Answer extracted from the The Think Small Podcast — listen to the full episode below.
What is the impact of using ChatGPT to review transactional documents during deal negotiations?
When clients use ChatGPT to review contracts before attorney review, the tool typically generates 10 to 20 flagged issues, but only two or three are valid insights the attorney wasn't already planning to address. This forces lawyers to spend extra time addressing non-substantive AI-generated concerns alongside the opposing party's actual comments, directly inflating client costs and potentially delaying deal closure.
The problem lies in how AI processes contractual language. ChatGPT generates queries and observations based on pattern recognition, not legal judgment. A contract review tool cannot distinguish between a genuinely problematic clause and a boilerplate provision that poses no real risk to the transaction.
When a client presents an attorney with both the opposing side's revisions and a long list of ChatGPT-flagged concerns, the attorney must methodically address each one—a process that consumes billable hours. Each non-substantive issue requires the same time investment as a legitimate legal point, since the lawyer must evaluate and explain why it is or isn't a true concern.
As Brian Calhoun explains in the full episode, this dynamic compounds negotiation friction. Clients who insert AI-generated commentary into back-and-forth exchanges create additional rounds of clarification, extending timelines and increasing overall legal spend.
The Hidden Cost: False Positives in Deal Flow
The 10-to-20 output rate from ChatGPT is not accidental—it reflects the tool's design. AI tools are trained to err on the side of verbosity, offering multiple angles and observations to appear thorough. In a transactional context, this behavior is counterproductive.
Business owners often believe they're being diligent by running contracts through AI before handing them to counsel. In reality, they're introducing noise into the process. A detailed discussion of this precise issue is available in this episode, where the attorneys also address what happens when AI-generated case law is relied upon in litigation—an even riskier scenario.
"ChatGPT will tell you anything that you want to hear. Your prompt that you put into OpenAI is discoverable in litigation."
Brian Calhoun — Managing Partner, Calhoun Bhella LLP. Founded the firm in 2010 alongside partner Gwen Bella after years working as an associate at major law firms in California and Dallas. Calhoun specializes in mergers and acquisitions, corporate transactions, vendor agreements, and general counsel services for mid-market businesses that lack in-house legal resources.
This quote underscores a dual risk: not only does ChatGPT generate incorrect or irrelevant observations, but any use of the tool leaves a discoverable record. In litigation, opposing counsel can subpoena the prompts you entered and the AI's responses—a fact business owners rarely consider when they run confidential deal terms through a public AI platform.
For transactional work, the practical lesson is clear: as discussed in depth in the podcast, the best approach is to rely on your attorney's expertise from the start, rather than attempting to pre-filter documents through an AI system designed for general conversation, not legal analysis.
See also
How does client involvement in litigation cases improve case outcomes and reduce legal costs?
When clients are actively involved and organize facts clearly for their attorney, it cuts out red tape inherent in large law firms. The more a client helps organize information, the more efficiently the attorney can work.
What are the risks of relying on AI case law citations in legal proceedings?
There was a recent ruling out of Dallas where an attorney relied on a case that ChatGPT generated, and it was entirely made up, resulting in serious professional consequences for the attorney.
How should business owners approach using AI-generated legal content when facing litigation?
Business owners should not rely on AI for litigation matters because prompts entered into OpenAI are discoverable in litigation and are not protected by attorney-client privilege.