As someone who covers AI tools every single day, I’m at a funny point where I can rule out AI subscriptions as a business expense. However, if you’ve been keeping up with subscriptions and usage limits, you’re likely well aware that the $20 tier that once got you practically unlimited usage no longer cuts it. In fact, the $20 tier is more akin to what once used to be the free tier. With just a few messages in and some heavy usage, you’ll hit the 5-hour limit on these plans, and eventually the weekly limit too.
Given just how much I write about these tools, being locked out of one mid-sentence isn’t something I can really afford. This means I’m typically on the higher-tier plans, and I’ve become somewhat of an expert in the very specific art of figuring out which subscriptions are actually worth their increasingly steep price and which ones I was paying mostly for the sake of being subscribed to it. I recently did an audit of all my AI subscriptions, and I ended up canceling three of them and deciding the remaining two were the only ones earning their place on my card.
Want to stay in the loop with the latest in AI? The XDA AI Insider newsletter drops weekly with deep dives, tool recommendations, and hands-on coverage you won’t find anywhere else on the site. Subscribe by modifying your newsletter preferences!
My AI stack is bigger than I’d like to admit
But we’re only talking the big players today
This might be getting a tad bit repetitive, but given the nature of my job, I have access to all sorts of AI tools and actively use them. I have a subscription to Reclaim AI, which is an AI-assisted calendar scheduling app. I use a tool called Read AI for meeting notes, and Recall is a great subscription that lets me keep track of everything I’ve read, watched, or saved. Since I’m majoring in computer science and have multiple math-heavy courses every semester, MathGPT is another subscription I can’t really do without.
The point I’m trying to make here is that I use a lot of AI tools, and most of them aren’t really up for debate. They each do one specific thing, and nothing else really does that thing, so they stay. The chatbots are the messy part. They all kind of do the same stuff, which means I was paying for five tools to solve one problem. So those are the ones I actually had to sit down and think about. Up until a few days ago, I was subscribed to Gemini, Claude, ChatGPT, Perplexity, and Copilot. I was on different tiers for each, depending on how restricting their limits felt and how much I actually leaned on them day to day.
Claude wasn’t even a question
ChatGPT put up a fight, but I’d pick Claude anyway
This shouldn’t surprise anyone, but Claude was the easiest to keep on the list. A lot of that comes down to the fact that I’ve built basically my entire workflow around it. Beyond just the chatbot, I’ve started relying on Projects for keeping my research organized, Cowork for the heavier multi-step stuff, Claude Design when I need to mock something up, and Claude Code for, well, coding. These aren’t features I tried once and forgot about, and now have a permanent spot in my workflow. Pulling Claude would mean rebuilding the whole thing from scratch. Now, ChatGPT does put up a real fight here. Its newer models are incredible, Codex has been shaping up fast, and its usage limits are often better.
However, Claude just offers me more value, even with the worst usage limits. I like its outputs more, I like how the model actually talks to me, and I like the direction Anthropic keeps heading in. The features they ship tend to feel like they’re trying something rather than copying whatever everyone else already did. ChatGPT is a solid second, and it’s definitely the subscription I’d have kept if Claude didn’t dominate in the areas above. I also tend to agree with Anthropic’s stance on safety more than I do with OpenAI’s. The OpenAI-Department of War deal earlier this year left a bit of a bad taste, and I’ve since found it harder to trust OpenAI. While that doesn’t really decide which tool is better or worse performance-wise, it’s the kind of thing that sits in the back of my mind every time I open one over the other. And when two tools are already this close, that’s enough to settle it.
Gemini was the easy keep I didn’t have to think about
NotebookLM Plus does most of the convincing
The other AI subscription I decided to keep was Gemini, and the biggest reason it stays is NotebookLM Plus, which comes bundled with Google AI Pro. The tier bumps up limits considerably, allowing you to have up to 600 sources per notebook, no watermarks on generations, more notebooks, priority access to the better models, and higher limits on the outputs I lean on most, like Audio Overviews and Slide Decks.
The rest of the Google ecosystem is also what makes it stick. With the subscription, you get access to the best models, Gemini wired directly into Gmail, Docs, Sheets, and the rest of Workspace. Copilot is the subscription that directly competes with Gemini here, but given I use the Google Workspace apps for basically everything, Copilot is the odd one out. It’s a strong tool if you live in Microsoft 365, but I don’t, so paying for AI baked into apps I barely open never made sense.
Perplexity, Copilot, and ChatGPT just don’t make sense for me anymore
Perplexity, while once among my favorite AI tools, doesn’t offer all that much that I can’t already get elsewhere now. Its whole appeal was sourced, cited answers with live search baked in. For a while, that genuinely was something the others didn’t do well. But that gap has basically closed now. Gemini’s Deep Research does the multi-step, cited research I used to open Perplexity for, and between that and the search built into the tools I already pay for, I kept noticing I just wasn’t reaching for it anymore. Paying for a tool I’d half-forgotten I had is exactly the kind of thing this audit was meant to catch!
And that’s really the thread running through all three cancellations. Copilot was AI baked into apps I don’t use, Perplexity was a research tool whose one trick the others learned, and ChatGPT was the second-best version of something I already get from Claude.
