger moves from "I used ChatGPT and didn't check" to "I trusted the expensive tool my firm bought." That kind of error is easier to miss, because a plausible case from a trusted vendor gets less scrutiny than an obvious fake.

The errors are also hard to catch because of how people read citations. A study of 24,000 Search Arena interactions found that users favored answers with more citations almost as much when those citations were irrelevant as when they were relevant Do users trust citations more when there are simply more of them?. AI judges show the same weakness: fake references and polished formatting raise their scores whatever the content says Can LLM judges be fooled by fake credentials and formatting?. A brief full of citations looks authoritative to an overworked reader, whether that reader is a solo practitioner reviewing their own draft or a partner skimming an associate's work.

Some researchers argue that "hallucination" is the wrong word. The model produces true and false citations by exactly the same process, so the fix isn't better grounding but checking the output afterward Does calling LLM errors hallucinations point us toward the wrong fixes? Should we call LLM errors hallucinations or fabrications?. Seen that way, firm size matters mainly because it decides who can afford the checking. Courts don't make up the difference: when they find fabricated citations they rarely punish the fabrication itself, and sanctions depend on harm and intent Do courts actually sanction fabricated AI citations when detected?. Academic publishing offers a useful comparison. Hundreds of suspect citations turned up in accepted NeurIPS papers How many accepted conference papers contain hallucinated citations?. ICLR treated confirmed fake references as a clear-cut reason to reject a paper, and sent the fuzzier detector flags to human reviewers How can conferences detect and handle LLM misuse in peer review?. Courts could adopt a similar process, and it would protect solo lawyers better than relying on each lawyer's own discipline.

The corpus has no data on actual misuse rates by firm size, so the honest answer is that the risks differ in kind, and nobody has shown that solo lawyers misuse AI more often. Solo lawyers are more likely to have errors that go unchecked and get caught. Large firms are more likely to have errors that come from trusted vendor tools and slip through quietly.