When the app surfaces content based on what a guest has told it they enjoy — types of routes, categories of places, event preferences — it can now explain why each result was suggested.
What the guest sees
Beneath or alongside a recommended item, there’s an option to see the reasoning. When they tap it, the app generates a short explanation — one or two sentences per matched preference — describing how that item might suit what they’re looking for.
For example, if a guest has said they enjoy walking and family activities, and the app recommends a riverside route, the reasoning might explain how the route relates to each of those preferences.
Importantly: it doesn’t invent facts
The model is instructed to be conservative. If it can’t make a clear connection between the content and a preference, it says so in general terms rather than making something up. Each response also includes a short disclaimer reminding the guest that the explanation was generated by Apple Intelligence and they should check the details themselves.
Caching
Once a justification has been generated for a particular item, it’s saved so the guest doesn’t wait again if they come back to it. The model only runs once per item per session.
Info
Reasoning is generated on-device and is not editable. The quality of the reasoning depends partly on how well-written the content is in the first place — accurate, descriptive content produces better explanations than vague or sparse descriptions.