Legal Research for Paralegals: What You Can Do, What You Can't, and How to Deliver It
Paralegals do real legal research every day — the trick is knowing where the unauthorized-practice line sits, how to take an assignment so you don't research the wrong thing, and how to deliver work an attorney can sign with confidence.
Yes, paralegals can do legal research — and in most small firms, they do the bulk of it. A paralegal can find controlling cases, pull statutes and regulations, read opinions, and draft a research memo, as long as a licensed attorney supervises the work, reviews the product, and stays responsible to the client. The one thing a paralegal cannot do is give the client legal advice. That single distinction — researching what the law says versus telling a client what it means for them — is the whole game.
This guide covers the four things that make a paralegal a research asset rather than a liability: knowing exactly where the unauthorized-practice line sits, taking the assignment so you never research the wrong question, building a free-tool stack that covers real primary law, and delivering work an attorney can sign without redoing it. It ends with a cautionary tale from a 2025 Utah case that every supervising attorney and every paralegal should read once.
Can paralegals do legal research? Yes — here's the line
The rule is settled across every authority that governs the profession. The ABA Model Guidelines for the Utilization of Paralegal Services let a lawyer delegate almost any task to a paralegal — including legal research and drafting — provided the lawyer maintains a direct relationship with the client, adequately supervises the work, and takes professional responsibility for the product. The NALA Code of Ethics and Professional Responsibility says the same from the paralegal's side: a paralegal must not give legal opinions or advice, establish the attorney-client relationship, or otherwise engage in the unauthorized practice of law. The NFPA Code of Ethics and Professional Responsibility draws the identical boundary.
So what actually crosses the line? ABA Model Guideline 3 is the clearest test. A lawyer may not delegate three responsibilities to a paralegal:
- Establishing the attorney-client relationship — deciding to take the client on.
- Setting the fee — deciding what the representation costs.
- Rendering a legal opinion to the client — telling the client what the law requires of them.
Everything short of those three is delegable under supervision. Locating authority, reading and summarizing cases, checking whether a statute is current, drafting the research memo, even proposing an analysis for the attorney's review — all fair game. The moment a paralegal answers a client's "so what should I do?" with anything other than "let me get the attorney," the work has drifted from research into advice, and that is the unauthorized practice of law. UPL paralegal rules are not about the difficulty of the research; they are about who owns the judgment call. Research is information. Advice is judgment. Paralegals produce the first so the lawyer can exercise the second.
The intake habit that saves your afternoon
The most expensive paralegal research skill has nothing to do with databases. It is taking the assignment correctly. The single biggest source of wasted research hours is a paralegal working hard on the wrong question, in the wrong jurisdiction, and discovering it at the deadline.
Before you open a single tab, get four things from the assigning attorney — in writing, even if "in writing" means a two-line email or a note you read back to confirm:
- Jurisdiction and court. State or federal? Which state? Trial or appellate? A perfectly researched California answer is worthless in a Utah district court. This decision drives everything downstream, because it determines what counts as binding.
- The precise question. Not "look into the statute of limitations" but "what is the limitations period for a breach-of-written-contract claim in Utah, and when does it start to run?" A vague question guarantees a vague answer.
- The deadline. A memo needed in an hour for a phone call is a different deliverable than one due next week for a brief. The deadline sets the depth.
- The format. Does the attorney want a two-sentence answer, a formal memo, or just the cases pulled and printed? Guessing wrong here means doing the work twice.
Writing these down is not bureaucracy. It is the difference between a paralegal the attorney trusts with ambiguous assignments and one who needs everything spelled out. It also protects you: when the question was ambiguous and you confirmed it in writing, a wrong turn is a shared miss, not your fault. Nail the assignment, then start.
A free-tool stack that covers real primary law
A paralegal does not need a $300-per-seat subscription to do sound research. A working free stack covers the primary law behind most assignments, and knowing it cold is a genuine paralegal research skill.
- Case law: CourtListener (from the nonprofit Free Law Project) and Google Scholar both hold full-text state-appellate and federal opinions at no cost. Our walkthrough on how to research case law covers the query technique, and the guide to Google Scholar for legal research explains the two features most people skip — court selection and the citing-cases view.
- Statutes and regulations: Cornell's Legal Information Institute hosts the U.S. Code and the Code of Federal Regulations for free; govinfo is the government's own authenticated source; and every state legislature publishes its current code online.
- Bar-member access: more than 80 state and local bar associations give members complimentary Fastcase (now part of vLex), a full primary-law library. If your supervising attorney is a bar member, you likely already have it — ask.
The free stack's limits are real. It is thin on editorial treatment and secondary sources, and no free case-law site will tell you at a glance whether a decision is still good law. That is exactly why the next two sections matter: what you deliver has to close the gaps the free tools leave open.
The deliverable attorneys actually want
Ask any supervising attorney what separates a great research memo from a merely correct one, and the answer is always the same shape: answer first, authorities linked, treatment checked. The attorney does not want a tour of your research process. They want the conclusion, the binding authority behind it, and the confidence that the authority is real and current — fast enough to use before the deadline.
Here is the checklist to run before you hand anything back.
| Element | What it means | Why the attorney needs it |
|---|---|---|
| Answer first | One or two sentences that directly answer the question asked | The attorney reads the conclusion before the reasoning |
| Question restated | The exact issue and jurisdiction you researched | Confirms you answered the right question |
| Controlling authority, linked | The binding cases and statutes, each hyperlinked to its full text | One-click verification, not a summary to take on faith |
| Treatment checked | A note on whether each key case is still good law | Keeps an overruled holding out of the brief |
| Persuasive authority labeled | Any non-binding source clearly marked as such | Keeps the authority hierarchy honest |
| Open questions flagged | What you could not resolve, and why | The attorney owns the judgment calls |
| Date and sources | When you ran the research and where you looked | Research has a shelf life |
Two of those rows carry most of the weight. "Controlling authority, linked" means you did the work of separating binding precedent from noise — our guide to finding controlling authority walks through the jurisdiction-and-hierarchy test that decides what binds. "Treatment checked" means you confirmed the case has not been reversed, overruled, or seriously questioned since; how to shepardize a case shows how to run that check with free tools when you do not have a paid citator. Skip either row and you have handed the attorney a research problem disguised as a research answer.
The supervision lesson: Garner v. Kadince
In 2025 a Utah case turned the abstract supervision rule into a concrete, published warning. In Garner v. Kadince, 2025 UT App 80, a petition filed with the Utah Court of Appeals cited a case called Royer v. Nelson — a case that does not exist. It had been generated by ChatGPT. The petition, the court explained, had been drafted by an unlicensed law clerk at the firm who used the chatbot, and the attorney who signed and filed it never independently checked the citations.
The court's response, as reported by KSL, was pointed. The attorneys "fell short of their gatekeeping responsibilities as members of the Utah State Bar." The signing attorney was ordered to pay the opposing party's attorney fees for the time spent responding, to refund his own client the fees tied to the petition, and to donate $1,000 to a Utah legal-aid nonprofit. The lesson the court drove home: attorneys must verify each source, given the technology's tendency to hallucinate.
For a paralegal, the takeaway is not "don't use AI." It is that the supervising attorney is on the hook for every citation regardless of who produced it — so the paralegal's job is to make verification easy, not to be a source of unverified work. Never hand up a citation you have not opened and read in its actual source. If a tool produced a summary, pull the real opinion behind it. Utah's guidance on this is worth reading in full; we cover it in our post on AI ethics for Utah lawyers, and the mechanics of catching a fabricated cite are in how to verify AI-generated citations.
This is also where the right tools earn their keep. When a paralegal and the supervising attorney work the same matter, a research tool that links every citation to its source — instead of generating plausible-looking cites from memory — turns verification from a chore into a click. CaseRead was built for exactly this handoff: on the Team plan, a paralegal can run matter-aware research across all 53 US jurisdictions and the firm's own case files, and every answer traces to a source the attorney can open and confirm. Retrieval, not fabrication, is the point — and it is the difference between the Garner pattern and a clean filing.
The bottom line
Paralegals are the research engine of most small firms, and the rules give them wide room to run — everything but the client-facing judgment calls. The technology is reshaping how that research gets done, not replacing the paralegals who do it. The paralegals attorneys trust with the hardest assignments do three things reliably: they pin down the jurisdiction, question, deadline, and format before they start; they lean on a free primary-law stack they know cold; and they deliver the answer first with every authority linked and its treatment checked. Do that, and you are not just doing legal research — you are making your attorney's signature safe.
And whoever drafts the research, verify before it is filed. Our Hallucination Shield checks every citation in any AI-drafted text for existence and support — free, no signup. It is the two-minute habit that keeps a fabricated Royer v. Nelson out of your firm's next brief.
Frequently asked questions
Can paralegals do legal research? Yes. Paralegals routinely find cases, statutes, and secondary sources and draft research memos, and the ABA Model Guidelines expressly permit it when a licensed attorney supervises the work and reviews the product. The line is legal advice: a paralegal may research what the law says but cannot tell a client what it means for them. That judgment call belongs to the supervising lawyer.
What can a paralegal not do in legal research? Under ABA Model Guideline 3, a lawyer may not delegate three things to a paralegal: establishing the attorney-client relationship, setting the fee, and rendering a legal opinion to the client. Everything else — locating authority, reading cases, summarizing holdings, drafting memos — is fair game under supervision. Giving a client legal advice, in any form, crosses into the unauthorized practice of law.
What should a paralegal confirm before starting a research assignment? Four things, in writing: the jurisdiction and court, the precise legal question, the deadline, and the format the attorney wants — a two-line answer, a full memo, or a set of pulled cases. Confirming these before opening a database prevents the most common cause of wasted paralegal hours: researching the wrong question in the wrong jurisdiction and finding out at the deadline.
What does a good paralegal research deliverable look like? Answer first, authorities linked, treatment checked. Lead with the direct answer to the question the attorney asked, then the controlling authority with a link to each source's full text, then a note on whether any case has been questioned or overruled. Flag what you could not resolve. The attorney should be able to verify every citation in one click, not take your summary on faith.
What is the paralegal supervision lesson from Garner v. Kadince? In Garner v. Kadince, 2025 UT App 80, a Utah petition drafted by an unlicensed law clerk using ChatGPT cited a fake case, Royer v. Nelson. The attorney signed without checking. The court sanctioned him — opposing counsel's fees, a client refund, and a $1,000 donation — for failing his gatekeeping duty. Supervision means the attorney verifies every citation, no matter who drafted it.
Do paralegals need paid tools like Westlaw to do legal research? No. A working free stack — CourtListener and Google Scholar for cases, Cornell's Legal Information Institute and govinfo for statutes and regulations, and Fastcase through most state bars — covers the primary law most research needs. Paid and AI tools add speed, editorial treatment, and matter context, but the free tier is a real library, not a teaser.
CaseRead Team
AI-powered legal research built for practicing attorneys.