Perplexity's Comet browser reviewed: screen-aware AI browsing is promising but slow
Jul 11, 2025
Key Points
- Perplexity's Comet browser reads screen content and auto-populates data across tabs, but page interaction is slow enough to undermine its core value proposition of collapsing multi-tab workflows.
- Comet requires explicit user prompts rather than proactive assistance, exposing a design tension in AI browsers between cluttering interfaces with low-signal suggestions or forcing constant user initiation.
- Headway's $200M ARR with 2x year-over-year growth despite free LLM alternatives suggests users value the art of good summarization over raw information, undercutting the premise that AI browsers solve information problems.
Summary
Perplexity's Comet browser is a screen-aware AI interface that reads and sometimes interacts with content visible on your screen. The core value proposition is collapsing multi-tab workflows. Users can ask Comet to extract data from a Google Sheet and have it auto-populate related information without manual copying into a separate chat window. Search queries through Perplexity's engine run quickly, but page interaction is slow enough to limit the product's usefulness.
Comet differs from Cluey, which handles video calls and audio input with proactive assistance. Comet requires explicit prompts and doesn't volunteer help unprompted. This reflects a broader design tension in AI-augmented browsers. Should the product push summaries and context proactively, like Chrome's auto-translation or Wall Street Journal's built-in article summaries? Or should it wait for user requests? Proactive features risk cluttering interfaces with low-signal AI suggestions. Pull-based interaction requires constant user initiation.
Layering AI summaries and context may not actually save time. Reading a three-sentence summary takes 10 seconds but retains far less than spending five minutes on an article. The information advantage compounds with time spent. A biographer or David Senra's carefully crafted book summaries hit differently from LLM summaries, which generally score mid. Spending an hour with AI summaries doesn't compete with five hours of original reading.
This dynamic undercuts the premise that AI browsers solve information problems, but it doesn't kill the market. Headway, a book summary app, just hit $200M ARR growing 2x year-over-year with 30% net margins. That business should theoretically be crushed by free LLMs yet thrives. People want the art of good summarization and the feeling of learning as much as the information itself.