AI tools for case law research keep getting better

Log in to listen to this article

Lawyers looking for AI-assisted case law research have, until recently, faced a stark choice. Pay a few hundred dollars a month for Westlaw Edge or LexisNexis' Protégé, or do without. Two new tools have changed that. Both are free, and one of them works inside Claude or ChatGPT.

In February, CanLII rolled out CanLII Search+, a limited integration of AI for searching the biggest and best free database of Canadian court decisions, statutes, and a growing body of secondary sources.

Continuous tweaks have made this an even better tool than when it first appeared, though some lawyers will find its daily use limits a hindrance.

A more powerful but still free alternative is an “AI Assistant” that works with a database of cases and statutes maintained by the Access to Algorithmic Justice (A2AJ) project, a joint initiative of faculty at the Osgoode Hall Law School and the Lincoln Alexander School of Law in Toronto.

A2AJ offers a plug-in that you connect — in one quick and easy step — to your ChatGPT or Claude account, letting you run legal searches within those apps. The tool is quick and effective in turning up relevant cases, and generates accurate summaries you can probe further through chat.

A2AJ's main limitation is coverage gaps, which it’s working to close.

More broadly, it points to a future where we do more of our legal research within our primary AI platforms.

CanLII’s clever integration of AI

CanLII Search+ enables you to pose a question in plain language and have AI automatically populate the keyword search field with a Boolean query. However, CanLII doesn’t do a great job of making Search+ visible.

It’s tucked into a narrow vertical sliver on the right side of your screen when you visit CanLII.org. I initially spent several minutes looking for it without success.

You have to log in to CanLII or create a free account to use it. Once you do, CanLII has a helpful guided tour to orient you.

As a beta tester using Search+ since the fall, I have found it highly effective at translating my plain English prompts into Boolean queries. The first crop of results has tended to surface what I’m looking for, reducing the time I now spend searching for case law on the site.

When you first run a query with Search+, the results are not well organized, and only some contain brief “AI-generated” summaries. But clicking the “Analyze” button, then “Rank by Search+ Score,” is where the magic unfolds.

The top 25 cases are reordered based on a relevance ranking out of five, and most impressively, you get a “Relevance Analysis” for each case, containing a roughly 200-word summary of which parts of the case, and precisely where, you’ll find the answer to your question. The summaries are consistently accurate and helpful, making this a great tool for quick queries.

But Search+ has a few key limitations. You’re limited to 10 search query generations and four relevance analyses per day. You still have to trawl through your results one by one. And once you pull up a case, you can’t use CanLII itself to dialogue with it in a chatbot.

A2AJ takes a different approach  

Rather than bolting AI onto a search portal, A2AJ brings its database to the AI platform you’re already using. It has no daily use limits. It generates more comprehensive summaries that you can dialogue with. And it’s faster and more reliable than Westlaw’s Edge or LexisNexis’ Protégé, which cost a few hundred dollars a month and are often inaccurate, on top of being slow and clunky to use.

A2AJ’s free connector lets you search their open-source dataset of Canadian court cases, statutes, and regulations. So far, coverage is extensive but not complete. It includes all Supreme Court of Canada decisions from 1877, all Federal Court cases for the past 25 years, and much of British Columbia, Nova Scotia, and the Ontario Court of Appeal. Big gaps remain, but A2AJ is working to fill them.

Even with the gaps, the tool works exceedingly well for searching the data it currently incorporates.

Queries you run in Claude or ChatGPT yield accurate, concise summaries of clusters of cases on point or apt summaries of lengthy appellate decisions. The quality and utility far exceed that of CanLII’s Relevance Analyses or, in my experience, Protégé or Edge for the searches I have run. And since you’re in a chatbot, you can continue the discussion by asking for more specific details.

On Claude, I ran multiple searches using the A2AJ connector of SCC cases I know well and was thoroughly impressed by the summaries. Queries about discrete issues under section 8 of the Charter turned up a host of B.C. trial decisions that were directly on point, helpfully summarized — and even followed by an analysis of underlying patterns in the decisions.

Rather than translating your query into a Boolean search string, A2AJ conducts a semantic search that grasps the meaning of your question with more nuance and precision. It's also faster and more accurate because it draws on its own case mapping rather than querying someone else’s portal.

A2AJ’s connector takes only a minute to set up and even works with the free tier of Claude.

The kind of solution A2AJ offers — a tool crafted to work with your main AI platform — looks to be the best way forward. With more lawyers doing more of their work with bigger, more versatile frontier models than with older legacy platforms like Westlaw, the best tool for research might be the one you’re already using.

Whichever tool you choose, the usual cautions still apply. You need to verify the cases, read the relevant passages yourself, and decide whether client information can be shared with an AI platform.

But the trajectory is clear. More powerful tools are becoming more accessible and arriving faster than we expect.