AI Legal Research Tools and NJ RPC 1.1: A Competence Checklist Before You Rely on Any Output

Most NJ solo attorneys who use AI legal research tools think about verification the same way they think about Westlaw cite-checking: run it at the end, confirm the citation is real, move on. That's not enough anymore, and the gap between that habit and what NJ RPC 1.1 actually demands is wider than most practitioners realize.

RPC 1.1 requires that attorneys provide competent representation, which means the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation. The NJ Supreme Court's comment to that rule explicitly includes keeping current with changes in the law as part of that standard. AI legal research tools introduce a new category of error that neither Westlaw nor Lexis ever created: confident, well-formatted, citation-complete output that is wrong at the substantive level, not just the citation level.

Here's what that means for a practical checklist.

---

Before You Run a Query

1. Know what the tool is actually searching.

Some AI research tools (Westlaw Precision, Lexis+ AI, Casetext CoCounsel) retrieve from live, indexed legal databases. Others generate from training data with a cutoff date and no live retrieval at all. ChatGPT with a legal plugin sits somewhere in between, depending on configuration. Before you run a query on any platform you haven't used before, identify whether the tool performs retrieval-augmented generation (RAG) against a real legal database, or whether it's synthesizing from a static model. For NJ practitioners, this distinction matters because New Jersey court rules, administrative regulations, and NJBA ethics opinions update frequently.

2. Set your jurisdiction explicitly.

Several AI tools will default to federal common law or majority-rule positions unless you force them to specify New Jersey. Queries about landlord-tenant notice requirements, domestic violence restraining order standards, or evidentiary rules can return technically accurate statements about other jurisdictions' law. Build a habit of including "under New Jersey law" and the relevant statutory scheme in every query, not just the topic.

---

When You Review Output

3. Confirm every cited case exists and says what the tool claims.

This is the step attorneys know they should take and still skip under deadline pressure. The reported incidence of hallucinated citations in general-purpose AI tools remains significant enough that verification can't be treated as optional. Pull the actual decision in Westlaw or Lexis, read the relevant passage, and confirm the proposition matches. A case can be real and still be misquoted.

4. Check the holding for subsequent history.

AI tools, even good ones, sometimes miss that a case was reversed, distinguished on the exact point you're relying on, or superseded by a statutory amendment. This is especially acute in NJ practice because the Appellate Division produces a high volume of unpublished opinions that shift the practical landscape in ways that don't always get indexed cleanly by AI training pipelines. Run subsequent history independently for any case you plan to cite in a brief or motion.

5. Identify the date problem on regulatory and court rule questions.

NJ Court Rules, IOLTA regulations, and New Jersey Register amendments have specific effective dates. Ask any AI tool you use when its training data or live retrieval was last updated, and then independently verify the current version of any rule or regulation the tool cites. This applies equally to tools marketed as "current" since update cycles vary by content category even within the same platform.

---

Before the Work Product Leaves Your Desk

6. Document your verification steps in your file.

Under RPC 1.1, competence is partly a process standard, and your file should reflect that the process happened. A short internal note, even a few sentences in your case management system, logging that you verified citations, checked subsequent history, and confirmed the current version of any regulation is your protection if a question ever arises about your reliance on AI-assisted research. It also keeps you honest about whether you actually did those steps.

7. Apply the "could I explain this to the court?" test.

Before citing any research output in a filing, ask yourself whether you could stand up in front of a NJ Superior Court judge and explain the holding, its procedural posture, and why it applies to your client's facts. If the answer is no, you haven't finished your research. The AI gave you a starting point, not a conclusion.

---

The NJ ACPE hasn't issued a formal opinion specifically addressing AI-assisted legal research yet, but the existing competence framework is fully capable of capturing failures in this area. Several NJ federal court orders in 2024 have already put attorneys on notice that AI citation errors don't get treated as innocent mistakes when there's no evidence the attorney reviewed what the tool produced.

For solo practitioners especially, the practical answer isn't to avoid AI research tools. They save real time on issue spotting and initial case identification. The answer is building a short, repeatable verification checklist into your workflow so that the efficiency gain doesn't come at the cost of the accuracy standard RPC 1.1 has always required.