This is general information, not legal advice, and reflects Canadian court and tribunal rules and publicly available legal commentary as of September 8, 2026. Effective dates and rule text below are drawn from each body's own published source or a named secondary report, linked throughout; confirm current text directly with that court, tribunal, or regulator before relying on it.
A different problem than fake citations
Two other pages on this site already track fake case-law citations: courts finding that a factum or motion cited cases that don't exist, or quoted real cases inaccurately. That's a drafting-and-research problem — AI inventing legal authority. This tracker covers something else: AI-generated or AI-altered material submitted as the evidence itself — a chatbot's analysis treated as expert opinion, a fabricated passage presented as a document excerpt, an entire research submission an AI produced end to end.
A July 6, 2026 analysis by the Canadian legal-affairs blog Slaw identified at least 28 Canadian decisions involving this broader category of AI misuse beyond fake citations — 25 of the 28 involving a self-represented litigant, 20 of the 28 from tribunals rather than courts. Seven of those rulings, named below, involve AI-generated evidence specifically or a hallucinated legal proposition offered with no citation attached at all.
Seven rulings, named
| Ruling | Court / Tribunal | Category | What happened |
|---|---|---|---|
| Iida v. Meadow Lane Equine Clinic Ltd, 2025 BCCRT 626 | BC Civil Resolution Tribunal | AI-generated evidence | Chatbot-generated veterinary-injury analysis, offered as supporting material for the claim, was rejected. |
| Re X Corp., 2025 BCCRT 1228 | BC Civil Resolution Tribunal | AI-generated evidence | A quoted passage attributed to a paragraph that doesn't exist in the underlying material was rejected. |
| Ren v. Area 09, 2025 BCPAAB 20253891 | BC Property Assessment Appeal Board | Costs consequence | AI-related misuse in the party's materials produced the highest costs award identified in Slaw's dataset: $910. |
| Lee v. AutoCanada MR Motors GP Inc., 2026 BCCRT 854 | BC Civil Resolution Tribunal | AI-generated evidence | Material generated by both ChatGPT and Claude was disregarded as evidence entirely. |
| Wyszynski v. The Owners, Strata Plan BCS1721, 2026 BCCRT 851 | BC Civil Resolution Tribunal | Hallucinated legal proposition | A self-represented litigant's AI-sourced legal propositions, unconnected to any real authority, were dismissed. |
| Chen v. The Owners, Strata Plan LMS3094, 2026 BCCRT 823 | BC Civil Resolution Tribunal | Hallucinated legal proposition | Fabricated legal interpretations offered in argument, with no real authority behind them, were dismissed. |
| Côté et Placements JF Desgagné inc., 2026 QCTAT 1311 | Tribunal administratif du travail (Quebec) | AI-generated evidence | An entire AI-generated causation-research submission on a workplace-injury claim was dismissed outright. |
Source: Slaw, "Beyond Fake Cases: The Other Ways AI is Going Wrong in Canadian Courts," July 6, 2026. These are 7 of the 28 decisions Slaw's analysis identifies, not the complete set; verify any citation directly on CanLII before relying on it.
What the table above deliberately leaves out
Slaw's dataset also includes a third bucket — fabricated quotations attributed to real, existing cases (as opposed to invented case names). That fact pattern sits closer to the fake-citation problem our companion tracker already covers, so it's excluded here to keep the two pages' claims from overlapping. See our fake-citation case tracker for that dataset, and our Ontario disclosure-duty tracker for the certification obligations that apply when AI touches drafting rather than evidence.
Three approaches, side by side
No Canadian court or tribunal has adopted a formal admissibility test for AI-generated evidence — there's no single national rule to check. What exists instead is a patchwork of outright bans, disclosure duties, and silence, and which one applies to your file depends entirely on where you're filing.
| Jurisdiction / Body | Approach to AI-generated evidence | Effective | Source |
|---|---|---|---|
| BC Civil Resolution Tribunal | Outright ban. Standard Rule 8.3(9) (and IIPO Rule 10.3(5)) prohibits submitting "fake evidence including evidence created or altered by an AI tool"; giving false or misleading evidence is an offence under the Civil Resolution Tribunal Act, with costs consequences. | Feb 1, 2026 | civilresolutionbc.ca |
| Federal Court of Canada | Disclosure duty, not a ban. Any AI-generated content in a filed document must carry a declaration in the document's first paragraph; doesn't prohibit AI-generated evidence outright. | Dec 20, 2023 (updated May 2024) | fct-cf.ca |
| Ontario courts, tribunals & regulators | Verification and certification duties layered across five separate bodies (Superior Court, Ontario Land Tribunal, a Rules of Civil Procedure regulation, the LSO, Legal Aid Ontario) — no evidence-specific ban among them. See our full Ontario tracker for the five duties individually. | 2024–2026, various | Our Ontario tracker |
| Most other Canadian courts and tribunals | No AI-generated-evidence-specific rule identified as of this post's research date. General rules of evidence and existing electronic-record authentication provisions still apply by default. | — | — |
Table verified against each body's own published source or a named secondary report on September 8, 2026. Rules change fast in this area; check the linked source before relying on a date or requirement above.
Before AI-touched material becomes evidence in your file
- Trace it back to a human-verifiable source first. Before any AI-assisted document, image, or analysis goes into an affidavit, exhibit list, or evidence bundle, confirm where it actually came from.
- Don't let a chatbot's output pass as expert analysis. Iida and Lee v. AutoCanada above are what happens when that line blurs — the tribunal disregarded the material entirely rather than weighing it.
- Check whether your specific court or tribunal has an evidence-specific rule yet. As of this post, only BC's CRT does; everywhere else runs on a disclosure duty or no AI-specific rule at all — the table above is the fast check.
- Ask directly when a client or witness supplies material. A photo, recording, or document handed to you isn't necessarily untouched — ask whether any AI tool generated or altered it before you rely on it as if it came straight from the source.
- Treat "no rule yet" as a gap, not a green light. Six of the seven rulings above happened without any jurisdiction requiring an evidence-specific ban — a decision-maker rejected the material anyway, on the merits.
Where to go deeper
This tracker deliberately stays narrow: seven rulings, one new rule, three approaches compared. For the fake-citation half of this problem — case names that don't exist, not evidence that was fabricated — see our fake-citation case tracker. For Ontario's five separate drafting-disclosure and certification duties, see our Ontario courts & tribunals tracker. For how each provincial and territorial law society is treating AI use generally, our province-by-province table covers all 13. And if litigation technology is the broader question, our complete guide to litigation practice management software in Ontario covers the full buyer's-decision picture.
Not ready to subscribe? Join the DROZlegal waitlist instead.