When to Outsource Legal Research and When to Do It Yourself

An Attorney Considering Whether to Outsource Legal Research

Legal research is one of the most consequential tasks in legal practice and one of the most frequently underestimated. The assumption that research is research, that a capable person with internet access can find what an attorney needs, has produced a growing body of sanctioned filings, overturned arguments, and embarrassed attorneys who trusted tools and people that were not up to the task.

This article is for attorneys who want a clear framework for when to do their own research, when to delegate it, and what questions to ask before they hand a research assignment to anyone. It is also informed by seven years of conducting legal research using Westlaw and LexisNexis across civil, criminal, immigration, family, and corporate matters in multiple U.S. jurisdictions. That experience has made one thing very clear: the difference between research that helps a case and research that creates liability is almost always a question of process, not effort.

The AI Problem Every Attorney Needs to Understand Before We Go Any Further

Before discussing when to outsource legal research or do it yourself, it is necessary to address the tool that a growing number of attorneys and self-described researchers are reaching for first: generative AI.

The cases are no longer hypothetical. They are decided, published, and increasing in number every year.

Mata v. Avianca and the case that changed the conversation

In 2023, a lawyer submitted a brief in the Southern District of New York that cited six cases. None of them existed. When the judge requested verification, the lawyer admitted he had used AI to prepare the brief and had even asked the AI whether the cases were real. The AI confirmed they were. The judge sanctioned the attorney and his colleague, ordering a $5,000 fine and mandating further legal education on the use of AI tools. The case, Mata v. Avianca, 678 F. Supp. 3d 443 (S.D.N.Y. 2023), became the reference point for AI hallucination liability in legal practice.

The sanctions have not stopped

Many courts have since issued dozens of orders sanctioning attorneys for including citations and statements of law fabricated by AI. The fallout for leading law firms caught inserting even a single AI-generated hallucination into a legal pleading has been described as catastrophic, almost like a data breach incident.

As recently as June 2026, the Ninth Circuit sanctioned two attorneys from Orange County, California, $2,500 each for submitting briefs in an immigration case containing citations to opinions that did not exist and fabricated quotes. The opinion cited a law article finding that AI tools from Westlaw and Lexis hallucinated 17% and 33% of answers, respectively, to a representative set of questions run in 2024.

In Johnson v. Dunn in the Northern District of Alabama, the court went further, disqualifying the defendants’ attorneys from the case entirely and referring the matter to the state bar. The judge noted that if fines and public embarrassment were effective deterrents, there would not be so many cases to cite.

In Alabama, the state Supreme Court sanctioned an attorney who had filed briefs laden with inaccurate AI-generated citations, including references to cases that did not exist. After being informed he had cited a fabricated precedent in one filing, the attorney promised it would not happen again, then cited nonexistent cases at the end of the very next sentence.

What this means for outsourced research

The AI sanction cases are not just a warning about tools. They are a warning about process. An attorney who outsources research to someone who uses AI as a primary source and does not verify findings through a credible database is not protected by the delegation. The attorney signed the filing. The attorney faces the sanction. Understanding this is the foundation of every decision that follows in this article.

Why Google Scholar and Justia Are Not Enough

Google Scholar and other generic databases like Justia are free, accessible, and covers a wide range of federal and state case law. For a general orientation to a legal question it has value. As a primary research database for work product that will be used in litigation or legal advice, it has a fundamental limitation that disqualifies it from serious research work.

There is no way to verify that a case is still good law

Westlaw has KeyCite. LexisNexis has Shepard’s Citations. These are citator services that track the subsequent history of every case in the database, flagging cases that have been overruled, distinguished, criticized, or limited by later decisions. When I Shepardize a case on LexisNexis, I can see at a glance whether a red stop signal indicates the case has been overruled, a yellow signal indicates it has been distinguished or criticized, or a green signal indicates it remains solid authority. KeyCite on Westlaw functions the same way with its own signal system.

Google Scholar or Justia have no equivalent. A case that appears in Google Scholar or Justia results may have been overruled by a subsequent decision in the same jurisdiction years ago. Without a citator, there is no reliable way to know. Citing an overruled case in a brief is not just an embarrassing error. Depending on the court and the circumstances, it can be sanctionable.

Coverage gaps in Google Scholar and Justia

Google Scholar and Justia’s coverage of state court decisions, particularly lower appellate and trial court decisions, is inconsistent. In practice, this means a researcher using these databases may miss controlling authority from an intermediate appellate court that directly addresses the issue being researched. Westlaw and LexisNexis carry substantially more comprehensive coverage of state court decisions across all fifty states, including unpublished opinions that, while not precedential in many jurisdictions, can be persuasive and are sometimes the only authority directly on point.

The verification process that separates real researchers from the rest

Seven years of conducting legal research with Westlaw and LexisNexis has taught me that the verification step is where most research either holds up or falls apart. The process is not complicated, but it is non-negotiable.

Every case that goes into a memo or a brief gets Shepardized or KeyCited before it leaves my hands. Every time. Not just the primary cases. Every case. A case that was good law when a memo was last updated six months ago may not be good law today. A case that a supervising attorney mentions from memory as a strong authority may have been distinguished so many times that its precedential value has eroded. The citator catches both. Skipping this step is not a time-saving measure. It is a liability transfer to the attorney who relies on the research.

The Jurisdictional Access Problem

This is the issue that attorneys evaluating a legal researcher should address before any other credential or rate discussion, and it is the one that gets skipped most often.

Database access is not universal

Westlaw and LexisNexis subscriptions vary significantly in what they cover. A contract legal researcher with a Westlaw subscription that does not include California primary sources cannot reliably conduct California legal research, regardless of how skilled they are. A subscription that covers federal case law but not state statutory annotations is not adequate for state law research. Before delegating a research assignment, an attorney needs to know exactly what jurisdictions and source types the researcher’s database subscription covers.

This is not a hypothetical concern. Database subscriptions are expensive, and many people offering legal research services have limited or partial access rather than the comprehensive coverage they may imply. A researcher who has access to federal case law but not California Reports or the California Code of Regulations is not equipped to support a California state court matter, even if their general research skills are strong.

Match the jurisdictional access to the assignment

A California attorney researching a state law issue needs access to California Supreme Court and Court of Appeal decisions, California statutory codes, California regulatory materials, and California secondary sources like the California Continuing Education of the Bar publications. Depending on the issue, they may also need access to sister state decisions from jurisdictions that have addressed the same question and whose reasoning California courts may find persuasive.

An immigration attorney working in federal court needs comprehensive coverage of the circuit court decisions for the relevant circuit, Board of Immigration Appeals decisions, and the Code of Federal Regulations. A family law attorney working in Texas needs Texas Family Code annotations, Texas Supreme Court and Court of Appeals decisions, and the secondary sources that explain how Texas courts have applied specific statutory provisions.

Before you outsource legal research to anyone, the question to ask is specific: do you have current, comprehensive access to the primary sources in this jurisdiction for this type of law? The answer to that question lets you know whether the assignment is appropriate for that researcher, regardless of anything else on their resume.

Verify the access, do not assume it

A legal researcher who lists Westlaw or LexisNexis on their profile has told you they have some access to those platforms. They have not told you what that access covers. The follow-up question is straightforward: can you confirm your database subscription includes current access to California primary sources, including case law and statutory materials? A researcher with comprehensive access answers that question immediately and specifically. A researcher whose access is limited will either hesitate or give a vague answer that warrants further inquiry.

Five Scenarios Where Outsourcing Legal Research Makes Sense

With the foundation of database access and verification process established, here are the scenarios where delegating research to a qualified researcher is the right decision.

Scenario 1: The matter is in an unfamiliar jurisdiction

If your practice is concentrated in California, you may occasionally handle a matter with significant Texas law components, or represent a client in a federal circuit where the case law landscape is unfamiliar. Conducting research in an unfamiliar jurisdiction is slower, more prone to missing significant authority, and more likely to produce a memo that identifies what the cases say without adequate understanding of how the courts in that jurisdiction have actually applied the law over time.

A researcher with demonstrated experience in that jurisdiction and verifiable database access to its primary sources can produce more reliable results in less time than an attorney researching outside their home jurisdiction. The key qualifier is demonstrated experience. For example, a researcher who has worked in Texas courts before knows that the Texas Supreme Court and the various Courts of Appeals do not always speak with one voice, and that which court decided a case often matters as much as what the case says.

Scenario 2: High-volume research across multiple issues

A complex litigation matter may require research across ten distinct legal issues simultaneously. A class action, a multi-defendant criminal case, or a large transactional matter with multiple regulatory questions can generate a research load that exceeds what a solo attorney can absorb while also managing client communication, court appearances, and the other demands of an active practice.

Delegating defined research assignments to a qualified researcher allows the attorney to divide the research workload without sacrificing quality on any individual issue. The critical discipline in this scenario is assignment specificity. A research assignment that says “research privacy law” returns something different from every researcher. An assignment that says “identify controlling authority in the Ninth Circuit on the third-party doctrine as applied to cell phone location data, with particular attention to decisions following Carpenter v. United States” returns something the attorney can actually use.

Scenario 3: Time-sensitive research with a hard deadline

Legal deadlines do not adjust for workload. A response to a motion for summary judgment due in ten days, a brief due in a circuit court in two weeks, or an emergency injunction application due tomorrow all require research to be completed on a timeline that may not be compatible with the attorney’s existing commitments, some of which may be unforeseen.

A researcher who can be activated immediately, who has the database access to work efficiently, and who understands the verification obligations of the assignment is a genuine operational asset in these situations. The risk in time-sensitive research is that speed pressure produces shortcuts. If you’re delegating urgent research, make the verification requirement explicit in the assignment: every case cited in the memo should be Shepardized or KeyCited, and the memo should note the date on which verification was performed.

Scenario 4: Research into emerging or rapidly changing areas of law

Some areas of law move fast enough that the landscape can shift significantly between the time a researcher starts an assignment and the time the memo is delivered. Privacy law, AI regulation, cryptocurrency and digital asset law, and certain areas of immigration enforcement are all examples of fields where the regulatory and case law landscape has changed substantially within short periods.

Research into these areas benefits from a researcher who follows developments actively, not one who is encountering the field for the first time through the assignment. The verification step matters especially here: a case that was decided six months ago in a fast-moving area may already have been distinguished or limited. Shepardizing recent decisions in emerging areas of law is not a formality. It is how a researcher confirms that a decision cited as current authority has not already been undermined by a later development.

Scenario 5: Research supporting a practice area the you are entering for the first time

Attorneys expand their practice areas. A family law attorney who begins handling immigration matters for existing clients, a corporate attorney who takes on an employment discrimination case as a favor to a longstanding client, or a solo practitioner who sees opportunity in a practice area adjacent to their existing work all face the challenge of doing legal work in territory they have not fully mapped.

Research in an unfamiliar practice area is a matter of understanding the framework, identifying the controlling statutes and regulations, recognizing which courts have jurisdiction over what types of claims, and knowing which secondary sources provide the most reliable orientation to the field. A researcher with experience in the target practice area can provide you with that orientation in the form of a structured memo that does the mapping work before diving into the specific issue, which saves the attorney significant time in getting up to speed.

Five Scenarios Where You Should Do the Research Yourself

Outsourcing legal research is not always the right answer. There are situations where the attorney’s direct involvement in the research produces better outcomes than delegation.

Scenario 1: The issue is strategically central to the case theory

When a legal question is at the heart of the case strategy, your direct engagement with the research is part of the strategic thinking, not separate from it. Research that you do yourself produces different insights than research that arrives as a completed memo. The process of moving through the cases, reading the reasoning, noticing what the courts emphasized and what they dismissed, and sitting with the contradictions in the authority informs strategic judgment in ways that a summary cannot fully replicate.

For the central legal theory of a significant matter, your time spent in the research is not inefficiency. It is engagement with the most important analytical work in the case.

Scenario 2: The matter involves highly sensitive client facts

Some matters involve client facts so sensitive that limiting the number of people who have access to them is a priority. A criminal defense matter with significant factual complexity, a family law matter involving allegations that the client has never disclosed publicly, or a corporate matter involving confidential business information may all fall into this category.

Research that requires the researcher to understand the specific factual context of the matter in detail is research that requires sharing that context. When the facts themselves are the most sensitive element of the matter, keeping the research in-house controls who has access to them.

Scenario 3: The question is narrow enough to answer quickly

Not every research question warrants a full delegation. A specific statutory question in a familiar jurisdiction, a quick check of a circuit split that you are already aware of, or a verification of a single case’s subsequent history are all tasks that, with database access, you can resolve in fifteen to thirty minutes. The time spent briefing a researcher on the assignment, reviewing the memo, and potentially following up with clarifying questions exceeds the time the research itself would take.

The judgment call here is honest: will this research take me more than an hour? If not, doing it yourself is almost certainly the more efficient path.

Scenario 4: The research requires real-time collaboration with the client

Some matters involve a client who is actively participating in the development of the legal strategy. A business client’s General Counsel making a real-time decision about a regulatory question, a client in an ongoing negotiation where the legal position is shifting, or a client who needs to understand the research to make an informed decision about next steps all benefit from the attorney’s direct engagement with the research rather than a mediated presentation of someone else’s findings.

Scenario 5: You are in a jurisdiction where your existing database access covers everything needed

If you have comprehensive Westlaw or LexisNexis access for the relevant jurisdiction and the research question is within your practice area, doing the research yourself eliminates the briefing time, the review time, and the dependency on another person’s process. A well-resourced attorney in their home jurisdiction, researching a question within their expertise, is often the most efficient researcher for that particular assignment.

What to Demand From a External Legal Researcher Before You Hire Them

The questions below are the ones I would want an attorney to ask me before delegating a research assignment, and the ones I would want them to ask anyone else.

Database access, specifically

Ask which databases they have access to and which jurisdictions those subscriptions cover. A researcher who lists Westlaw and LexisNexis as credentials without being able to specify their coverage is telling you less than you need to know. The follow-up is: can you confirm current access to primary sources in the specific jurisdiction of this assignment?

The verification process they use

Ask how they confirm that a case is still good law before including it in a memo. The answer should reference Shepard’s Citations on LexisNexis or KeyCite on Westlaw, and it should describe the process as standard rather than optional. A researcher who describes verification as something they do when they have time or when a case seems important is describing a process that will eventually produce a problem.

Their experience in the relevant practice area and jurisdiction

General legal research experience is not the same as experience in family law in Texas. Ask specifically whether they have conducted research in the practice area and jurisdiction of your assignment. Ask for examples of research they have done in similar matters. A researcher with seven years of experience conducting research across civil, immigration, family, and corporate matters in California, New York, Texas, Illinois, and Oregon can speak to those jurisdictions specifically. A researcher without that background should say so.

Their memo format and delivery timeline

Ask to see a sample research memo before committing to an engagement. The memo format tells you a great deal about whether the researcher organizes findings in a way that is useful for attorney review, whether they distinguish controlling from persuasive authority, whether they flag uncertainty rather than projecting false confidence, and whether they understand the difference between what the cases say and what they mean for your specific issue.

Their policy on AI use in research

Given the sanction record that now exists, this is not an optional question. Ask directly whether they use AI tools in the research process and if so, how. A researcher who uses AI as a starting point for orientation and then conducts all verification and authority confirmation through Westlaw or LexisNexis is describing a defensible process. If they use AI to generate citations and do not independently verify every one of them through a credible database, they describing exactly the process that has gotten attorneys sanctioned.

The Process Behind Research That Actually Holds Up

For any attorney reading this who wants to understand what rigorous legal research looks like from the inside, here is the process I follow on every assignment.

The first step is clarifying the question before opening a database. A vague research question produces vague research. Before beginning, I confirm the precise legal issue, the jurisdiction, the procedural posture of the matter, and any deadlines. If the assignment is unclear, I ask before I start, not after I have spent time going in the wrong direction.

The second step is secondary sources before primary sources. I use secondary sources, including practice guides, legal encyclopedias, and jurisdiction-specific treatises available through LexisNexis or Westlaw, to orient myself to the legal landscape before diving into case law. This prevents the mistake of building a research structure around the first case that appears to be on point without understanding whether it represents the mainstream of authority or an outlier.

The third step is identifying controlling authority in the jurisdiction. I search for decisions from the highest court in the jurisdiction first, then intermediate appellate courts. I note which courts have mandatory authority over the issue and which have only persuasive authority.

The fourth step is Shepardizing or KeyCiting every case before it goes into the memo. Every case. Not just the ones I am unsure about. Not just the ones that seem old. Every case, on the day the memo is delivered, so that the verification reflects the state of the law as of the date the research is used.

The fifth step is researching the counterarguments with the same rigor as the primary position. An attorney needs to know where the weaknesses in their argument are before opposing counsel finds them. A memo that presents only the favorable authority is a memo that leaves work for the attorney to do before they can rely on it.

The sixth step is flagging uncertainty honestly. When authority is thin, when the circuits are split, when the issue is unsettled in the relevant jurisdiction, the memo says so. Projecting false confidence about an unsettled legal question does not help an attorney make strategic decisions. Honest assessment of the strength of the authority does.

The Bottom Line

Outsourcing legal research makes sense in a significant number of situations. It does not make sense when the person doing the research does not have verifiable access to the right databases for the right jurisdiction, cannot describe a rigorous verification process, or is relying on AI-generated output that has not been confirmed through a credible citator.

The attorneys who have been sanctioned for AI-hallucinated citations were not cutting corners out of malice. They were trusting tools and, in some cases, people who were not equipped for the responsibility they were given. The standard for legal research has not changed because new tools have arrived. A case cited in a brief needs to exist, needs to say what the brief claims it says, and needs to still be good law. Meeting that standard requires access to Westlaw or LexisNexis, a disciplined verification process, and the professional judgment to know the difference between research that holds up and research that only looks like it does.

If you are looking for legal research support that uses Westlaw and LexisNexis, Shepardizes every case before delivery, and has seven years of experience across multiple U.S. jurisdictions and practice areas, feel free to schedule a free discovery call.