Not long ago, when a client had a legal question, they called a lawyer. Now, they call after they have already asked the robot. By the time you get on the phone, the conversation feels different. They are not just asking questions. They are testing your answers. They have a theory of the case, a few newly minted legal terms, and a level of confidence that only a two-minute AI search can produce.

“ChatGPT said…” has quietly replaced “I read online…” And, in some ways, it is harder to deal with. When people use AI tools like ChatGPT or Google Gemini, the answer shows up fully formed, well written, and delivered with a kind of confidence that makes it sound settled.

So now you are not just advising a client. You are managing a client who thinks they may already have the answer.

The problem is not that clients are using AI. That part is not going away. The problem is that AI is very good at sounding right, even when it is not. It gives clean, general answers. It smooths over the messy parts. It does not know your client’s full set of facts, the personalities involved, the local rules, or the small procedural details that end up mattering most. It is relying on a prompt from someone who doesn’t even know what questions to ask.

But the client does not see any of that. They see a clear answer. And now they want to know why yours is different.

So, how can you deal with these clients? Telling them, “That’s wrong,” usually does not land well. A better approach is to meet them where they are and then shift the frame: “That’s a good starting point. What it’s giving you is a general answer. What we need to do is look at how that applies to your specific situation.”

That does two things. It respects the effort they made, and it gently reminds them why they hired you. You are not competing with the tool. You are doing something the tool cannot do—exercise judgment.

Most clients have never really had to think about the difference between legal information and legal advice. AI lives entirely on the information side of that line. AI does not verify facts. It does not make judgment calls. It does not carry any responsibility if it gets the answer wrong. It does not know your jurisdiction, your judge, or your timeline.

You do. And that is where the value is.

So, part of your job as a lawyer today is to manage these clients who are using AI. Part of that management is advising them about the risks of entering confidential information into AI platforms. That creates real risk. Decisions like United States v. Heppner suggest that communications entered into consumer AI tools may not be protected by attorney-client privilege or the work product doctrine. In practical terms, a client may be exposing information they assumed was protected.

If clients insist on using AI to work on their own case, strongly urge them not to input confidential information and to use AI only at the direction of the lawyer. This will make it more likely that the input and output will be protected by the work product doctrine.

The best time to deal with this is at the beginning of the representation. A short conversation or even a paragraph in your engagement agreement can prevent a lot of trouble later. Let clients know that it is fine to be curious and to learn, but there are limits. Explain the risks in plain English. Set some guardrails.

At the end of the day, clients have always come in with outside information. What is different now is the speed, the polish, and the confidence behind it. Your role has not changed, but the environment has. You are not just providing answers. You are helping clients understand why some answers do not quite fit, where the risks are, and what actually applies to their situation. The client may have asked the robot first, but they still hired you for your judgment, experience, and accountability.