Ask Copilot and ChatGPT the same question and the answers can look suspiciously alike. That is not a coincidence. Microsoft Copilot runs on OpenAI’s models, so underneath the interface you are often talking to the same brain. The gap opens up in everything wrapped around that brain: which apps the assistant can reach, which files it can open, what you pay, and how far you can push it before it starts handing back generic filler.
I have run both side by side for months on writing, spreadsheets and code. Here is where each one genuinely holds up.
The engine is shared, the cockpit is not
Copilot lives inside Microsoft’s estate. It appears as a Windows key shortcut, a sidebar in Edge, a panel in Word and Outlook, and a chat window at copilot.microsoft.com. Its default answers are grounded in the live web through Microsoft’s search index, so you usually get citations attached when you ask about something that happened this week.
ChatGPT is a standalone destination. It does not know your calendar unless you connect it, and it cannot see the document you have open in Word unless you paste it or upload the file. What you lose in context you gain in focus: one clean workspace with no ribbons, no task panes, and no corporate chrome competing for attention.
That one distinction drives almost every practical difference below.
What the free tiers actually give you
Both free versions are genuinely useful, and both will nudge you toward a $20 monthly plan within about ten minutes of heavy use.
Free Copilot gives you chat, web-grounded answers with sources, image generation, voice conversation, and a capped number of priority responses during busy periods. The browsing side is especially generous, which makes it a decent replacement for a search engine on factual lookups.
Free ChatGPT gives you access to a strong reasoning model with a daily message limit, file uploads for PDFs and spreadsheets, image generation, and Projects for keeping related chats together. The ceiling on hard reasoning tasks is higher, but you hit that ceiling faster.
Where Copilot earns its place
It already knows where you work
Writing an Excel formula against your actual columns, summarising a forty-message Teams thread you were copied into at 8am, drafting an Outlook reply that matches the thread’s tone. Copilot does that natively. ChatGPT can only do it after a copy-paste relay that loses half the formatting.
Answers arrive with receipts
Ask Copilot about a recent policy change or a company announcement and it typically links its sources. That matters more than it sounds. Verifying a claim takes seconds instead of a separate search, and you notice quickly when the cited page does not support what the answer claimed. If you want a deeper look at where that reliability holds and where it slips, this breakdown of the Microsoft Copilot chatbot’s strengths and weak spots is worth ten minutes.
The Windows shortcut is habit-forming
Alt+Space on a Windows machine brings up Copilot over whatever you are doing. It is a small thing that changes behaviour: you start asking instead of opening a browser. Copilot also handles screen-aware questions reasonably well on Windows 11, so “what does this error mean” works without describing the error.
Where ChatGPT still pulls ahead
Harder reasoning and longer documents
Feed both tools a 60-page contract and ask for every clause that shifts liability. ChatGPT holds the thread more reliably across that length. Copilot tends to summarise the shape of the document and lose the specific details buried on page 43.
Custom GPTs, memory, and Projects
ChatGPT lets you build a reusable assistant with your own instructions and files attached, keep persistent memory across conversations, and organise long-running work into Projects. Copilot’s equivalent features are thinner and more locked down, which matters if you want an assistant that remembers your house style.
Prose that does not read like a slide deck
Copilot defaults to corporate register and loves bullet lists. ChatGPT will write actual paragraphs if you ask, and its creative drafts need less de-bulleting. For a fuller head-to-head on which tool survives a normal working day, this comparison of ChatGPT versus Copilot in a real daily workflow digs into the specifics.
Coding: closer than most people assume
GitHub Copilot’s free tier in VS Code gives you a couple of thousand code completions and a modest monthly chat allowance. That is enough to feel the value on a side project. ChatGPT is the stronger partner for architecture discussions, strange stack traces and explaining an unfamiliar repository, partly because you can paste in more context at once.
Neither is a dedicated coding tool, though. If autocomplete is the thing you actually care about, this test of Blackbox AI against Copilot, Cursor and ChatGPT covers which one earns its tab key on real codebases.
What you pay, plainly
- Copilot free: $0. Web grounding, image generation, voice, daily limits.
- Copilot Pro: $20/month. Priority access, Copilot inside desktop Office apps, better image generation.
- Copilot for Microsoft 365: around $30 per user per month, annual commitment. This is the version that reads your actual mail, files and meetings.
- ChatGPT Plus: $20/month. Higher limits, advanced reasoning, file analysis, custom GPTs.
- ChatGPT Pro: $200/month. Only worth it if you are burning through heavy reasoning tasks daily.
Note the trap in that list. The $20 Copilot Pro plan does not include the Microsoft 365 tier, which is where the real work integration lives. If your employer already pays for Microsoft 365, check whether Copilot is switched on before you spend anything yourself.
Choosing without second-guessing yourself
The honest answer is that the decision depends less on model quality than on where your work already sits.
- You live in Outlook, Teams and Excel: Copilot, and push for the 365 tier through work.
- You write, research or code for a living: ChatGPT Plus gives you more room to think out loud.
- You want cited answers to factual questions fast: Copilot’s grounding is the better default.
- You need a reusable assistant with your own house rules baked in: a custom GPT wins comfortably.
- You are on Google Workspace all day: neither is the obvious pick, and this look at Gemini versus Copilot explains why.
The two-tab setup most people end up with
After the novelty wears off, the pattern is remarkably consistent. Copilot stays pinned for anything grounded in the web or tied to a Microsoft document: quick lookups with sources, a formula in Excel, a summary of a meeting transcript. ChatGPT gets opened for the slower work, the 2,000-word draft, the messy debugging session, the problem you need to argue with for twenty minutes before it cracks.
It is worth knowing where the other assistants sit too, since the field is wider than two names. This comparison of Copilot against ChatGPT, Gemini and Claude maps the whole set against common workflows.
One habit makes either tool noticeably better: give it the raw material. Paste the email thread, the error message, the messy notes, the three paragraphs you already wrote and hate. Both Copilot and ChatGPT produce bland output mostly when they are guessing, and specific output when they are working from something real. The model matters less than what you hand it.

