Access to legal information matters. But for legal information to be truly valuable, users also need a quick way to understand what a document covers, why it may be relevant, and whether it deserves a closer look. That is where AI-generated summaries make a real difference. By helping readers grasp the subject matter and significance of a decision or statute more quickly, they make legal research easier to navigate and more efficient in practice.
At Lexum, we have been actively developing and refining our AI enrichments to strengthen the research experience for our clients end users. In Canada, that work takes on particular importance because legal research unfolds in a bilingual environment shaped by two official languages: English and French. On a platform like CanLII, that reality affects not only how legal information is consulted, but also how it is discovered. As we explained in our earlier article, A New Chapter Begins for Bilingual Legal Research on CanLII, documents available only in English historically did not surface in search results when users conducted their research in French. As a result, highly relevant English-language content could remain out of sight for legal professionals working only in French. This is precisely where bilingual AI-generated summaries have begun to change the picture. By generating summaries in both official languages and progressively indexing them into CanLII’s search engine, Lexum has helped reduce a long-standing barrier in Canadian legal research. Searches conducted in French can now surface documents available only in English, making it easier for legal professionals to work in their own language while still discovering relevant material that might not previously have surfaced.
Building on that progress, the latest rollout extends AI-generated summaries and enrichments across both case law and legislation in several jurisdictions. New bilingual summaries are now available for Federal, Nova Scotia, and Newfoundland and Labrador case law and legislation, while additional case law coverage in the second official language has been added for Nunavut, Yukon, and the Northwest Territories. British Columbia is also included with legislation summaries in the original language, as well as AI-generated summaries for a selected corpus of case law in the original language. The rollout also adds significant new case law coverage for Ontario in both official languages, along with additional AI enrichments such as keywords, titles, and short summaries. In Saskatchewan, Alberta, and Manitoba, AI-generated English summaries have been added for decisions available only in French. Here is a summary by jurisdiction for the latest projects:
| Jurisdiction | Collection | Summaries | Exclusions | Completion Rate |
| Federal | Case Law | 503 595 | 3 996 | 99% |
| Legislation | 12 126 | 1 187 | 90% | |
| Nova Scotia | Case Law | 118 511 | 6 250 | 95% |
| Legislation | 5 990 | 39 | 99% | |
| Newfoundland and Labrador | Case Law | 14 719 | 116 | 99% |
| Legislation | 2583 | 0 | 100% | |
| British Columbia | Case Law | 104 270 | 1 017 | 99% |
| Legislation | 3 229 | 872 | 73% | |
| Ontario | Case Law | 167 891 | 4 205 | 97% |
| Enrichment (Keywords) | 83 534 | 1 | 99% | |
| Enrichment (Titles and short summaries) | 85 527 | 5 | 99% | |
| Saskatchewan | Case Law | 14 | 0 | 100% |
| Alberta | Case Law | 48 | 1 | 99% |
| Manitoba | Case Law | 14 | 1 | 99% |
| Nunavut | Case Law | 1 130 | 9 | 99% |
| Yukon | Case Law | 3 824 | 76 | 98% |
| Northwest Territories | Case Law | 12 393 | 118 | 99% |
The scope of this work is considerable. As part of the projects listed above, more than 950,000 AI-generated summaries have now been added to CanLII’s case law and legislation collections, along with more than 169,000 additional enrichments for Ontario. Together, these additions make CanLII more useful for users conducting research across different jurisdictions, languages, and document types.
Just as important as the quantity of AI-generated summaries is their quality. Over the last few months, Lexum’s teams developed new optimized prompts for large language models to improve the clarity, consistency, and reliability of the generated summaries. The enhanced capabilities provided by the latest model led to concrete improvements in the handling of references, content structuring, and legal language, particularly in the naming of parties and the consistent use of terminology in bilingual contexts. This means that we are not only expanding coverage but also refining how useful these summaries are in practice for people reading legal documents.
This focus on quality is also reflected in the findings of an external study conducted by Lux Insights in December 2025. The study revealed that CanLII’s AI-generated summaries are primarily valued as a screening tool: they help users quickly determine whether a document is worth reading and provide a structured summary before conducting a more in-depth analysis. Users also expressed greater confidence in CanLII’s AI, as it links directly to the official text of the decision, avoids misquotes, and appears more reliable than other generic AI models.
This latest rollout shows how AI-generated summaries have now become an integral part of the legal research experience on CanLII. Their purpose is not to replace the reasoning of legal professionals, but to facilitate the early stages of their research, particularly by helping them identify relevant documents, understand their context, and determine which ones warrant special attention. As these improvements continue to be rolled out across all of CanLII’s collections, they will help more users find and evaluate legal information more effectively.
