Copilot and Gemini can both write your meeting notes, but only if a few settings are on before the call. Here is what to check so names, tasks, and quotes come out right.

You finish a call, open the AI summary, and half of it is wrong: the wrong person is credited with a decision, the action items are vague, and the one quote you needed is paraphrased into mush. It is tempting to blame the tool and switch to the other one. But in 2026, Microsoft Copilot and Google Gemini produce good meeting notes only when a handful of settings are turned on before the meeting starts, and the choice between them is mostly about which ecosystem your call already lives in. Get the settings right and either one works; get them wrong and neither will.

At a Glance: Copilot is built for Teams and Microsoft 365, while Gemini is built for Google Meet and Workspace, so the practical rule is to use whichever matches where your meeting happens rather than treating them as interchangeable. Both need transcription turned on before the call to capture anything usable, and both attribute names correctly only when participants join signed into their real accounts instead of dialing in or appearing as a guest. Action items export well when someone states them out loud with an owner, and accurate quotes depend on a verbatim transcript being enabled. Check those before the meeting and the export lands clean; skip them and you get the garbled summary.

So which one should you use?

Use the one that matches where your meeting already happens. If your calls run in Microsoft Teams and your files live in Microsoft 365, Copilot is the natural fit because it reads the meeting transcript and your surrounding documents from inside that ecosystem, as Microsoft’s guide to Copilot in Teams meetings lays out. If your calls run in Google Meet and your work lives in Google Workspace, Gemini is the equivalent, drafting notes straight into a Google Doc. Trying to make the off-ecosystem tool cover a call it was not present for is where most disappointment comes from. So the first decision is less about which is smarter in the abstract and more about which one is actually in the room when you meet.

Turn on transcription before the call

Neither tool can summarize a conversation it never recorded. Copilot works from the Teams meeting transcript, so transcription or recording has to be switched on, and there is a mode that only lets Copilot work if a transcript is being kept. Gemini’s take notes feature has to be started for the meeting too, and it writes to a Doc as the call goes. The common failure is realizing halfway through that nothing was capturing, at which point the first half of the meeting is simply gone. Make turning on transcription part of starting the meeting, the same reflex as unmuting, so the notes have raw material to work from. If your notes keep coming back thin or empty, a missing transcript is the first thing to rule out.

Why the names come out wrong

Speaker attribution is where both tools stumble, and the cause is usually the same. To label who said what, the tool needs each speaker tied to a real signed-in account. When someone dials in by phone, joins as an anonymous guest, or shares one laptop for two people, the transcript cannot tell them apart, so quotes and decisions get pinned on the wrong person or on Speaker 2. The fix is to have everyone join signed into their own account on their own device, which is easy inside an organization and harder with outside guests. If a call mixes signed-in staff and dial-in clients, expect the client side of the attribution to be shaky. Our walkthroughs of Copilot meeting notes coming out wrong and Gemini exporting the wrong names go deeper on cleaning this up.

Getting tasks and quotes to export cleanly

Action items export well when they are spoken like action items. If someone says out loud who owns a task and by when, both tools tend to capture it as a clear to-do; if the group only implies the task, the summary often misses it or leaves it vague. A useful habit is to verbally restate decisions and owners near the end of the call, which gives the AI an unambiguous line to lift. Quotes depend on the transcript being verbatim rather than a cleaned-up paraphrase, so if you need someone’s exact words, keep the full transcript and pull the quote from there instead of trusting the summary’s version. When the tasks keep vanishing from your summaries regardless of tool, our fix for meeting notes missing action items covers the recovery steps.

Language, accuracy, and the honest limits

Both tools are strongest in clear English audio and get shakier elsewhere. Gemini’s note-taking supports a limited set of languages and its accuracy varies with accents, crosstalk, and audio quality, and Copilot has the same real-world ceiling: heavy background noise, people talking over each other, or a bad connection all degrade the transcript that everything else is built on. Google’s overview of its Workspace AI features frames these as assistants rather than court reporters, and that is the right expectation. Treat the auto-notes as a strong first draft you skim and correct rather than a final record you forward unread, and the difference between the two tools matters far less than whether you gave either one clean audio and a real transcript to work from.

What you want	Setting to check first	If it is off
Any notes at all	Transcription or recording on before the call	Summary is empty or covers only part
Correct speaker names	Everyone signed into their own account	Quotes land on the wrong person
Clear action items	Owner and deadline stated out loud	Tasks are vague or missing
Exact quotes	Full verbatim transcript kept	Wording is paraphrased and softened
Right tool for the call	Copilot for Teams, Gemini for Meet	Off-ecosystem tool misses the meeting

Both Copilot and Gemini write notes from the meeting transcript, so whether names and tasks come out right is decided by the settings you turn on before the call starts. Frequently Asked Questions Is Copilot or Gemini better for meeting notes?

Neither is universally better; the right one is whichever matches your meeting platform. Copilot is built into Microsoft Teams and 365, so it fits calls and files that already live there, while Gemini is built into Google Meet and Workspace and drafts notes into a Google Doc. If your organization runs on Microsoft, Copilot will feel smoother, and the reverse is true on Google. The quality gap between them matters far less than turning on transcription and having people sign in, which affects both equally.

Why does the AI keep putting the wrong name on quotes?

Because it can only label speakers it can identify, and it identifies them through their signed-in accounts. When someone dials in by phone, joins as a guest, or shares a device with another person, the transcript cannot separate them, so decisions and quotes get attributed to the wrong person or to a generic speaker label. The fix is to have each participant join signed into their own account on their own device. Mixed calls with outside guests are where attribution stays unreliable no matter which tool you use.

Why are my meeting notes empty or missing the first half?

Almost always because transcription or recording was not on when the meeting started, or was started late. Both Copilot and Gemini summarize from the transcript, so anything said before capture began simply is not available to them. Make turning on transcription the first thing you do when a meeting starts, and confirm it is running rather than assuming. If notes are still thin afterward, poor audio or heavy crosstalk may have produced a transcript too rough to summarize well.

How do I get clean action items instead of vague ones?

State them out loud like tasks. Both tools capture a to-do far better when someone says who owns it and by when, rather than leaving it implied in discussion. A reliable habit is to spend the last minute of the call restating each decision, its owner, and its deadline, which gives the AI an explicit line to lift into the notes. After the meeting, skim the action-item section and add anything that was agreed indirectly, since the summary tends to catch only what was said plainly.

Can I trust the quotes in an AI meeting summary?

Treat them as approximate unless you check them against the transcript. Summaries often paraphrase and tidy up wording, so a quote in the notes may not be word for word. If you need someone’s exact statement, for a decision record or a follow-up, keep the full verbatim transcript and copy the quote from there. The summary is a good draft for skimming, but the transcript is the source of truth when precise wording matters.

Key Takeaways

• Pick the tool that matches your meeting: Copilot for Microsoft Teams, Gemini for Google Meet.
• Turn on transcription or recording before the call, since both tools summarize only what was captured.
• Have everyone join signed into their own account so names attach to the right quotes and decisions.
• Say action items out loud with an owner and deadline to get clean tasks rather than vague ones.
• Keep the full transcript for exact quotes, and treat the summary as a draft you check and correct.

The Bottom Line

The Copilot-versus-Gemini debate usually asks the wrong question. The one that actually decides whether your notes are usable is which tool was in the room, whether it was allowed to transcribe, and whether the people on the call were signed in. Match the tool to your platform, turn on transcription as reflexively as you unmute, ask participants to join under their own accounts, and speak your action items plainly with owners and dates. Do that and both assistants produce a clean, accurate draft you only need to skim; skip it and the smartest model still gives you a garbled summary. The settings, rather than the brand, are what you are really choosing. For more on getting AI tools to behave, browse Venture-Lab’s Tech section.