Which AI Should You Pay For? Here's My Honest Answer.

I don’t think it will surprise you, but it’s also not what you might expect either.

Update September 5, 2026: This is my most popular post on 30 Plus Days of AI, but since I wrote it in March, 2026 a lot has changed. My primary tool is Claude now, and Gemini as a backup. I’ve lost count of the number of new model releases. Google has gone from a leader to distant third. I’ve updated this post with new information and perspectives, but the way things are going I’ll need to update it again in two months.

In nearly every workshop I hold, the same question comes up:

Which AI should I use? It’s there a best or right one? Which one is the right one to pay for?

First off, ChatGPT, Gemini, and Claude are all amazing. You’re not going to go wrong with any of them. It’s like the old saying “No one got fired for buying IBM,” if you pick ChatGPT or Gemini or Claude; it’s fine. The lead for which of the models is “best” changes every time one is updated. They are all powerful, capable tools with the potential to do amazing things.

The question is then: If all of them are great, then how do I pick?

The answer is as simple as this:

You’re picking the one that fits inside the work you’re already doing, the tools you’re already using, and the budget you can justify.

Look at what you’re already using and start from there. The goal of AI is to improve and enhance your work, not make things harder because you’re shoehorning a new AI tool in because “it’s cool.”

No magic answer

The question isn’t which AI is best; it’s which AI fits best with the tools you use, the people you work with, and what you want to get done. The only way to learn that is to try all the tools and see which one works best for you.

And don’t be worried that you’re not creating a mind blowing app with Claude Code, automating your work with Claude Cowork, using OpenClaw to run your start up, or making a Gem that reads your thoughts and writes posts for you. In truth, only a teeny tiny fraction of people do any of that. Most people have a few prompts they like and copy-paste them in as they need them. A few people have GPTs they use a lot.

Keep in mind, my wife has been paying for ChatGPT for almost three years, uses it to help her be a better teacher. Does she have saved prompts? Nope. Does she let me help her create custom GPTs to do repeated tasks more easily. Nope. Doest she get a tremendous amount of value using it?

Hell yes.

Let’s walk through a few scenarios to give you my take on each of them.

If your team runs on Google Workspace—hang tight it will get better

The first AI tool I started paying for was Gemini as part of Google Workspace. It was a better deal and gave me more value than ChatGPT did; I hadn’t even started playing with Claude. In hindsight I wish I had made paid more attention to Claude, but that was nearly a year ago now, so ’dems the breaks I guess.

At the time Google’s models were class leading. It’s connections into Google Workspace were best-in-class and the difference between paying for Gemini via Workspace and personal Gmail wasn’t worth worrying about. Now it’s a different story.

Google has fallen way, way behind. The new 3.8 Flash models are solidly good at coding, but ChatGPT and Claude’s middle of the road models (Terra and Sonnet respectfully) perform at Gemini Pro 3.1 level. Speaking of Gemini Pro, the launch of 3.5 hasn’t happened yet and the rumor is that Gemini 4 will be the next Pro model out the gate.

When is anyone’s guess.

Over the summer Google made Workspace customers second-class citizens in for Gemini. Glacial feature releases, no access to Antigravity, and a price tag that just didn’t make sense. In contrast a Google One subscription at Pro AI level is only $26CAD/mo, includes 5 TB of storage, lots of usage, full Antigravity, and regular new feature releases. Oh, and I can share it all with my family. Oh and I get credits on Google Cloud Platform—credits that let me build StickyMapper, my bespoke training site, and more.

No brainer to switch.

So…Workspace customers? I have to believe Google will get its head out of its butt and bring Workspace—Enterprise users especially—into the modern AI era. Google still has the model, the compute, the chips, and the ecosystem. Something only Microsoft can even really get close to—and they don’t have the chips. I’m not counting Google out, but they have serious work to do to get back on top.

If your team runs on Microsoft 365

Here’s my honest take: I’ve used Copilot a good bit and I’ve been underwhelmed (this hasn’t changed much since my the first edition of this post). Even with its own new models coming out, I have yet to hear anyone in the AI community wax poetic about Copilot. Yes, it’s been getting better with new features and tools, still talking about Copilot seems to always be an afterthought. The “oh and yes Copilot…” statement when talking about AI tools.

Except.

If your company is a Microsoft shop, Copilot is going to be the tool you’ll get. It might not be awesome. It might be doing just okay work for you. However it’s integrated into all the tools your team uses—just like if you’re in the Google Workspace world. Bolting Claude Cowork into Copilot—and building on it—will be great for a lot of people. Admittedly, I sometimes dabble a bit with Copilot at work because the basics are included with our M365 subscription. Still not the first place I go when I need to do work.

When I had access to insider pricing, I thought the per user costs of Copilot were pretty intense compared to Google, ChatGPT, or Claude. Since Microsoft has a massive install base and the cost of switching from M365 to Google Workspace (really the only other option) incredibly high, they don’t have price pressure other companies do. Sure, there is a risk of people opting for Claude or ChatGPT (with their integrations into M365 tools), but for a big company it’s a lot easier to just “turn on” Copilot than do a Claude or ChatGPT roll out.

If you’re starting from scratch

If you’re a small business or solo-preneur and have all the options available to you—try them all. The free tiers of all the tools are good to start with and kick the tires on. I find Gemini has the most generous free tiers. Both ChatGPT and Claude’s free tiers are getting tighter and tighter. Still enough to get a feel for them, but not for a lot of serious work. At least without taking a break for a few ours. One plus for ChatGPT and Google is you can try the paid tiers on a 30 day trial, but Claude doesn’t offer that perk.

See which of the tools work the best for you. In your professional circle (of people who use AI), which ones do they use? Having folks who will help you get going with the tools really helps you get the most out of AI quickly. Personally I think the Trust Insights Slack community is awesome and highly recommend joining it. The group ranges from the super power user Chris Penn to us mere mortals who have to look up commands on a regular basis.

My personal recommendation leans more toward Claude and ChatGPT. It used to be Claude and Gemini, but Gemini has fallen so far behind I can’t put it on the “this should be your primary tool” list. The bottom line is, you have look at what you do day-to-day and which tool helps you get your work done—not which one is “cool” right now.

Claude has set the bar pretty high for “the tool that lets you get sh*t done.” Cowork, Code, Design, and Remote, plus Skills and MCPs set the standard for how AI tools should work. I spend hours a day using all of them. Often starting Code and Cowork sessions on my phone when I need to do to something on my personal machine while I’m at work. If I could turn back time (yes, cue Cher), I would put my money down on Claude.

I still don’t understand why, as awesome as Claude is, it still can’t create images like Gemini or ChatGPT can.

Yeah, doesn’t make sense to me either.

Okay, but what if I can have two?

This is where I’ll tell you what I’d pay for: Claude and Gemini. Hands down, no question. I’m using Claude Pro now (finally) and Gemini as my backup (like when I max out my Claude usage or need to make an image). I might downgrade my Gemini Pro AI subscription, but I the Google Developer Program benefits and access to new features ahead of free users is pretty niche.

I do my heavy lifting in Claude—coding, writing, data analysis, design. Claude is where I go to get work done. Claude has great access to a lot of tools—even Google ones—I use all the time (at home and work). I think Gemini has better access to Google tools, but Claude has been able to clean house on Google Drive and help me consolidate files and emails from my Workspace account to person (that’s a whole other discussion).

Since I doubled down on Claude about a month ago, I haven’t made a single new Gem, but tons of SKILLs. I have played with Gemini Spark (Google’s answer to Cowork, at least I think it is) to see what’s what, but except for making images, it’s all Claude.

Comparing AI tool pricing—Commoditizing knowledge

Only Google and OpenAI have paid tiers in the <$20 range with plans that are around $10/month (USD). These might be okay starter plans, but I expect that the feature limitations aren’t going to be worth the bargain.

Why should I pay for a tool if free is good enough?

I don’t think I can answer that for everyone, but for me it comes down to—getting work done and access to new features. I know I said you shouldn’t chase new features, but if you’re in a tool’s ecosystem and you hear about the new tool to use, yeah you want to try it.

Hitting a usage limit is a whole different thing altogether.

When you’re in the middle of a project—keeping in mind that you can’t use Claude Cowork, Code, or Design unless you pay for at least a Pro account—and hit that usage wall you lose a lot of momentum. Not just the time factor, but momentum. Sometimes it’s not so bad. You can keep editing something on your own or work on your deck “the old fashioned way,” but hitting that wall for me always makes me think—what the Hell do I do now‽

It’s like when the internet goes out at home. Suddenly all the things you usually do just don’t work. A year ago, I didn’t think I’d be so enmeshed in my AI tools that I’d have to pause and wonder what I can do, but…this is how fast things have changed. From “okay compile all this research” or “look at these reports and see where they need strengthening” or “just do this menial task for me so I can focus on something else…” AI and my work are pretty tightly woven together.