Added on: 18 August, 2026
AI drafting tools have made it faster than ever to turn a stack of medical records into a demand letter or medical chronology. What they haven't solved is accuracy at the level a personal injury case actually requires. Across the AI-generated demand letters and medical chronologies FactAssess reviews through our AI Output Audit service, the same categories of error show up again and again — some subtle, some serious enough to undermine an entire claim.
Below are the eight most common issues we find, why each one matters, and what independent, source-traceable human review catches that AI drafting tools consistently miss.
1. Incorrect Billing
AI-generated damages calculations don't always distinguish between billing that belongs in the claim and billing that doesn't. We regularly see total billing figures that quietly include unrelated medical visits, administrative fees, or even costs like record copying charges — line items that inflate the damages total without actually representing compensable treatment. Once an adjuster or defense counsel catches one inflated number, it casts doubt on every damages figure in the letter, not just the one that was wrong.
2. Unrelated Visits Taken as Related
AI drafting tools often work from date proximity and record volume rather than a true clinical causation analysis. That means medical visits that happen to fall within the treatment window — but that address a separate, unconnected health issue — can get folded into the treatment chronology and damages total as if they were part of the claim. This is one of the most common causation errors we catch, and one of the easiest for defense counsel to exploit if it isn't caught first.
3. Prior Injuries Not Flagged or Considered
A prior injury to the same body region as the current claim is exactly the kind of detail defense counsel searches for to argue the client's condition pre-existed the incident. AI-generated chronologies frequently fail to surface these prior injuries at all — not because the information isn't in the records, but because identifying its legal significance requires judgment the AI tool isn't applying. A demand letter that doesn't proactively address a documented prior injury hands the defense a ready-made argument instead of getting ahead of it.
4. Incorrect or Inconsistent Formatting
Beyond factual accuracy, AI-generated documents frequently have formatting inconsistencies — mismatched date formats, inconsistent provider name references, irregular section structure, or formatting that doesn't match a firm's standard demand letter template. These issues don't affect the substance of a claim, but they undermine the professional presentation an adjuster expects, and inconsistent formatting can make an otherwise strong demand letter look rushed or unreviewed.
5. References to Other Medical Providers Ignored
Medical records routinely reference treatment from other providers — a referral, a specialist consult, an emergency room visit noted in a follow-up appointment — that may never have been directly supplied as part of the record set. AI drafting tools typically only process the documents they're given, and don't flag when a record references outside treatment that hasn't been accounted for. Missing a referenced provider can mean missing treatment history that's directly relevant to both causation and damages.
6. Incorrect Citations or Statutes When Citing Liability
Perhaps the most serious error category: AI-generated demand letters can cite liability clauses, statutes, or contractual obligations that read as authoritative but don't actually exist, or that misstate what the applicable law or clause actually says. This is a form of AI hallucination — confident, specific-sounding legal language with no basis in the actual source material. An unsupported citation is often disprovable with a single phone call, and it can undermine the credibility of an otherwise well-documented claim.
7. Unrelated ICD/CPT Codes Included
Incorrect or mismatched ICD and CPT codes are a recurring issue in AI-generated medical chronologies — diagnostic or procedure codes that don't match the documented treatment, or codes carried over from an unrelated visit or condition. Beyond weakening the chronology's clinical accuracy, incorrect coding can create friction during claims review and give an adjuster grounds to question the accuracy of the entire document.
8. Handwritten Records Incorrectly Captured
Handwritten provider notes remain a common source of AI transcription errors — misread dosages, misinterpreted shorthand, incorrect dates, or missed diagnoses buried in difficult-to-read clinical handwriting. Optical character recognition and AI extraction tools have improved significantly, but handwritten records are still where AI-generated chronologies are most likely to contain simple, avoidable factual errors that a human reviewer familiar with medical documentation would catch.
Why These Errors Keep Happening
None of these eight issues are the result of an AI drafting tool malfunctioning. Most AI drafting platforms are built to organize, summarize, and draft quickly from the records they're given — genuinely useful for accelerating the first-draft stage of a demand letter or medical chronology. What they're generally not built to do is apply the clinical and legal judgment needed to distinguish related from unrelated treatment, recognize the significance of a prior injury, verify a citation against actual law, or catch a missing referenced provider.
That's the gap independent AI output audit and certification is built to close — not replacing AI-assisted drafting, but adding the verification layer that catches what AI alone consistently misses.
How FactAssess Catches These Errors
Every AI Output Audit & Certification engagement checks specifically for the error categories above — source-traceability of every factual claim, hallucination detection on cited authority, damages and billing accuracy, ICD/CPT code verification, causation and prior-injury analysis, and formatting and compliance review — before a document is certified and delivered with a full audit trail memo.
FAQs
Common questions from this post.
Know which of these eight errors are hiding in your next demand letter — before you send it.
Our Services
About Us
Get in Touch with Us!
© 2026 FactAssess. All rights reserved. Privacy Policy