In 2026, a good lawyer should utilize AI but never, ever rely on it. When it comes to research, AI can give a lawyer a good start; likewise, when it comes to reviewing or drafting discovery, AI can help by providing a discovery request an attorney may not have considered by reviewing the attorney’s work, and it can help with document review by spotting crucial documents and issues buried amongst thousands of pages in an astoundingly short period of time. However, AI can also fabricate both precedential cases and conclusions out of thin air, miss crucial issues and evidence, and, occasionally, just give terrible advice. When it comes to general legal questions, AI frequently gets the answer wrong (one 2025 study found that both legal-specific and general AI are around only 80% accurate). From our experience with the leading software, AI does all of this, both good and bad.
Thus, AI can a help a lawyer in a case, but it can also cause harm. And, if lawyers rely on AI for research without checking the citations and/or conclusions, they can find themselves severely sanctioned, as has happened several times over the last year. In fact, even those on the bench fall victim to these errors, as a judicial opinion cited to non-existent cases that were nothing more than artificial intelligence hallucinations.
In order to discuss how attorneys should use AI, it must first be understood how the technology works. Artificial Intelligence products utilize a combination of natural language processing, machine learning algorithms, and large language models to analyze prompts and documents. Natural language processing allows the technology to “understand”, interpret, and understand language. This technology is commonly encountered in products such as Siri, Alexa, and all autocomplete suggestions across a variety of products. Machine learning allows the program to improve its accuracy via feedback. Finally, large language models allow the AI to understand, process, and then generate human language via large data sets and statistically analyze the relationships between words.
In addition to hallucinations, such models have also had issues with, for lack of a better term, sycophancy. In order to further incentivize re-engagement with the product, AI is sometimes known to be overly positive in feedback, telling users that they are correct when they are not. The best way to avoid falling into this trap is to ensure that one uses neutral language when inputting prompts.
So, should lawyers stay completely clear of AI so as to avoid mistakes? Absolutely not, in fact, it’s a great resource; but at this stage of AI development, one must be extremely careful with leveraging its powers. Regardless of how advanced AI gets, a lawyer should never substitute the requisite work needed to litigate a case properly by relying solely on artificial intelligence.
For example, when it comes to document review, which can be extremely cumbersome and difficult, AI can be a great resource, but, again, it’s an imperfect one. AI can sift through medical records or business documents and provide very insightful narratives, annotations, and commentary. Moreover, AI can do all this in an extremely short time, while a lawyer might take hours or even days to do the same thing. But the problem is that AI can miss evidence or not appreciate the value of certain documents as it goes through records. Moreover, and perhaps most importantly, the lawyer is the one litigating the case, not the AI, and, in order for the lawyer to pursue strategies and advocate properly, there’s no substitute for reading, re-reading, and absorbing the source evidence itself. Indeed, sometimes a lawyer can review documents multiple times, and then, only, after looking at a document numerous times, draw a certain conclusion that, for some reason, was not obvious even after several lawyers looked at it previously. The lawyer needs to make the arguments about what documents mean and in order to do that, there’s no substitute for the lawyer going through the documents and getting to know them herself. But, because AI can get through documents quickly and most of the time with significant utility, it is a tool that lawyers should use, but never rely on solely. It is also important that attorneys abide by confidentiality rules, as many AI products record all prompts and responses for continued learning. Such an upload and retention to a third-party website would violate attorney client privilege rules, and all attorneys should review the product’s terms and security measures before any use.
As to writing, we are firm believers that AI should never be used to draft briefs, or even to do an initial draft of part of a brief. Not only does AI hallucinate and frequently get things wrong, but, if a lawyer does not use his or her brain to contemplate how the facts and law interrelate, in order to deduce and devise arguments, the lawyer will fail to grasp the full meaning and scope of a case. By relying on AI to do first drafts of briefs, especially dispositive motion briefs, a lawyer will be unable to fully comprehend the strengths, weaknesses, and scope of risks involved. On the other hand, if a lawyer gets hit with writer’s block or seeks to have another rendition of the wording of a sentence or phrase, AI can come in handy in those limited circumstances to help with finalizing a brief.
Legal research evolved with technology substantially before the advent of AI. When I was in law school from 1994-97, we learned to research via book by using a descriptive word in a “Digest” to find a relevant case, and then “shepardizing” the case to see if any subsequent decisions analyzed it, distinguished from the facts of another case, or potentially overruled it. We also learned to use software such as LexisNexis and Westlaw, to find cases by using searches that looked for cases that contain certain terms and connectors, known as “Boolean” searching, which was far easier, quicker, and productive than going to a Digest. Now, a lawyer can give a general question to AI via legal software from the leading AI companies and get an answer in memo form with legal citations. Sometimes these memos are very helpful, but they can also be flat wrong, miss leading cases and even cite to authorities that do not exist. I usually take an “all of the above approach” for important legal issues using Boolean searches and AI, and when I use AI, I use it as a starting place, not an end. I always read a case that the AI cites, and search for any other cases that cite to the authority. As with preparing discovery requests or sifting through discovery responses, AI can be a helpful tool, but it cannot be the only resource relied upon by a lawyer. Otherwise, the lawyer would be serving their client short, and maybe even get into trouble.
Besides document review and legal research, as with any emerging technology, there are other inventive uses for AI technology. For example, if an attorney is in a trial, and feels like the jury found one exhibit especially compelling, AI can create an exhibit for use in the closing argument in a short amount of time. Additionally, such exhibits could be used in internal memos, for the creation of graphics to help detail what an injury or accident scene looked like, to allow a co-counsel or partner to easier understand the relevant facts of the case. All, of course, while properly checking such images are accurate representations of reality.
In short, lawyers involved in litigation today should utilize the incredibly powerful tool of artificial intelligence, but they need to do so wisely. A lawyer should never rely on artificial intelligence, but should frequently consult AI for assistance for more basic, time-consuming tasks. Most importantly, if a lawyer is going to serve her client’s interest, the lawyer must live the case by understanding the law firsthand, by absorbing the evidence obtained through discovery, and by making the connections between the law and the facts. If a lawyer overly relies on AI, the lawyer will never be able to do this, and the client will suffer as a result.