Perplexity’s Comet Assistant Gets Major Upgrade With More Transparency, Control and Safety
Perplexity AI has rolled out one of its most significant enhancements yet to the Comet Assistant, with CEO Aravind Srinivas announcing a new focus on transparency, user choice and safer automation for web-based tasks. Revealed on Saturday, the update aims to give people a clearer understanding of how the browser-based AI agent works and ensure it operates only with user-approved permissions.
According to Srinivas, the latest version of Comet has been engineered to behave more like a responsible human assistant—visible, accountable and careful about when to act independently. The company hopes these improvements will help users feel more confident about letting the assistant handle complex online tasks.
Comet Now Shows Its Actions Step-by-Step
In a detailed blog post, Perplexity explained that users will now have a front-row view of everything the assistant is doing. Whether Comet is browsing a webpage, clicking through links, scrolling, or following a multi-step process, its actions will be displayed in real time.
A new sidecar panel brings an added layer of clarity by revealing the assistant’s reasoning and allowing users to follow its workflow as it unfolds. This design, Perplexity says, ensures people always know why Comet takes certain actions and how it arrives at an answer.
Users Stay in Charge at Every Step
Another major improvement addresses how and when Comet takes control of tasks. Now, when someone enters a query in the omnibar or search interface, the assistant may ask if it should carry out the task autonomously. Users can choose from several options—handling the task themselves, allowing Comet to perform the action once, or granting it ongoing permission.
Perplexity emphasizes that this system ensures the assistant never oversteps, giving people the freedom to let it automate tasks only when they feel comfortable.
Stronger Guardrails for Sensitive Actions
Recognizing that certain actions require extra caution, the upgraded Comet Assistant pauses when it faces high-risk steps such as logging into accounts or proceeding with purchases. In these cases, the assistant will explicitly ask for permission before moving forward.
Perplexity says this mirrors the judgment of a trusted human aide—someone who understands when a decision is too important to take without checking in first.
Part of a Broader Vision for Responsible AI
The changes to Comet fit into Perplexity’s larger mission of creating reliable, transparent and user-centric AI tools across its ecosystem, which includes Deep Research, Labs, Email Assistant and Background Assistants.
As the company puts it, the goal is to shift the user mindset from wondering what an AI can do to deciding what they want the assistant to accomplish—confident in the knowledge that it operates safely and visibly at all times.
If you'd like, I can also create a shorter version, add subheadings, prepare social-media snippets, or draft a push-notification summary.