Insights / AI in Personal Injury Litigation
AI-Generated Medical Chronologies: What Attorneys Need to Know Before Relying on Them
By Breen LNC.
Short answer: An AI chronology is a draft. It can look complete and still leave things out or misstate them. Before relying on one, check that every entry cites a page, read the cited line yourself, and have a clinician decide what the entries mean and what is missing.
A polished draft is not a checked one
Arizona’s committee guidance for lawyers warns that AI can produce work that looks polished yet is “incomplete, incompetent, biased, and inaccurate.” It adds that “legal professionals must not defer to its judgment.” A medical chronology is a good place to apply that advice, since a tidy table of dates reads as settled fact.
What the studies show, and what they do not
We did not find a published study measuring error rates in litigation chronologies. The nearest evidence involves clinical documents.
- In JAMA Internal Medicine in 2025, AI discharge narratives had 2.91 errors per summary against 1.82 for physicians, and were less comprehensive, though overall quality scores were similar.
- A 2025 study in npj Digital Medicine reported a 1.47% hallucination rate and a 3.45% omission rate per annotated sentence across 18 test configurations. The authors work at a company building these tools, and they report that refining prompts and workflows reduced major errors.
These rates come from specific setups and cannot be carried over to the product your firm is considering. They do show that error rates vary and that omissions are a measurable problem of their own.
Five questions to ask of any chronology
- Does every entry cite a page you can open in one step?
- Does the cited line actually say what the entry says?
- Who decided whether the entry is clinically significant, and what are their credentials?
- What happened to pages the software could not read, such as handwriting or poor scans?
- Are the gaps marked, or does the timeline simply jump from one date to the next?
Keep it useful as a working document
The Michigan Bar Journal’s chronology advice applies here even though it isn’t specific to medicine: record discrete facts, cite the source, keep disputed facts in, mark uncertain dates, and rate how much each fact matters. A chronology that does this helps you think about the case. One that only lists encounters leaves that thinking still to do.
At Breen, a nurse decides every entry against its page before an attorney sees it.
Frequently asked questions
How accurate are AI medical chronologies?
We did not find a published error rate for litigation chronologies. Studies of AI clinical summaries report varying rates depending on the setup, which is why each chronology should be checked against its source pages.
What should every entry in a medical chronology include?
At minimum a date, the provider, what happened, and a citation to the exact page. Disputed or uncertain facts should be marked, not smoothed over.
Sources
- Supreme Court of Arizona, Steering Committee on AI and the Courts, Generative AI: Ethical Best Practices for Lawyers and Judges (Nov. 14, 2024)
- Williams et al., Physician- and Large Language Model-Generated Hospital Discharge Summaries, JAMA Internal Medicine (2025)
- Asgari et al., A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation, npj Digital Medicine (2025)
- Greg Krehel, Chronology Best Practices, Michigan Bar Journal
General information for legal professionals. This is not legal or medical advice.