Perplexity Moves From Search To Action
Perplexity Computer vs Claude Cowork
This marks the point where search stopped being the product and became just one feature inside a bigger AI workspace. Once ChatGPT, Claude, Gemini, and Copilot all shipped web search with citations, Perplexity could no longer win just by being the cleanest answer engine. Comet was the next move, a browser that not only reads pages but also clicks, fills forms, shops, and books travel, so Perplexity could own the place where intent starts and where tasks get finished.
-
The original edge was simple, cited answers for hard web queries. By mid 2025, the major labs had folded that into their own assistants, which turned search from a distinct product category into table stakes and forced Perplexity to move up the stack.
-
Comet changes the workflow from asking a question to delegating a job. Instead of returning links, it can use browser context and take actions across websites, like filling checkout fields or making a booking, which creates more time spent, more subscription value, and more room for transaction revenue.
-
This is the same pattern seen elsewhere in AI. Labs and apps start with one killer wedge, then the interface layer becomes the real prize. The battle has moved from best model answer to who controls the desktop, browser, or app where the user actually works.
The next phase is a land grab for default agent surfaces. Perplexity is pushing from answer engine to action engine, while Anthropic, OpenAI, and others do the same from their own starting points. The winners will be the products that become a daily habit for getting real work done, not just a place to ask occasional questions.