AI legal research tools typically combine a search of legal materials with a generative model that turns selected sources into a conversational answer. The answer and its citation links are a starting point—not proof that the cited case is real, applies to the question, supports the stated proposition, or remains good law. Those checks still require a researcher to open and assess the authorities.
How does AI find relevant case law?
Many legal research assistants let a user ask a question in ordinary language. The system then identifies candidate material from a legal content collection and uses a generative model to compose an answer based on some of that material. This is often described as retrieval-augmented generation, or RAG. It is a useful high-level description, not a claim that every product uses the same architecture or reveals all of its search and ranking steps.
](https://m.media-amazon.com/images/I/515uFKS7D4L._SL160_.jpg)
1. It interprets the question
The system has to identify the legal issue behind the user’s wording. That can be harder than matching keywords. A question may contain a mistaken premise, combine several issues, or use a phrase that resembles a different legal doctrine. If the tool misunderstands the issue, it may retrieve material that sounds relevant but does not answer the question.
2. It retrieves candidate sources
The tool searches material available in its collection. Depending on the product and subscription, that material may include cases and other primary law, secondary sources, editorial content, or practice guidance. Coverage can differ by jurisdiction and content type. Public product descriptions do not disclose every ranking detail, and a vendor’s description of its collection is not independent proof that the collection is complete for a particular question.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
3. It synthesizes an answer and may attach citations
A generative model drafts a response using selected material. Some products provide links to source documents or other citation signals. For example, Thomson Reuters says CoCounsel Legal is grounded in Westlaw, Practical Law, and firm knowledge and includes linked citations. Lexis describes linked citations and Shepard’s verification features. Thomson Reuters describes Westlaw Deep Research as using Westlaw and Practical Law tools and content, including primary law, administrative materials, secondary sources, and current awareness. These are vendor descriptions of their products; available features and content depend on the product and subscription.
A link gives the researcher a route to inspect a document. It does not, by itself, show that the answer accurately states the decision, that the source is relevant to the specific issue, or that the authority remains controlling.
Rank #2
Can AI verify a legal citation?
AI can help surface citations and, in some products, provide access to citator tools. It cannot replace the checks needed to establish whether an authority is usable. A citation may point to a real case and still be unsuitable because it comes from the wrong jurisdiction, addresses a different issue, or has been limited or overruled. A correct case name also does not establish that the generated sentence faithfully describes the holding.
Shepard’s and KeyCite can help identify citing references and signals about subsequent treatment. Read those signals as prompts for further review, not as a final answer about whether a case controls your issue. Check the decision itself and its legal context.
Rank #3
How do I check whether an AI-cited case is real and still good law?
For each authority that materially supports the answer, follow the citation into the source and check it against the question you are researching.
- Confirm the document. Open the cited source and verify that the case exists and that the court, date, case name, and citation match.
- Check jurisdiction and legal context. Confirm that the court and governing law are relevant, and that the case addresses the issue and procedural context at hand.
- Read the supporting passage. Locate the text said to support the answer. Compare the generated description with the court’s actual words and reasoning; do not rely on a summary or citation link alone.
- Check subsequent history and treatment. Use an available citator, such as Shepard’s or KeyCite, to look for later proceedings and decisions that discuss, limit, distinguish, or overrule the case. Read the relevant later materials rather than treating a status indicator as a conclusion.
- Assess the legal fit. Decide whether the authority’s holding applies to the facts and issue you have, and whether other relevant authority changes the analysis.
Thomson Reuters advises users: “Use Deep Research to accelerate your research, not to replace it.” Its help documentation also warns that Deep Research can be inaccurate and should not be relied on without additional research.
Where can AI legal research go wrong?
Errors can happen before, during, or after retrieval. The peer-reviewed evaluation by Magesh et al. describes systems retrieving authority about “moral turpitude” when asked about the distinct “moral wrong doctrine.” It also reports mistakes involving authority from the wrong jurisdiction or legal context, and generated descriptions that misstated a court passage. The underlying problem is not limited to a model inventing a citation: finding the right source requires identifying the actual legal issue first.
- Issue mismatch: the tool answers a similar-sounding question instead of the one asked.
- Jurisdiction mismatch: it retrieves authority that does not govern the relevant jurisdiction.
- Synthesis error: it finds a real source but overstates or misdescribes what the court decided.
- Incomplete research: a plausible answer may omit a central authority or important later treatment.
The Law Society of England and Wales warns that members have encountered cases that, when checked, “have turned out to be a fake citation, a misrepresentation of a document, or even a piece of legislation from another jurisdiction incorrectly described as English and Welsh law.” That warning reinforces why citation checking must include the source’s substance and jurisdiction, not just whether a link opens.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallBest Value
What do published evaluations tell us about accuracy?
Published results show meaningful error risk, but they describe specific evaluations—not a universal error rate for every legal AI product or query. Magesh et al.’s 2025 peer-reviewed paper reports hallucination rates between 17% and 33% across the three tools in its benchmark. The New York State Unified Court System Advisory Committee’s 2025 report, discussing the Stanford evaluation, gives a 17% hallucination rate and 65% accuracy rate for the tested Lexis product, and a 33% hallucination rate and 42% accuracy rate for the tested Westlaw product.
Those figures belong to the systems and query sets evaluated. They do not predict the result of an individual query or establish the performance of a current product release. The products and features have changed since the studies were conducted, so the reported rates should not be generalized beyond those tests.
The same New York report describes summer 2024 trials of AI-enhanced legal research platforms involving nearly 100 judges, court attorneys, law clerks, and law librarians. Participants saw possible time savings in preliminary research tasks, including finding on-point sources and preparing first drafts, while also finding the tools imperfect and in need of review and correction. The Advisory Committee states: “Even when using the AI-enhanced features that have been incorporated into established legal research platforms, any content generated by AI should be independently verified for accuracy.”
How should you compare AI legal research tools?
Compare the research workflow and source access available to you, rather than relying on a general claim that a tool is accurate. Ask:
- Which jurisdictions and types of legal material are included in your subscription?
- Does the answer link to primary authority, and can you locate the passage supporting each important proposition?
- Is a citator available, and can you inspect subsequent treatment and history?
- Can you see the source trail and understand what materials informed the answer?
- How does the system handle a mistaken premise, procedural posture, or jurisdictional limit?
- What human review and information-security controls does your practice require?
Vendor pages can explain features and content claims. Independent evaluation and inspection of the underlying authorities are needed to assess how well a tool performs for a particular task.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
