AI Prompts Can Become Discoverable Evidence in Expert Reports

# When AI Prompts Become Part of the Evidence

A recent Connecticut federal court decision raises an important issue for anyone using AI in expert work: when does an AI prompt stop being a private drafting aid and become part of the expert's discoverable methodology?

An expert witness reportedly used ChatGPT while preparing a report, including asking the system to help produce an analysis designed to support the client's position. Opposing counsel challenged the use of AI, and the court ordered production of the prompts because the AI-assisted process was treated as part of the expert's methodology rather than merely a protected drafting note.

The lesson is bigger than one case

The key question is no longer simply:

> Did the expert use AI?

It is:

> How materially did AI influence the methodology, analysis, or conclusions?

That means expert engagements should increasingly address:

• Whether AI tools may be used
• What purposes they may be used for
• Whether prompts must be retained
• Whether material AI outputs must be preserved
• What independent verification is required
• Which conclusions remain the expert's own professional judgment

AI does not remove expert responsibility

An AI-assisted report can still be defensible. But the expert should be able to explain what the system contributed, what sources informed the analysis, what was independently checked, and how the final conclusions were reached.

The important shift is from AI use to AI provenance.

If AI materially contributes to expert methodology, the record should make that contribution understandable rather than leaving it hidden behind the final report.

That is why auditability should be designed into the workflow from the beginning—not reconstructed after discovery demands it.

The real question is not whether experts can use AI. It is whether they can prove the integrity of the work produced with it.