AI use and limits
Last updated July 15, 2026
How the AI features in Raisonance work, and their limits. Raisonance assists documentation; it does not make clinical decisions.
Human oversight
The audiologist must review every AI-generated output before relying on it. Raisonance produces drafts and suggestions, not decisions.
Accuracy limits
AI output can be incomplete or incorrect. It should be checked against the source information and the clinician’s own judgement.
No autonomous diagnosis
Raisonance does not diagnose, triage, or recommend treatment, and it is not clinical decision support. See the Regulatory status page.
Data flow
The model runs locally on your workstation. Prompts and patient data are not sent to a remote model provider. When PubMed evidence search is used, Raisonance sends only generalised clinical search terms to NLM/NCBI in the United States; it does not intentionally send patient names, identifiers, or clinical notes. Opening a citation contacts PubMed in your default browser.
Model improvement
Your data is not used to train any model.
Logging of AI activity
Logging of AI activity is designed to minimise patient information.
Language and bias
Quality may vary by language and context. Review output with particular care where wording is sensitive or clinically significant.
Not for emergencies
Raisonance is not intended for emergency or urgent decisions.