Why AI Search Needs Separate Evaluation for Factual and Exploratory Queries

Not every search query has the same goal.

Some users want one factual answer.

Others are exploring a topic and expect multiple perspectives.

AI search should be evaluated differently for each case.

For a Malaysia-focused platform such as WINMYR, factual queries need strong source accuracy and freshness.

A question about a sports result has a correct answer.

A question about industry trends may require several sources and more interpretation.

Exploratory queries benefit from breadth.

Factual queries benefit from precision.

Using one evaluation score for both can hide weaknesses.

A system might be excellent at summarising broad topics but poor at retrieving exact dates.

Another might answer simple facts well but oversimplify complex analysis.

For WINMYR, the broader AI lesson is that search quality depends on user intent.

Evaluation should reflect the type of question the product is actually expected to answer.