Start here. This is the direct spoken answer to practice first.
Why this question matters
Modern observability is moving toward OpenTelemetry, but interviews care about judgment more than package names. This drill matters because the production behavior is shaped by platform configuration, code boundaries, and the operational path the team will use during a real incident.
Both approaches can send useful telemetry from a .NET API to Application Insights. The classic Application Insights SDK is often straightforward for Azure-focused apps. OpenTelemetry gives a more standard instrumentation model and better portability across backends. I would choose the option that preserves request, dependency, exception, log, metric, and trace correlation with the least operational surprise for the team.