As most AI tools started paywalling more and more of their features, I never thought I’d actually give in and pay the subscription. I usually default to free tools whenever I can - if you can cover your bases with the free features, I don’t see a reason to pay. But I made an exception for Perplexity because I wanted to see how good an AI tool could actually get once you remove the limits.

It didn’t last very long, however, not when Claude’s free tier kept outperforming in areas that actually mattered to me. I didn’t need an AI that specialized in search, but rather something that could help me synthesize and interact with those search results and turn them into something coherent. Claude’s prototyping abilities also far surpass anything similar Perplexity can do, even in its free version, which is one of the main features I was interested in.

My brief time with Perplexity Pro

And why I ditched it

I paid for a Perplexity Pro subscription for one month only, so it’s not exactly like I was bleeding money. But had I given Claude a real chance before that, I probably wouldn’t have signed up in the first place. I mainly got the subscription for better models and more importantly the generative graphics features via Labs (now renamed to “create files and apps”), hoping to be able to create some interactive prototypes.

The problem with Labs, to me, is that it behaves more like a project engine than a prototyping tool. It generates decent dashboards and web apps, but it feels more like requesting a finished deliverable than actually designing and iterating - I needed something more flexible. Not to mention, I barely touched the deep research mode since I already have access to multiple AI tools that offer this for free. So it just wasn’t really money well spent, especially not since Claude entered the picture…

I didn’t need a more powerful AI search engine

Claude ticks all the boxes I actually need

As far as search engines go, I’ve been using Google Search for work and Brave Search for studies and personal queries for years. Both are AI-powered now via Gemini and Leo, respectively. While Perplexity has its advantages of synthesized citation-rich answers, Brave and Google maintain the traditional and familiar SERP layout, just with the addition of the AI overviews and chat interfaces above that. I never really needed to pay for an AI search engine when the tools I already rely on handle discovery well enough.

If I were going to use an AI tool, it would need to help me think through ideas, question assumptions, and shape concepts into something I could visualize, not just gather information. That’s exactly the gap Claude ended up filling. It has much better reasoning abilities because it’s model-first, whereas Perplexity is search-first. Plus, it not only lets me visualize my design ideas through prototypes with its Artifacts feature, but it also lets me iterate on them, instead of handing me a fully developed output I have to work around like Labs do.

Perplexity is great at surfacing information quickly and synthesizing it according to your prompt, but most of my time isn’t spent hunting for sources. It’s spent trying to make sense of it and pushing rough concepts into something more concrete. For anyone doing any kind of work where the goal is to create, build, or solve something, a reasoning-focused assistant like Claude is just a better fit than something designed around search.

Claude has superior prototyping abilities

It lets me focus on designing

My primary goal with trying most AI tools is to have them assist and accelerate my design projects and studies. So when I discovered that the free version of Claude lets me create detailed, interactive UI design prototypes, that was enough to make me drop Perplexity on the spot - Labs doesn’t come close.

I’m not talking about Claude Code, which is a paid feature from Anthropic that turns Claude into a coding assistant capable of writing and running code tasks directly within development environments. I’m talking about the completely free Artifacts feature. Artifacts are like little containers or workspaces that keep specific tasks from cluttering up your main chats.

Here, I can actually instruct Claude to generate prototypes of my designs or design ideas, and it will use code to build them in real-time. I can use plain language, and Claude can infer layout structure, components, and interactions. However, the more information you give it to work with, the more accurate the prototype will be. Claude can create these designs in downloadable, editable HTML and SVG files. I actually opened some of the SVG files in Figma, and all the layers were intact.

The free Claude beats paid AI

Perplexity Pro isn’t bad, it just solves a different problem than the ones I actually have. My work involves finding information, but then also turning it into ideas and working concepts. Claude made the decision easy by helping me think through projects, build prototypes, and iterate without feeling boxed into a single output.