What Is Perplexity Comet and Why Is Everyone Switching Their Browser?
Perplexity Comet is a free AI-native web browser launched on iOS, Android, Windows, and Mac in March 2026. Unlike Chrome with an AI extension bolted on, Comet is built around an assistant that can read the page you are on, answer questions about its content, and execute multi-step tasks like booking, form-filling, and summarising on your behalf.
The headline numbers explain the buzz. Perplexity reported 45 million monthly users and over 1 billion queries per month in early 2026, and Comet hit number 3 on the US App Store at launch. Max-tier subscribers can now select the model powering the agent, with Claude Opus 4.6 as the default and GPT Realtime 1.5 powering voice mode.
For practitioners, the interesting part is not the marketing. It is what the browser actually changes about your daily workflow. Several tasks that used to require five or six tabs and a copy-paste dance now collapse into a single sidebar prompt. This article covers five of the most useful workflows we have actually tested.
How Does the Comet Sidebar Assistant Actually Work?
The Comet sidebar is a context-aware assistant that knows which tab you have open, what is on the page, and what you have done in the previous turn. You ask questions in natural language and it answers using the page content plus a fresh web search, with citations to both sources.
The key mental model: treat it as a research partner sitting next to your screen, not a search engine. It can read the article you opened, compare it to what you read three tabs ago, and write a draft email referencing both. Chrome with extensions cannot do this without manual copy-paste between tools.
One concrete behaviour worth knowing: if you ask a question and it does not find the answer on the current page, Comet automatically extends to the open web and tells you it did so. This is different from a normal AI chat which has no idea what page you are looking at.
Workflow 1: How Can You Use Comet for In-Page Research?
The fastest payoff comes from in-page research. Open any long article, report, or PDF in Comet, then ask the sidebar to do work that would normally take 20 minutes of skimming and note-taking. The savings compound across a research-heavy day.
Three prompts that work consistently well. First, "summarise this in five bullets, each focused on a different argument." Second, "list every claim that has a number attached and tell me which ones cite a source." Third, "compare this article's view to the last article I had open."
The third one is where Comet pulls ahead of standalone tools. Because the browser remembers your recent tabs, it can do cross-document comparison without you pasting anything. Practitioners report this is the workflow that converted them from "interesting demo" to "this is my main browser now."
The honest caveat: Comet still occasionally hallucinates a quotation when the source document is long and dense. Always click into the cited paragraph before you put a number or quote into a deliverable.
Workflow 2: How Do Multi-Step Agentic Tasks Work in Comet?
The most distinctive Comet feature is its ability to take a high-level goal and execute multiple browser steps to complete it. You tell it "find me three coworking spaces in Sheung Wan with day passes under HKD 400 and email me the results," and the agent opens tabs, runs searches, extracts the data, and drafts the email.
The clearest test of whether agentic mode is right for your task: count the steps. If a task involves four to ten predictable browser actions, agentic mode usually wins. If it requires judgment in the middle (like "decide which option fits our brand"), you are better off doing it manually with the sidebar as helper.
A specific workflow that has held up well in testing: travel research. "Compare flights from HKG to Tokyo on May 15-19 across at least three airlines, list the cheapest non-stop option, and tell me which has the best on-time reliability for that route." Comet opens the tabs, runs the comparison, and writes a clean summary.
The limit you hit fastest is rate-limiting and login walls. Sites with aggressive bot detection (some airlines, banking portals) will block the agent and you will have to take over. This is improving, but it is the single biggest reason agentic tasks fail today.
Workflow 3: How Can You Use Comet for Email Triage and Form Filling?
Comet integrates with Gmail and Outlook. Open your inbox in a tab and the sidebar can sort, summarise, and draft replies based on natural-language instructions. Practitioners report saving 30 to 45 minutes per day on email triage alone, mostly by replacing manual scanning with one prompt.
A reliable starting prompt: "Show me only emails I owe a response to today, sorted by who is waiting longest. For each one, draft a one-paragraph reply that I can edit." This reframes the inbox from a backlog into a prioritised task list.
Form-filling is the second productivity unlock. Comet can autofill repetitive forms (RFP templates, expense reports, signup pages) using context from your previous tabs. The example most people remember is filling a 30-field vendor onboarding form using information already in an attached company profile PDF.
Two cautions before you adopt this. First, never let the agent submit forms with sensitive financial fields unless you have personally reviewed every value. Second, anything involving an SSO or 2FA prompt should be done manually. The agent is good at filling fields, not at making access decisions.
Workflow 4: How Should You Compare Comet With Chrome Plus AI Extensions?
The honest answer is that Comet wins on tasks that need cross-tab context and loses on tasks that need a stable, predictable web environment. If your day involves heavy research, summarisation, or repetitive form-filling, Comet collapses workflows that Chrome cannot. If your day involves complex web apps, internal corporate tools, or anything that depends on specific Chrome extensions, the picture is mixed.
Three practical examples. Research-heavy day: Comet wins by a wide margin. Standard SaaS workday in Notion, Slack, and Salesforce: roughly tied, slight edge to Chrome because of mature extension ecosystems. Heavy regulated work (banking, brokerage, internal HR systems): use Chrome. The agent will get blocked, and the audit trail matters.
The realistic 2026 setup most practitioners settle on: Comet as the default browser for research, writing, and personal productivity; Chrome retained as a secondary browser for legacy apps and anything that absolutely must be Chromium-baseline.
One specific feature gap worth knowing: as of April 2026, Comet does not yet support all Chrome extensions natively. If your workflow depends on a specific password manager extension, dictionary lookup, or specialised research tool, check before you switch.
Workflow 5: How Can You Use Comet for Faster Long-Form Writing?
Long-form writing is one of the strongest Comet use cases that nobody talks about. Open three to five reference articles in tabs, then ask the sidebar to draft sections of your piece using each tab as a source. The model can pull specific claims, quotes, and data points across all of them in a single prompt.
A useful drafting prompt template you can copy and use today.
Try This Prompt:
You are reading three articles I have open in tabs. I am writing a 1,200-word feature for a HK business audience on [topic]. Draft a 200-word section on [specific subtopic] that synthesises the strongest claims from at least two of the three sources. Cite which tab each claim came from. Use direct, peer-to-peer voice. No corporate filler. No em-dashes.
The output is not publishable as-is, but it gives you a structured first draft with sources marked, which is faster than manual research-then-write. Most practitioners report 40 to 60 percent time savings on first drafts when the source material is already loaded into tabs.
The same technique works for anything narrative: client briefs, post-meeting summaries, internal proposals. The pattern is always the same. Open the source tabs first, then ask the sidebar to draft using them as evidence.
What Are the Honest Limitations You Should Know About?
Three limitations worth flagging before you switch your default browser. First, privacy. The agent reads page content to function. If you work with sensitive client data or regulated information, audit Perplexity's enterprise privacy controls before using Comet for that work. Second, occasional silent failures. The agent sometimes "completes" a multi-step task by reporting success even when one step failed. Always verify outputs from agentic tasks before you trust them.
Third, model variability. Because agent quality depends on which model is powering it, your experience changes when Perplexity updates the default. Practitioners on Max tier can pin to Opus 4.6 for consistency. Free users get whichever model Perplexity has currently routed.
The 2026 reality is that Comet is genuinely useful, genuinely limited, and changing fast. Test it on your specific workflow for two weeks before deciding. Most practitioners find at least two workflows that materially change with Comet, even if they keep Chrome as a backup.
懂AI，更懂你 UD相伴，AI不冷。AI tools are launching faster than any team can evaluate alone. UD has spent 28 years helping Hong Kong businesses cut through tool overwhelm and build workflows that actually deliver.
Find the Right AI Tool Stack for Your Team
Choosing the right AI tools for your specific workflow is harder than it looks, and switching costs add up fast when you pick wrong. Browse UD's curated AI directory of tools tested for Hong Kong business workflows, and we'll walk you through every step of building a stack that fits your team.
