What it actually does
Comet is Perplexity's computer-use agent, positioned as a research assistant rather than a general-purpose web automator. The framing matters: the same underlying capability, optimised for a different workflow.
What is good
- Multi-source research with synthesis is genuinely useful. The product workflow assumes you are gathering, not transacting.
- Citation quality is competitive with Perplexity's search product, which is its strongest feature.
- Faster than hand-rolling for well-bounded research tasks.
What is broken or surprising
- Not a transactional tool. Booking, form-filling, payment flows are not what Comet is positioned for; OpenAI Operator is the better fit there.
- Cost-per-task on long research sessions tracks Operator's shape; cap session length and depth.
- Brittle on sites that block automation. Like all computer-use agents, blocked surfaces are the limit.
When you would choose it
Pick Comet for research tasks where citation quality matters. Skip Comet for transactional browser tasks; use Operator. Skip Comet for tasks an API can serve; the API is always cheaper and more reliable.
Cost at scale
Subscription with usage tiers. Cost is competitive for research workloads at small to medium scale. At larger scale, per-task economics tighten and worth profiling against your specific research mix.
Read next
Oliver runs Digital Signet, a research and product studio that operates ~500 production sites with AI agents as the engineering layer. The Digital Signet portfolio is built using a continuous AI-agent build pipeline, one of the largest agent-operated publishing operations on the open web. The handbook draws directly from those deployments: real cost data, real failure modes, real recovery patterns.