Rarely does anything good come for free. While I’d be willing to bet almost everyone reading this is subscribed to at least one service like Netflix, Spotify, Apple Music, YouTube Premium, or some combination of them all — AI is a different story. As someone who writes about AI daily (and has realized that AI is an incredible tool when used the right way), I’ve spent the better part of the last two years putting each AI tool through its paces every single day.

This includes ChatGPT, Perplexity, Claude, and Gemini, and the thing about AI tools is that their free versions are shockingly capable at zero cost. While I’ve subscribed to each tool at least once (typically simultaneously), I’ve come to a clear conclusion: only one of them is worth keeping on my credit card, while the rest have free tiers that are more than enough for most people.

Claude is the only AI subscription worth paying for

Worth. every. cent.

As with any other subscription you’re considering opting for, the AI tool that’s ideal for you depends on your specific needs and how you actually use it day to day. Having rotated between all four, I now think Claude is the only AI tool where opting for the paid plan makes a substantial difference.

Here’s the thing: Claude’s free tier is more or less unusable. While its responses are still great on the free tier, you’ll hit the usage limit within a handful of messages. We’re talking maybe nine or ten exchanges before you’re locked out for the next five hours. If you’re using Claude for anything beyond a quick one-off question, that limit really, really gets in the way.

Claude currently has three paid tiers: Pro, Max 5x, and Max 20x. The Pro plan costs $20/month, while the Max 5x costs $100/month, and the 20x plan costs $200/month. All the paid plans unlock Cowork, Claude Code, Projects, Research, Memory, Connectors, and Claude in Excel, PowerPoint, and Chrome. These features are not currently available on the free tier and are only unlocked when you subscribe to one of the paid tiers.

Claude Code is simply one of the most compelling reasons to pay, whether you’re a developer or not. It brings an AI programmer right to your terminal and lets it make changes to your codebase directly. Cowork is another feature I’ve been loving, and I detailed my entire experience with it in a separate article.

But even features aside, usable limits are what you’re really paying for. The Pro plan gives you roughly five times the usage of the free tier, which translates to around 45 messages every five hours. The Max 5x and 20x plans multiply that further, but unless you’re spending eight-plus hours a day inside Claude, Pro is more than enough. When you combine Claude’s incredibly powerful models (which I think are beating every other AI tool right now in writing quality and reasoning) with the features you unlock on the paid tier, no other AI subscription comes close in terms of value.

Finally, the Opus models, including the most recently released Opus 4.6, which is Anthropic’s most powerful model, are only available on the paid tiers. Free users are limited to Sonnet and Haiku, which are still capable, but the jump to Opus makes a bigger difference than you’d think.

As much as I wish Claude’s free tier would suffice, its limits don’t make that possible. The free tier gives you just enough to see how good it is, and the paid tier gives you the full experience. So, if there’s one AI plan you’re paying for in 2025, make it Claude.

Gemini is the only other AI plan I'd consider paying for

But only if you're already in the Google's ecosystem

If you've read even one article of mine before or chatted with me, something that you'll know is that I constantly praise Gemini and Google's AI efforts. Even though I prefer Claude's subscription, that doesn't in any way mean that Gemini isn't worth your attention. Google currently has three paid tiers: AI Plus, AI Pro, and AI Ultra. Plus retails for $7.99/month, the AI Pro costs $19.99/month, and the AI Ultra costs a whopping $249.99/month. Google's AI Pro plan at $19.99/month is still hands-down the best value bundle in this space. As with Claude, Gemini's paid plan gets you access to more features and higher limits.

While I do prefer Anthropic’s models over Google’s and think they handle coding and complex reasoning better, Google is doing a lot of things right in AI right now, and it would be dishonest not to acknowledge that. NotebookLM is hands-down my favorite AI tool, and if you’ve read my work before, you already know that. It’s the best AI tool out there for students, and nothing else comes close.

While I still think NotebookLM’s free plan suffices for many people, the biggest benefit of Google AI Pro is that you get 5x the limits. You can generate significantly more audio overviews and slide decks, and ask far more questions.

With Google’s paid AI plans, you also get Gemini integrated directly into the Workspace apps you rely on every day, including Gmail, Docs, Sheets, Slides, Vids, and more. I’ve tested all of these capabilities extensively, and I think they make a massive difference when you’re already using these tools as part of your workflow.

That said, the reason I didn’t name a Google AI plan as the one AI subscription you should be paying for is simple: the value lies in the ecosystem, not the AI model itself. If you stripped away the storage, the Workspace integration, and NotebookLM, and judged Gemini purely on the quality of its responses, it doesn’t beat Claude.

If you’re deep in Google’s ecosystem, a NotebookLM fanatic like me, and don’t need Claude-level reasoning for your day-to-day work, Google AI Pro is the only other AI subscription I’d genuinely recommend.

Perplexity was my topic pick for the longest time

It isn't what it used to be

A couple of months ago, Perplexity would have been at the top of this list. Even though I got the Pro subscription for free as part of several promotions Perplexity ran, I canceled my subscription a few months later when it was caught quietly downgrading paid users’ queries to cheaper, less capable models without telling them.

I’ve detailed what happened in a separate article, but long story short, Pro subscribers started noticing a significant drop in answer quality and eventually found out that Perplexity was silently routing queries to smaller models. For example, it was swapping out Claude Sonnet for Haiku at times, while the interface still showed the premium model you selected. Ever since that happened, I’ve struggled to trust Perplexity’s outputs.

More recently, Perplexity massively slashed daily query limits. Mind you, it isn’t a marginal difference. For instance, Pro subscribers have reported having their Deep Research limits cut from 600 per day to just 20 per month. Read that again: 600 per day to 20 per month. Rather than easing subscribers into this, Perplexity made the change with no warning whatsoever. When users reached out to support, they were simply told that “rate limits have been updated” and were prompted to upgrade to the $200/month Max plan.

For a tool that built its entire reputation on transparency and trust, this is about as bad as it gets. I genuinely loved Perplexity. Its real-time search with citations, the ability to switch between multiple AI models, and Deep Research were features no other tool was offering in one place.

ChatGPT is a jack of all trades

The free tier handles everything well

ChatGPT is the tool that began the AI boom, and it’s easily the most recognizable name in the space (and for good reason). ChatGPT does just about everything well, including coding, research, image generation, voice conversation, and more.

OpenAI currently offers three paid tiers: Go, Plus, and Pro. The Go plan is ChatGPT’s most budget-friendly plan and retails for $8/month, the Plus plan costs $20/month, and the Pro plan costs $200/month. I subscribed to the Plus plan for quite a few months, but quickly realized that the subscription doesn’t really offer anything groundbreaking that the free tier doesn’t already give you.

ChatGPT’s free tier is, in my opinion, the most generous free AI experience available right now. You get access to GPT-5.2, image generation, web browsing, and voice — all without spending a cent. The Plus plan gets you higher message limits, GPT-5.2 Thinking for more complex reasoning, and basic Sora access for video generation, but the actual quality of responses between free and paid never felt dramatically different to me. It’s not like Claude, where upgrading unlocks an entirely different caliber of model.

With ChatGPT, you’re essentially paying $20 a month to use the same tool more often and with fewer interruptions. And for most people (including me), the free tier’s limits are generous enough that you’ll rarely feel the need to upgrade. If you’re someone who’s never tried an AI tool before, or you just want something that handles a little bit of everything without paying a dime, ChatGPT is the one I’d point you to. I use the free tier daily and don’t feel the need to re-subscribe again.

It all comes down to your needs and priorities

After reading the above, you’ve probably realized that the pricing of each AI tool isn’t very different. What really sets them apart is how those prices translate into value. At the end of the day, it all comes down to your specific needs, priorities, and how you actually use AI day to day.