KeyCite vs Shepard's: Which Citator Catches More Bad Law?

KeyCite and Shepard's do the same job on rival platforms: they tell you whether a case has been reversed, overruled, criticized, or otherwise weakened by later decisions. KeyCite runs on Westlaw, Shepard's runs on LexisNexis, and neither is meaningfully more accurate than the other. The most cited study on the question found each missed or mislabeled roughly a third of negative citing relationships. If a case is load-bearing in your brief and you can reach both, check both.

Most comparisons of these two are written by one of the two vendors, which tells you how much weight to give them. What follows is the practitioner version: what each tool signals, where they genuinely differ, what the independent research says about accuracy, and what to do if you cannot afford either.

What is the difference between KeyCite and Shepard's?

The difference is mostly platform and presentation, not capability. Shepard's is the older name and the origin of the verb: lawyers said "shepardize" long before either service went online, back when it meant pulling a bound volume of Shepard's Citations off the shelf. It now runs on LexisNexis. KeyCite is Thomson Reuters' answer, built into Westlaw, and it uses a flag system instead of Shepard's signals.

Both take a citation, pull every later document that cited it, and classify the treatment. Both separate direct history (what happened to this case on its own appeal) from citing references (what other courts said about it afterward). Both let you filter by treatment type, jurisdiction, and depth of discussion. If you learned one, you can use the other in about twenty minutes.

The differences that show up in daily use are narrower than the marketing suggests:

• Organization by jurisdiction. Shepard's sorts citing references by jurisdiction and court by default. This is the single most useful practical difference, because the only treatment that can actually bind you comes from a court above you in your own jurisdiction. KeyCite makes you filter to get there.
• Alert summaries. Shepard's offers summarized alerts when treatment changes on a case you are tracking. If you are monitoring a live issue across months, that matters.
• Editorial versus algorithmic classification. Lexis has long argued that Shepard's citing references are editor-analyzed while much of KeyCite's classification is machine-generated. Thomson Reuters disputes the characterization. Treat this as a vendor argument rather than settled fact, because the accuracy research does not show a clear winner.
• Overruling Risk. KeyCite added a feature on the Westlaw Edge tier that flags cases implicitly undermined by later rulings on a different point of law, the sort of quiet erosion a conventional flag misses. Shepard's has no exact equivalent. It sits behind a premium tier, so whether you have it depends on your contract.

What do the KeyCite flags and Shepard's signals mean?

The symbol systems look different and mean roughly the same things. This is the translation table worth keeping:

Here is the part that trips people up, and it trips up experienced lawyers too. A red flag does not mean the case is dead. It means at least one point in the opinion is no longer good law. Long appellate opinions cover many issues, and a case can be overruled on an evidentiary holding you do not care about while remaining the controlling authority on the proposition you are actually citing. The flag is an index. The answer is in the citing references.

The reverse is also true and more dangerous: a clean flag is not proof the case is safe. Which brings us to the research.

Which citator is more accurate, KeyCite or Shepard's?

Neither, in any way that should change your purchase decision. The best independent evidence is Paul Hellyer's study in volume 110 of the Law Library Journal (2018), which reviewed 357 citing relationships that at least one citator had labeled negative and compared how Shepard's, KeyCite, and BCite handled each one.

The findings, stated plainly:

• Shepard's and KeyCite each missed or mislabeled about one third of the negative citing relationships.
• BCite, Bloomberg Law's citator, missed or mislabeled more than two thirds.
• Serious errors, the kind that would matter in a filing, appeared in all three.

Read that carefully, because it is easy to draw the wrong conclusion. It does not mean citators are unreliable and you should skip them. They remain the only systematic tool for this job, and doing it by hand is worse. What it means is that the treatment symbol is a starting point rather than a verdict, and that the two market leaders disagree often enough that running a critical case through both will surface negative treatment that either one alone would have hidden from you.

If you have access to both platforms, through a firm subscription plus a bar benefit, or a co-counsel relationship, use both on the cases your argument depends on. If you have one, use it properly: open the negative citing references and read them rather than trusting the color.

Do KeyCite and Shepard's cover the same cases?

Broadly yes for US federal and state case law, which is where the overlap is nearly complete. Both cover published federal and state appellate decisions comprehensively. Divergence appears at the margins: unpublished dispositions, administrative decisions, some older state material, and secondary sources that one platform licenses and the other does not. Each also cites into its own ecosystem, so Shepard's surfaces treatment in Lexis-exclusive secondary sources and KeyCite does the same for Westlaw's.

For the everyday question of whether a case is still good law, the coverage difference is not the deciding factor. The classification difference is, and that is what the Hellyer numbers describe.

Is there a free alternative to KeyCite and Shepard's?

There is no free product that matches either one. There are two workable substitutes, and one of them you may already be paying for without knowing it.

Check your bar membership first. More than sixty state and local bar associations include a Fastcase or vLex library in the dues you already pay, and it comes with Authority Check, a citator that flags negative treatment and counts citing references. It is not as deep as Shepard's or KeyCite, but it is a real citator and it costs you nothing additional. A remarkable number of solos buy research access their bar was already providing. Clio completed its acquisition of vLex in November 2025, so watch for changes to what your bar benefit includes at renewal.

Read subsequent history manually. CourtListener and Google Scholar both let you see cases citing a given decision. You can work through them yourself. This is genuinely slower, requires you to make the treatment judgment a citator editor would otherwise make, and is easy to get wrong on a long citation list. It is fine for a preliminary read. It is not appropriate as the last check before a filing.

What is not an option anymore is SmartCite, Casetext's citator, which was the affordable middle ground for a lot of solo practitioners. It retired with Casetext on April 1, 2025, about eighteen months after Thomson Reuters bought the company for 650 million dollars, and its function now reaches users through Westlaw and CoCounsel Legal under the KeyCite flags. We covered what happened and where users went in our piece on the Casetext shut down.

Which one should you buy?

If the choice is genuinely open, buy the platform, not the citator. Neither citator is good enough to justify a platform you otherwise dislike, and the accuracy research says as much. Decide on coverage, secondary sources, judge analytics, drafting tools, and the quoted price, then accept whichever citator comes attached. Our Westlaw vs LexisNexis comparison works through that decision, and the legal research platforms comparison covers the field including the vendors that will not publish a rate at all.

Two situations do tip the choice toward Shepard's. If you appear regularly before a specific court and need to see fast whether a binding court has treated a case negatively, the jurisdiction-first organization saves real time on every check. And if you monitor issues over months rather than checking cases once, the alert summaries are better. Two tip toward KeyCite: if you already live in Westlaw's secondary sources, and if Overruling Risk is included in your quote, because implicit erosion is a genuine blind spot in ordinary citator logic.

The workflow that actually protects you

Finding a case and validating a case are separate jobs, and conflating them is how bad citations reach filings. The find step is about search: getting to the on-point authority without having to guess the exact words the court used. The validate step is about history: confirming that authority still stands. A tool that is excellent at one may do nothing for the other.

A sequence that holds up:

1. Find the on-point authority, however you find it fastest.
2. Read the opinion itself, not a summary of it, and confirm the holding says what you think it says.
3. Verify the citation against the official reporter.
4. Run it through a citator, open the negative citing references, and weight them by jurisdiction.
5. For anything load-bearing, run it through a second citator if you can.

Step four is the one people skip under deadline, and it is the one that produces sanctions. Courts have disciplined lawyers for briefs resting on cases that had been overruled, and for cases that a general-purpose AI tool fabricated outright. Our note on where AI legal research goes wrong collects those examples. The lesson is the same regardless of what you use to find the case: verify the citation, then validate the case.

This is also why firms that handle the research well tend to handle their own institutional memory well. The brief where a colleague already worked through this exact issue is usually sitting in a shared drive somewhere, and being able to find any answer across your own files often saves more hours than any citator feature. Past work product is the cheapest research library you own, and the least used.

Cases handles the first step of that sequence and is deliberately explicit that it does not handle the fourth. Ask your question in plain English, get on-point US federal and state precedents back as headnote cards with the holding surfaced and a real citation to check. Then take that citation into whichever citator you have access to. Our guide to running a citator check covers every option and what each one costs, and how to shepardize a case walks the Lexis workflow step by step.

All of this is informational research, not legal advice. Verify every citation in the official reporter before you rely on it.