If you have practiced law long enough, you have probably represented a client who graduated from Google University School of Law.  You know the type. They do a Google search about their legal matter and are ready to argue before the Supreme Court or take over the negotiations of a complex business deal. They just need you to bless their work and put your name on it. These same clients have now earned their LLM from ChatGPT School of Law. What was once a mildly annoying problem for lawyers has become a much more complicated challenge.

Whether it is ChatGPT, Claude, Gemini, Copilot, or some other generative AI platform, clients are increasingly using these AI assistants to draft contracts, wills, corporate documents, pleadings, and virtually any kind of legal document you can imagine. They then show up at your office with what appear to be polished legal documents or strategies. They explain to you that they have saved you an enormous amount of time and saved them the cost of all those hours you would have billed to complete the same work. It is a win-win, right?

Generative AI systems can feel almost magical because they are trained on staggering amounts of information. These systems analyze huge datasets that may include books, articles, websites, legal opinions, contracts, treatises, public filings, academic papers, and countless other forms of human writing. Rather than “thinking” like a lawyer or truly understanding the law, the system identifies patterns in language and predicts what words are statistically likely to come next in response to a prompt. The result can sound remarkably confident, polished, and sophisticated.

To many clients, this creates the impression that AI “knows everything” and can generate anything from a motion for summary judgment to a complex estate plan with the click of a button. But the appearance of intelligence can be misleading. Generative AI does not actually verify facts, exercise professional judgment, understand a client’s objectives, or appreciate the legal and practical consequences of the documents it produces. It is, at its core, a prediction engine built on patterns drawn from vast quantities of data.

Please don’t misunderstand my point here. I am not suggesting that AI is bad. I regularly use AI in my own work. I’m not even suggesting that clients shouldn’t use AI. I’m simply making the point that AI is not the magic pill that clients often think it is. Because clients often see only a polished and authoritative-looking output, without understanding the limitations behind it, they may struggle to accept the lawyer’s explanation that the document or strategy is flawed.

The reality is that this problem is not going away. In many ways, this is just the modern version of clients using Google to work on their own legal matters. The difference is that generative AI is far easier to use than Google and can instantly produce work product that looks like it was prepared by a lawyer.

This is quickly becoming the ordinary reality of modern legal practice. Clients will increasingly show up with AI-generated contracts, pleadings, estate plans, business documents, legal research, and litigation strategies already in hand. But this does not mean that lawyers can simply accept AI-generated work product at face value or set aside their professional and ethical responsibilities. At the same time, reflexively dismissing everything the client has generated by AI may not be the right answer either. In some cases, these materials may help clarify client objectives, organize information, identify issues, or provide a useful starting point for further analysis.

The real challenge for lawyers is learning how to navigate the middle ground. Lawyers must still exercise independent professional judgment, carefully verify the accuracy and reliability of the information provided, explain the limitations and risks associated with AI-generated content, and ensure that any final work product complies with the lawyer’s professional obligations and standards of competence.

So, what should lawyers do when clients show up with AI-generated work product in hand? Lawyers may not be able to stop clients from using AI, but they can take steps to manage these situations effectively. Let’s first discuss how lawyers should handle clients who arrive at the initial meeting with AI-generated documents or strategies and then look at proactive measures lawyers can take before the client ever walks through the door.

Start With Curiosity Instead of Defensiveness

One of the worst things a lawyer can do when a client arrives with AI-generated work product is to become defensive or dismissive. In many cases, the client genuinely believes they are helping the lawyer and reducing fees. If the lawyer immediately reacts with irritation or ridicule, the conversation can quickly become adversarial.

A better approach is to respond with curiosity. Ask the client what AI platform they used, what information or prompts they entered, and what they were trying to accomplish. It may also be worth asking whether the platform reserves the right to use user inputs or outputs to train or improve its models. Many clients have never considered that the information they entered into the platform may not have remained private.

Explain That Polished Does Not Mean Accurate

One of the challenges with generative AI is that incorrect information is often delivered in a highly polished and authoritative format. AI platforms can generate contracts, pleadings, legal analysis, and case citations that appear entirely legitimate while containing serious factual or legal errors.

Lawyers should take time to explain to clients that AI systems are capable of “hallucinating” information, including nonexistent cases, incorrect legal standards, fabricated quotations, or inaccurate factual assertions. Courts around the country have now sanctioned lawyers for filing AI-generated pleadings containing hallucinated authorities and other errors. Clients often assume that something that sounds sophisticated must also be correct. Lawyers should help clients understand these risks.

Explain Your Ethical Duties and Why Review Time Matters

Clients may assume that because they used AI to generate a document or strategy, the lawyer’s work has largely been completed. Lawyers should explain that their professional obligations require them to independently review and evaluate any material used in the representation.

Duties of competence, diligence, and independent professional judgment do not disappear simply because the first draft was created by an AI platform instead of a human being. If the lawyer’s name is going to appear on the document or the lawyer is going to rely on the information in advising the client, the lawyer must verify the accuracy of the content and determine whether it is legally and strategically appropriate.

Lawyers should also candidly explain that reviewing and correcting AI-generated work product may take as much time as drafting the document from scratch. That can be difficult for clients to hear because they understandably believed they were saving time and money. A little empathy goes a long way here. Lawyers can acknowledge why the client thought the AI-generated work would be helpful while still explaining why careful review is necessary.

Warn Clients About Confidentiality and Privilege Risks

Lawyers should also discuss the risks associated with entering confidential or sensitive information into consumer AI platforms. Depending on the platform’s terms of service and data usage policies, the client may have disclosed information to a third party in a way that could jeopardize claims of confidentiality, attorney-client privilege, or work product protection.

Courts across the country are beginning to wrestle with these issues, although many jurisdictions, including North Carolina, have not yet provided definitive guidance on all aspects of AI-related privilege and work product waiver. Until the law becomes more settled, the safest course is to advise clients not to enter confidential information into public AI platforms without first consulting counsel.

If the lawyer is meeting with the client before confidential information has been entered into an AI platform, the lawyer should specifically advise the client not to do so going forward except at the lawyer’s direction. Taking this step may help strengthen later arguments that the client intended to preserve confidentiality and work product protections.

Remember That You Can Decline to Use the AI-Generated Work Product

Lawyers should also remember that they are not obligated to use or review AI-generated work product simply because the client brings it to the meeting. In some situations, the risks associated with hallucinations, inaccuracies, or confidentiality concerns may outweigh any potential benefit.

There may be times when the lawyer reasonably concludes that it would be more efficient, less expensive, and less risky to start from scratch rather than attempt to salvage AI-generated material. Lawyers should feel comfortable communicating that decision to the client in a professional and respectful manner.

Address AI Risks in the Engagement Agreement

Finally, lawyers may want to consider addressing AI-related issues directly in their engagement agreements or initial client communications. For example, the agreement could warn clients about the risks of entering confidential information into public AI platforms and instruct clients not to use such tools for sensitive matters without consulting the lawyer first.

In addition, if the lawyer personally uses an AI platform that reserves the right to access, review, or share user inputs, the lawyer should carefully evaluate whether informed client consent is necessary before entering confidential information into the system. The North Carolina State Bar’s position is that entering confidential client information in an AI platform that reserves the right to review, access, or share that information is a violation of Rule 1.6.

Like it or not, generative AI is now part of the lawyer-client relationship. Clients are going to continue using these tools because they are easy to access, inexpensive, and capable of producing work product that appears sophisticated and authoritative. The challenge for lawyers is not learning how to eliminate AI from the practice of law but learning how to responsibly manage its growing presence in client interactions. Lawyers who approach these situations with professionalism, curiosity, clear communication, and sound judgment will be in the best position to protect both their clients and themselves.