TL;DR: Legal research in the United States means working across two parallel systems - federal and state - each with its own statutes, regulations, and three tiers of courts. The fastest modern workflow pairs free primary sources such as CourtListener, Google Scholar, govinfo.gov, and supremecourt.gov with AI that scopes to the right jurisdiction, cites the exact page and quoted passage, and lets you confirm good law before you rely on it. The technology speeds the search; you keep the judgment.
American legal research has a deserved reputation for complexity. A single question - whether a non-compete is enforceable, say - can have fifty-one different answers depending on which state's law applies, layered on top of any federal rule that pre-empts it. Add a court hierarchy that runs from trial courts to the Supreme Court, a split between primary and secondary sources, and the constant duty to confirm a case is still good law, and it is easy to see why research consumes so many billable hours. This guide explains how AI changes the speed of that work without changing the standard of care, and how to do it with sources you can actually verify.
Why is US legal research uniquely layered?
The United States is a federal system, which means two sovereigns make law at once. Congress and the federal agencies produce statutes and regulations that apply nationally; each of the fifty states, plus the District of Columbia and the territories, has its own constitution, code, and body of common law. Federal law governs matters such as bankruptcy, patents, and federal crimes, while most contract, tort, property, and family law is state law. Where the two overlap, doctrines of pre-emption decide which prevails. For a researcher, the first question is never just what the law says, but which jurisdiction's law controls the matter in front of you.
That layering is the reason a careful researcher scopes before searching. A perfectly correct California holding is useless in a Texas dispute, and a federal circuit's reading of a federal statute may not bind a court in a different circuit. AI does not remove this complexity, but a tool that lets you scope to a jurisdiction and shows the deciding court on every result makes it far easier to stay inside the boundaries that matter.
How do the federal and state systems fit together?
Both systems share the same three-tier shape - trial court, intermediate appellate court, court of last resort - but they run in parallel, and authority flows down each separately.
The federal hierarchy
Federal cases begin in the U.S. District Courts, are appealed to one of the thirteen U.S. Courts of Appeals (the circuits), and may ultimately reach the Supreme Court of the United States. A Supreme Court decision binds every court in the country on a question of federal law. A circuit's decisions bind the district courts within that circuit but are only persuasive in other circuits - which is why a circuit split is often what draws the Supreme Court's attention. When you research a federal question, you therefore care intensely about which circuit you are in.
The state systems
Each state mirrors this structure with its own trial courts, an intermediate appellate court in most states, and a state supreme court whose word is final on questions of that state's law. A state supreme court binds all courts within the state; a decision from another state is merely persuasive. Names vary - in New York the trial court is confusingly called the Supreme Court - so confirming the deciding court and its level is part of reading any result correctly.
What is the difference between primary and secondary sources?
Primary sources are the law itself: constitutions, statutes, regulations, and judicial opinions. They are what binds a court and what you cite in a brief. Secondary sources - treatises, law review articles, practice guides, and legal encyclopedias such as American Jurisprudence - explain, summarise, and critique the law. They are invaluable for orienting yourself in an unfamiliar area and for finding the leading primary authorities quickly, but you generally do not rely on them as controlling authority.
A sound research method uses both in order. Start with a secondary source to understand the doctrine and identify the key cases and statutes, then move to the primary sources to confirm exactly what they say and whether they still stand. AI is well suited to this pattern: it can summarise an area the way a secondary source would, then point you to the primary authority behind each proposition - provided it shows you the source so you can read it yourself. For a broader treatment of AI research mechanics across jurisdictions, see our legal research tools guide; this guide stays focused on the US specifics.
Which free databases can you use for US legal research?
You do not need an expensive subscription to reach authoritative US primary law. A handful of free, reputable sources cover most of what a researcher needs, and they are exactly the kind of grounded sources an AI tool should rely on rather than its own memory.
| Source | What it covers | Best for |
|---|---|---|
| CourtListener | Millions of federal and state opinions, oral arguments, and dockets | Finding and reading case law, tracking dockets |
| Google Scholar | Case law and law review articles, with a citation graph | Quick case look-up and seeing who cited a case |
| govinfo.gov | The U.S. Code, Federal Register, CFR, and official federal documents | Authoritative statutes and regulations |
| supremecourt.gov | Slip opinions, orders, and the docket of the Supreme Court | The latest Supreme Court decisions |
The catch with free sources is that they are powerful but not stitched together - each has its own search interface, and none verifies whether a case is still good law. That is where an AI layer earns its place: it can search across grounded sources, retrieve the on-point passage, and present it with a pinpoint citation, leaving you to do the confirmation that no database does automatically.
How do you confirm a case is still good law?
Finding a case is only half the job. Before you cite it, you must know whether it remains good law - whether it has been reversed on appeal, overruled by a higher court, superseded by statute, or quietly undermined by later decisions. The traditional tools for this are commercial citators: Shepard's on LexisNexis (to Shepardize a case) and KeyCite on Westlaw, which flag negative treatment with signals. The concept is essential even if you do not have those subscriptions: you are tracing a case forward through its subsequent history to make sure the ground has not shifted beneath it.
No AI tool should be trusted to declare a case good law on its own, and Judicio does not pretend to be a citator. What good AI research does is make the manual check fast: it surfaces later decisions that cite your case, quotes how they treated it, and links you to the primary text so you can judge for yourself. Treat good-law verification as a deliberate, separate step every time, and never let a confident summary substitute for reading the subsequent history. Our guide on AI citation checking and good-law verification walks through a full workflow.
How do you cite US authority correctly?
American legal citation largely follows The Bluebook, the dominant style guide, with some courts using their own local rules or the ALWD guide. A case citation identifies the volume, reporter, and page - for example, 347 U.S. 483 points to volume 347 of the United States Reports at page 483 - usually with a pinpoint to the specific page you are relying on. Statutes cite the title and section of a code, such as a provision of the U.S. Code. Getting these details right signals care; getting them wrong, or citing a source that does not exist, signals the opposite.
AI can help compile and standardise citations, but the pinpoint matters: a citation is far more useful when it points to the exact page the proposition came from. A tool that already attaches page-level citations to every answer gives you most of what a court expects, leaving you to confirm the format against the relevant court's rules.
What is the hallucination risk, and how do you avoid it?
The most publicised danger of generative AI in law is the hallucinated citation - a plausible-looking case name and reporter string for a decision that was never handed down, or a real case quoted for a holding it never reached. This is not hypothetical. In the now-infamous Mata v. Avianca matter, lawyers filed a brief containing fabricated cases generated by a general-purpose chatbot and were sanctioned for it. We cover that episode in detail in the Mata v. Avianca story.
The defence is structural, not a matter of trusting a better model. Use a tool that grounds every answer in a retrieved source and quotes the exact passage, then open that passage and read it before you rely on it. General chatbots invent because they generate from memory; research tools that retrieve from real databases and cite the page let you catch any error in seconds. The rule is simple and unforgiving: never cite a case you have not read in the original. For the verification habit in full, see how to verify AI legal research.
How does jurisdiction scoping keep AI results relevant?
Because the US has more than fifty overlapping bodies of law, an unscoped search is a recipe for confident irrelevance. Scoping is the act of telling the tool which jurisdiction governs - a circuit, a state, the federal system - before it retrieves. Judicio's Legal Research lets you set the jurisdiction explicitly and will also auto-detect it from your question and confirm before assuming, so a query that mentions a California statute is researched against California law rather than a national average. When a matter spans forums, Deep Mode can explore up to five angles in parallel, but you still anchor each to the jurisdiction that controls it.
How does Judicio research US law?
Judicio approaches US research as a citation-first workflow. Its Legal Research draws on dedicated legal-database integrations - including CourtListener for US case law - among 33 jurisdictions with native connections, and reaches 100-plus jurisdictions elsewhere through curated legal web search. Every answer carries the exact page and the quoted passage it relied on, with deterministic citation labels, and every web source is archived as a permanent PDF at the moment it is retrieved, so a citation cannot quietly rot between research and filing. You can export an evidence pack of the whole set for your file or your supervising partner.
Just as important is what it does not claim. Judicio is a research and drafting assistant, not a substitute for your judgment, and its outputs are not legal advice. It will not tell you a case is good law on its own authority; it gives you the pinpoint citation and the subsequent decisions so you can decide. Because one upload into the File Library feeds every tool, the authorities you find can flow straight into Drafting or a timeline without re-uploading. You can see the whole feature set in one place.
How do you get started?
Begin with one real question from a live matter. Scope it to the right jurisdiction, ask it in plain language, and read the cited passages rather than the summary. Confirm each key case against the primary source on CourtListener or govinfo.gov, check its subsequent history, and only then carry it into a brief - formatted to the court's citation rules. Run that loop for a week alongside your normal process and compare the time spent against the reliability you get.
You can try the workflow on your own matters with Judicio's 7-day free trial - 500 credits, no credit card required - and move to the Professional plan at $200 per month for 5,000 credits when you are ready. For a walkthrough tailored to your practice, get in touch. For the full picture of how Judicio covers the country, see AI legal research for the United States. The tools do the finding and the reading at speed; the verification and the judgment stay with you.