I get asked this question more than almost anything else these days, usually by someone standing in a hallway holding a coffee, looking a little exasperated. “Which one am I actually supposed to be using?” It’s a fair question, and it doesn’t have a fair answer, because the honest response is (and that’s the consultant speak coming out in me here): “it depends.” It depends on what you’re doing and, frankly, what your company already pays for.
That’s not a cop-out. It’s the actual shape of the problem. These three tools aren’t interchangeable flavors of the same thing, even though the marketing sometimes makes them sound that way. They come from different companies with different priorities, they’re priced in different (and in one case, genuinely confusing) ways, and they show up in your day differently depending on whether you live in Word and Outlook or a browser tab.
So let’s sort it out properly: not “which one is best,” but which one is best for what.
The short version, if you’re in a hurry
Claude tends to be the strongest pick for writing, analysis, and anything where you need the AI to actually follow a detailed set of instructions and reason through a problem rather than just generate plausible-sounding text. ChatGPT is the most broadly capable generalist with the biggest ecosystem of plugins, custom GPTs, and third-party integrations. It’s the one most people have already tried. Copilot wins the moment your work already lives inside Word, Excel, Outlook, and Teams, because it’s stitched directly into those apps in a way the other two aren’t, even when they’re technically “just as smart.”
None of that means one is universally better. It means they’re solving slightly different problems, and the right tool depends on where you’re standing.
What each one is actually good at
Claude, from Anthropic, has built a reputation, deserved in my experience, around long-document work, careful instruction-following, and coding. If you paste in a 40-page contract and ask it to flag anything unusual, or hand it a rough architecture diagram and ask it to poke holes in your design (see my earlier piece on AI prompts for exactly that kind of use), it tends to stay on task rather than drift. The current lineup runs from a free tier up through Claude Sonnet 5 and Claude Opus 5 on paid plans, with Claude Haiku 4.5 as the fast, lightweight option for quick tasks. Anthropic also has a research/agentic tier above that called Mythos, but that’s not something most of us will touch day to day. (Or one that I can personally have access to, but here’s to hoping one day.) 😉
ChatGPT, from OpenAI, is the one with the biggest surface area. It was first, it has the most name recognition, and it has the deepest ecosystem: custom GPTs, a genuinely useful voice mode, image generation built in, and a coding agent called Codex for people who live in a terminal. Pricing runs from a free tier (which now shows ads) up through a $20-a-month Plus plan and a $200-a-month Pro tier for people who want the least-restricted version of everything. If you want one tool that does a little bit of everything reasonably well, and you don’t mind bouncing between models depending on the task, ChatGPT is the safest generalist bet.
Copilot, from Microsoft, is a different animal entirely, because “Copilot” isn’t really one product: it’s a brand slapped across several different things. There’s the free Copilot you get in Windows and Edge. There’s Copilot Pro, a $20-a-month individual subscription. There’s Copilot bundled into Microsoft 365 Personal and Family plans with a monthly allotment of AI credits that quietly ticks down as you use it. And there’s Microsoft 365 Copilot for business, which runs your company about $30 per user per month on top of whatever Microsoft 365 license you’re already paying for. If that paragraph felt like a lot, that’s the point. Copilot’s real advantage isn’t raw model quality, it’s that it’s already sitting inside the apps you use eight hours a day. Ask it to summarize a Teams call you missed or draft a first pass at a PowerPoint from a Word doc, and it does that with zero context-switching. That’s genuinely valuable. It’s also genuinely confusing to shop for.
Where this actually matters: your data
Here’s the part that doesn’t get talked about enough outside of IT circles, and it’s the part I actually care about as an architect: what happens to what you type in.
If you’re using any of these tools through a free or personal-tier account, assume (unless you’ve explicitly gone into settings and turned it off) that your conversations may be used to improve the underlying models. Enterprise and business tiers of all three vendors offer contractual protections and admin controls that the consumer versions don’t. That’s not a knock on any one vendor; it’s just how the economics of “free” work across the entire industry. (As a general rule, if you are getting a product for free, you are not the consumer, you are the product.)
If you’d rather not wonder, here’s where the toggle lives for each: Copilot, ChatGPT, and Claude. Takes about two minutes total, and it’s the kind of thing you do once and forget about.
This is exactly the ground rule I open every AI-prompts conversation with: never paste credentials, customer data, internal hostnames, or anything with compliance implications into a public AI tool, personal account or not. If your company has stood up an enterprise instance of any of these three, with the right data handling agreement in place, use that one for anything that touches real work. If it hasn’t, sanitize first and treat the free tier like what it is: a demo, not a filing cabinet. Protecting what you type into these tools is one side of data protection; the other side is what happens if you lose that data entirely, which I get into in Backups vs. DR: Why You Need Both.
Honorable mentions: the ones outside the big three
The “big three” framing is a little unfair, because it leaves out a handful of tools that are genuinely worth knowing about, even if they didn’t make my daily rotation. (And since I always try to add in bonus material as a thank you for your time reading this, here are a few more for your consideration.)
Google Gemini is the one with the most reach you might not realize you already have: it’s baked into Gmail, Docs, and Android, the same way Copilot is baked into Microsoft’s stack. Pricing runs from free up through a lower-cost “AI Plus” tier and a roughly $20-a-month “AI Pro” tier. If you’re already deep in Google Workspace instead of Microsoft 365, Gemini is your version of the “it’s already sitting inside my apps” argument I made for Copilot above.
Perplexity is the one I’ll cop to using myself: I got in early as an adopter, and I still run into it on the free tier’s usage limits more often than I’d like to admit. It’s less a chatbot and more a research engine: you ask it something, and it goes out, reads multiple sources, and hands you an answer with citations attached, rather than a wall of text you have to fact-check yourself. If your use case is “find me the current answer to this and show your work,” it’s worth a look ahead of a general-purpose assistant. I have also used it extensively for “Tell me how to do {insert task here} using Copilot, step by step, with the appropriate prompts.” That has been very beneficial for learning these other platforms and finding new capabilities along the way.
Grok, from xAI, is Elon Musk’s entry, tightly wired into X (formerly Twitter). It’s genuinely capable on benchmarks, and if you’re already paying for X Premium+ it comes along for the ride; otherwise SuperGrok runs around $30 a month standalone. Worth knowing it exists, particularly if real-time X data matters to what you’re doing, that’s its actual differentiator, not just “another chatbot.”
DeepSeek gets the asterisk. It’s a Chinese-developed model that made real waves for being genuinely capable and remarkably cheap to run. Here’s my architect/security hat going on: DeepSeek’s own privacy policy states that user data is stored on servers in China, and multiple governments, including Australia at a federal device level, have restricted or banned its use over exactly that concern. I have clients in other countries that are very sensitive to data and compute sovereignty for very legit reasons, and having everything living in China should raise concerns even if you’re not the tin-foil-hat type. I’m not going to tell you never to try it out of curiosity. I am going to tell you the same thing I’d tell you about any tool handling sensitive data: know where your data physically lives and who can compel access to it before you make it part of your workflow, not after. (Hey CLOUD Act, I am looking at you here too.)
So which one do you actually pick?
If you’re doing serious writing, research, or need something to reason carefully through a long, messy problem, lean Claude.
If you want one flexible tool that also does images, voice, and has the widest plugin ecosystem, lean ChatGPT.
If your day is already Outlook, Word, Excel, and Teams, and you want AI help without leaving those apps, Copilot is going to feel the most natural, and you may already be paying for a chunk of it without realizing it.
And here’s the thing nobody wants to hear: a lot of us end up using at least two of these, not one. I keep a Claude tab open for anything that needs real drafting or analysis, and I lean on Copilot inside Outlook because that’s where my inbox already lives, and most of my meetings are in Teams and with transcription on, it takes notes and flags action items and follow-up tasks way better than my dyslexic typing skills ever could. There’s no badge of honor for picking exactly one AI tool and sticking with it out of loyalty. Pick based on the task in front of you, not the logo.
The Monday takeaway
Don’t ask “which AI is best.” Ask “what am I about to do, and which of these tools is already sitting closest to that work?” Nine times out of ten, that question answers itself faster than any comparison chart, including this one.