Easy answer why... because you have to provide more when using the API... eg. by setting search grounding with Google APIs. The UIs are agent systems which will figure out if they need to use something like a search tool, APIs require you to set this manually.

https://ai.google.dev/gemini-api/docs/google-search

Not worth wasting your time reading this as the author missed the obvious "discrepancy"