Artificial intelligence is now being used to support diagnostics, triage, imaging interpretation, workflow prioritisation, and clinical decision-making across veterinary settings. As adoption accelerates, a fundamental question emerges: how should veterinarians evaluate the tools they are being asked to trust? At the core of that question sits transparency.
AI in Clinical Practice Is Not One-Size-Fits-All
AI diagnostic and clinical support tools do not perform uniformly across all scenarios. Performance varies by condition, species, prevalence, data quality, and clinical context. Yet many products are still presented using high-level claims that offer little insight into how a system behaves in day-to-day practice.
For clinicians, this creates risk. Without clear, condition-specific performance data, it becomes difficult to judge:
- When AI outputs can meaningfully support decision-making
- When uncertainty is high and additional validation is required
- How much weight AI results should carry in clinical conversations with clients
In medicine, opacity does not protect clinicians – it leaves them exposed.
Transparency Is a Professional Obligation, Not a Marketing Choice
Professional bodies such as the ACVR and ECVDI have already articulated this clearly in the context of imaging AI, highlighting the lack of transparency and validation as a key challenge in the current market.
Any AI tool that influences clinical decisions carries an ethical responsibility to demonstrate how it performs, where it works well, and where its limitations lie.
Publishing performance data:
- Supports evidence-based adoption
- Enables independent evaluation and peer-review
- Reduces over-reliance and misuse
- Reinforces the clinician's role as the accountable decision-maker
Setting an Example of What "Good" Looks Like
Vetology AI's recent decision to publicly release detailed performance metrics across its diagnostic classifiers demonstrates open disclosure of condition-level performance, including limitations. By making sensitivity, specificity, and sample sizes publicly available across applications, the company has shown that transparency at scale is achievable. Importantly, the data does not present AI as infallible, but as a decision-support tool that must be interpreted within a clinician-led workflow.
Why This Matters Beyond Individual Products
For veterinarians, transparent AI performance data supports safer, more confident clinical decision-making. For researchers, it provides the foundation for independent validation and meaningful comparison between tools. For regulators and professional bodies, it enables proportionate, evidence-based oversight rather than reactive governance driven by uncertainty. And for AI developers, transparency builds long-term trust – the currency that ultimately determines whether tools are adopted, rejected, or regulated out of relevance.
Raising the Bar for the Entire Market
As AI becomes more deeply embedded in veterinary clinical workflows, performance transparency must become a baseline expectation, regardless of whether a tool supports imaging, diagnostics, triage, or clinical decision-making.
Veterinary medicine has always demanded accountability, evidence, and professional judgement. AI should be held to the same standard.
Because when clinical decisions are involved, transparency is not optional – it is essential.