---
id: copilot-modes-and-grounding
lang: en
track: copilot
title: "Copilot modes and grounding"
seoTitle: "Copilot Quick response vs Think Deeper"
description: "Choose between Quick response and Think Deeper in Copilot, see how it grounds answers on the web by default, and know what to use when neither mode is enough."
part: getting-started
partTitle: "Getting started with Copilot"
partOrder: 2
order: 3
level: Beginner
durationMinutes: 4
outcome: "Pick between Quick response and Think Deeper, spot when Copilot is already searching the web, and know what to use when neither mode is enough."
summary: "Quick response vs. Think Deeper controls how hard Copilot reasons, and Copilot grounds its answers on the web by default, unlike Claude out of the box."
draft: false
updated: "2026-10-14"
furtherReading:
- title: "Overview of Microsoft Copilot Chat"
url: "https://learn.microsoft.com/en-us/copilot/overview"
note: "Microsoft's overview of Copilot for work and school accounts."
- title: "How grounding works with a work or school account"
url: "https://support.microsoft.com/en-us/topic/how-grounding-works-with-a-work-or-school-account-a3ccafb3-7fd3-4d51-bbd2-4245a7a1ef16"
note: "How Copilot pulls in web results and cites its sources."
- title: "Get started with Researcher in Microsoft Copilot"
url: "https://support.microsoft.com/en-us/microsoft-365-copilot/get-started-with-researcher-in-microsoft-365-copilot"
note: "How to open and use the Researcher agent."
- title: "Get started with Analyst in Microsoft Copilot"
url: "https://support.microsoft.com/en-us/microsoft-365-copilot/get-started-with-analyst-in-microsoft-365-copilot"
note: "How to open and use the Analyst agent."
keywords:
- Copilot Think Deeper
- Copilot Quick response
- Copilot web grounding
- Copilot modes
- Copilot citations
- Researcher and Analyst agents
quiz:
intro: "Five quick questions. Ace them all and the next lesson unlocks."
passScore: 5
questions:
- q: "What's the difference between Quick response and Think Deeper?"
options:
- "Quick response searches the web; Think Deeper never does"
- "Think Deeper is a separate agent you open outside the chat"
- "Quick response answers immediately; Think Deeper takes longer and reasons more thoroughly through harder, multi-step problems"
answer: 2
explain: "Quick response answers immediately; Think Deeper takes up to around ten seconds and reasons more thoroughly through harder, multi-step problems."
- q: "How does Copilot's default web behaviour differ from Claude's?"
options:
- "Copilot often grounds answers in web results with citations by default; in Claude, web search is something you turn on"
- "Claude searches the web by default; Copilot only does when you ask"
- "Neither of them can search the web"
answer: 0
explain: "Copilot often grounds answers in web results with citations by default; in Claude, web search is something you turn on."
- q: "When should you reach for Think Deeper instead of leaving it on Auto?"
options:
- "For every question, just to be safe"
- "For a genuinely hard, multi-step problem with real trade-offs"
- "For a quick lookup like the syntax of an Excel formula"
- "For a sourced research report across many sources"
answer: 1
explain: "For a genuinely hard, multi-step problem with real trade-offs, not a quick lookup. On a simple question it just makes you wait longer for the same answer."
- q: "Why check citations even on an answer Copilot already grounded for you?"
options:
- "Copilot only adds citations when it's unsure"
- "Citations only work when Think Deeper is on"
- "Copilot's citations never link to real sources"
- "A citation means Copilot found a source, not that it read it correctly or that the source is reliable"
answer: 3
explain: "A citation means Copilot found a source, not that it read it correctly or that the source is reliable. Verify anything you'll act on or repeat to someone else."
- q: "You need to dig across many sources for a real research question, not just a chat answer. What should you reach for?"
options:
- "Quick response"
- "Think Deeper"
- "Researcher, one of Copilot's built-in agents"
- "Facilitator"
answer: 2
explain: "Researcher, one of Copilot's built-in agents, rather than Quick response or Think Deeper. It digs across your files, the web and connected data and comes back with a sourced report."
pass:
badge: "Perfect score"
title: "You know which mode to pick."
body: "Full marks. You can match Quick response, Think Deeper and the built-in agents to the task, and you check citations before you trust them. Next up: Copilot Pages, for turning a chat answer into a document you can keep and co-edit."
retry: "Check the marked answers and give it another go. You've got this."
---
> **Level:** Beginner ยท **Reading time:** ~4 min
>
> **What you'll be able to do:** pick between Quick response and Think Deeper, spot when Copilot is already searching the web, and know what to use when neither mode is enough.
## Quick response vs. Think Deeper
Copilot Chat has a mode selector under the message box.
- **Quick response**: answers immediately. Good for straightforward asks where you don't need deep reasoning ("what's the syntax for a VLOOKUP," "draft a one-line Slack message").
- **Think Deeper**: takes up to around ten seconds and reasons more thoroughly through multi-step or ambiguous problems: planning, debugging, weighing a decision with real trade-offs.
- Left on its default (**Auto**), Copilot picks between the two for you based on the question. That's fine for most everyday use.
## Copilot grounds on the web by default
Here's the real contrast with Claude. Web search is something you switch on in Claude, but Copilot's answers are often already grounded in current web results with citations, without you asking. Ask about a competitor's latest pricing, for instance, and Copilot will typically come back with an answer and a source link already attached, the way a colleague who had just looked it up would, rather than waiting for you to ask it to search.
That's useful, since answers stay current, but it changes the habit you need: trust, but verify. A citation means Copilot found a source, not that it read the source correctly or that the source itself is reliable. On anything you'll act on, repeat to a colleague, or put in front of a client, open the cited source and check it yourself before you rely on it. Let Copilot do the finding; you own the fact.
Here's what that looks like in practice. Ask Copilot for a competitor's current pricing and it answers with a number and a linked source. Click through before you repeat that number anywhere: sometimes the source is the competitor's own current pricing page and the citation holds up; sometimes it's a two-year-old blog post or a reseller's page quoting an old rate, technically a real citation but a stale fact. The check takes ten seconds. Skipping it is how a wrong number ends up in a client email with your name on it.
## When neither mode is enough
Quick response and Think Deeper both answer inside the chat, in seconds. Some questions need more than that: a genuine research pass across many sources, or real analysis of a dataset rather than a quick read of it. That's what Copilot's built-in agents are for, not a third mode but a separate tool you call on deliberately. **Researcher** digs across your files, the web, and connected data to answer a broader question with a sourced report rather than a quick reply, the kind of task that would otherwise mean fifteen browser tabs and an hour of your afternoon. **Analyst** works with data directly, spreadsheets and structured files, to find patterns and answer quantitative questions instead of you building the pivot table yourself. There's also **Facilitator**, a third pre-built agent, for running and summarising a meeting rather than researching or crunching numbers. None of these are modes you toggle mid-chat the way Quick response and Think Deeper are; you open them as their own agent when the task genuinely calls for it. Reach for one of them when the honest answer to "could Quick response or Think Deeper handle this" is no, not as your default.
## Matching mode to task
Put the three together and a simple habit falls out. Stay on Quick response, or just leave it on Auto, for everyday lookups and drafting. Switch to Think Deeper the moment a question is genuinely multi-step or has real trade-offs to weigh, not for something you could answer yourself in ten seconds. Reach for Researcher or Analyst when the task is a proper research or data job, not a chat question at all. And whichever mode answers you, check the citations before you trust anything time-sensitive.
Take a concrete week: "what's the syntax for this Excel formula" stays on Quick response. "Should we push this deadline given the client's last three emails" is worth Think Deeper, since it's weighing a real trade-off rather than looking something up. "Pull together everything public about this prospect before Thursday's call" is a Researcher job, not a chat question at all. Do not fiddle for the sake of it: cranking Think Deeper on for a simple question just makes you wait longer for the same answer Quick response would have given you.
## Try it
Ask Copilot the same multi-step planning or logic question twice: once on Quick response, once on Think Deeper. Compare the two answers. Then ask something time-sensitive (recent news in your industry, a current price) and check what, if anything, it cites as its source, and click through to verify it.
## Key terms
[Quick response](/glossary#quick-response), [Think Deeper](/glossary#think-deeper), [Web grounding](/glossary#web-grounding), [Researcher](/glossary#researcher), [Analyst](/glossary#analyst), [Facilitator](/glossary#facilitator), [Copilot](/glossary#copilot)