Looker: structured exploration for people who will learn it
Against ThoughtSpot, Looker's interface is more structured. A user opens an Explore, chooses dimensions and measures the LookML model exposes, filters, and saves the result to a dashboard. That takes some training in dimension-and-measure thinking, but it gives precise, repeatable results, and the same model powers the API and embedded views through the embed SDK.
Looker has been adding natural language too: Gemini in Looker brings conversational analytics and explore assistance on top of the model, which narrows the gap with search-first tools. The heavy part is unchanged, though. LookML is written and maintained by analytics engineers, and the model has to be good before anything on top of it works well.