In McCandlish Farmhouse Confectionery Limited and Anr v Michael Malone and Anr [2026] CSOH 85, Lord Sandison's judgment is ostensibly about trademark infringement and passing off, but it contains a clear running criticism throughout: AI-generated material is alone not enough to build a legal case that can withstand the tests applied by the courts.
At the conclusion of proof, Lord Sandison reiterated his criticisms, branding the claim "a salutary lesson in how AI may prompt lay persons to say the right things in pleadings, but cannot furnish the evidence needed to establish those statements". Later, in dissecting the pursuer's quantification of damages, he observed that "all of these claims existed in the AI realm only".
The mere attempt by lay representatives to advance litigation should not be criticised; full access to the courts is a fundamental aspect of the Rule of Law. AI models can be helpful within that process. However, the judgment draws a clear and important line. AI models may also be genuinely useful for structuring an argument, suggesting the right legal categories to plead, or producing a plausible outline of submissions, particularly valuable for a litigant without legal training or resources.
However, despite its exponential advancement in recent years, AI still falls short of providing a fully-packaged case that is sufficiently comprehensive to withstand judicial scrutiny. The result, as this case shows, is pleadings and submissions that might initially read persuasively, but quickly fall apart in practice. This lesson is not exclusive to lay representatives and can equally be applied to solicitors.
Background
The dispute centred on the sweets "Chelsea Whoppers": chocolate-covered fudge fingers originating from the Second World War. The pursuer, a small family-run confectionery business, claimed trademark rights and goodwill in the product's name and packaging, and sued a sole-trader for infringement and passing off who was selling similarly-branded confectionary. Damages were "conservatively" estimated at over £1 million. The pursuer's case was largely advanced by a lay representative, a company director. The defenders also represented themselves.Legal context and decision
Lord Sandison ultimately found that one of the two trademarks relied upon had been infringed for a period, but declined to award any of the damages sought, and refused permanent interdict given an undertaking already offered by the defender to refrain from further producing the sweets. The passing off claim failed outright for lack of evidence of goodwill in the pursuer's product, or actual consumer confusion between the two. Central to this outcome was the court's finding that the pursuer's claims of brand dilution, reputational harm, and rebranding costs were unsupported by any evidence at all, with promised independent survey evidence never materialising.Criticisms of the use of AI
Lord Sandison observed early on that much of the material presented by the pursuer "had every appearance of being abstract and generic content produced by artificial digital means", and that the presentation of the case seemed "influenced at least as much by what that content indicated should be said as by demonstrable facts". This seemingly shaped the entire trajectory of the litigation. Affidavits made claims about both defenders' involvement, that prior court records from a previous claim directly contradicted. Repeated claims of market confusion arose solely from family testimony. The Pursuer placed heavy reliance on one particular email that, read in full, actually discredited their claim.At the conclusion of proof, Lord Sandison reiterated his criticisms, branding the claim "a salutary lesson in how AI may prompt lay persons to say the right things in pleadings, but cannot furnish the evidence needed to establish those statements". Later, in dissecting the pursuer's quantification of damages, he observed that "all of these claims existed in the AI realm only".
The mere attempt by lay representatives to advance litigation should not be criticised; full access to the courts is a fundamental aspect of the Rule of Law. AI models can be helpful within that process. However, the judgment draws a clear and important line. AI models may also be genuinely useful for structuring an argument, suggesting the right legal categories to plead, or producing a plausible outline of submissions, particularly valuable for a litigant without legal training or resources.
However, despite its exponential advancement in recent years, AI still falls short of providing a fully-packaged case that is sufficiently comprehensive to withstand judicial scrutiny. The result, as this case shows, is pleadings and submissions that might initially read persuasively, but quickly fall apart in practice. This lesson is not exclusive to lay representatives and can equally be applied to solicitors.
