ChatGPT, Claude, and Friends Walk Into a Bar: We Pit the Top AI Writing Tools Against Each Other

Everybody and their grandma has an opinion about AI writing tools right now. Half the internet thinks they're going to replace every writer on the planet. The other half thinks they're glorified autocomplete with a marketing budget. After spending a few weeks throwing real-world tasks at ChatGPT, Claude, Perplexity, and a couple of others, we're landing somewhere in the messy middle — and we've got receipts.

We didn't run these tools through sterile benchmark tests. We used them the way a normal person actually would: drafting a passive-aggressive email to a landlord, trying to write the opening chapter of a thriller, debugging some gnarly JavaScript, and brainstorming names for a hypothetical hot sauce brand (yes, really). Here's what we found.

The Contenders

For this comparison, we focused on the tools most people are actually using or considering:

• ChatGPT (GPT-4o) — OpenAI's flagship, available free with limits or via ChatGPT Plus at $20/month
• Claude 3.5 Sonnet — Anthropic's model, with a generous free tier and a Pro plan at $20/month
• Perplexity AI — more of a research-first tool, but it has writing chops worth testing
• Gemini Advanced — Google's entry, bundled with Google One at $20/month

We tested each on the free tier where possible, and paid tiers where the feature gap was significant.

Task 1: Write a Difficult Email

We asked each tool to draft a firm but professional email disputing a surprise charge on a credit card bill — the kind of thing most people spend 20 minutes agonizing over.

ChatGPT nailed the tone right out of the gate. The email was assertive without being combative, hit the right formal-but-human register, and even suggested including specific dates and amounts as placeholders. Solid.

Claude went a slightly different direction — warmer, almost conversational. The email it produced felt more like something a real person would actually send, which honestly might work better in a lot of situations. It also added a brief note explaining its tone choices, which was a nice touch.

Perplexity struggled here. It's really built for research and Q&A, not polished prose. The email it produced was functional but stiff — you'd want to rewrite most of it.

Gemini was competent but a little generic. It felt like it was playing it extremely safe, which isn't always what you want when you're trying to get your $87 back.

Winner for emails: Claude, narrowly over ChatGPT.

Task 2: Creative Writing — The Opening Chapter Test

This is where things got interesting (and occasionally weird). We asked each tool to write the opening 300 words of a psychological thriller set in a small Pacific Northwest town. No other constraints.

ChatGPT produced something competent and atmospheric — moody rain, a protagonist with a secret, a body discovered in a lake. It hit every genre beat correctly. A little too correctly. It felt like it had read a thousand thrillers and averaged them into one.

Claude took a more unusual approach. The opening it generated had a genuinely unsettling voice, leaned into unreliable narration from the first sentence, and avoided the most obvious tropes. It wasn't perfect, but it felt like it had an actual point of view. Of all the outputs, this was the one we'd most want to keep reading.

Gemini gave us something polished but oddly cheerful for a thriller. It kept softening the dark edges in ways that undercut the tone. Strange choice.

Perplexity produced a brief outline rather than actual prose, then apologized and tried again. The second attempt was fine but forgettable.

Winner for creative writing: Claude, and it wasn't particularly close.

Task 3: Coding Help

We brought in a real problem — a React component that was re-rendering way too aggressively, causing performance issues in a side project. We described the behavior and asked for help diagnosing and fixing it.

ChatGPT crushed this. It identified three likely culprits within its first response, explained each clearly without being condescending, and offered working code fixes for all three. This is where GPT-4o genuinely earns its keep.

Claude was also excellent here — arguably better at explaining why the problem was happening, which is useful if you're trying to actually learn rather than just copy-paste a fix.

Gemini did well, leveraging its Google ecosystem knowledge, though it occasionally cited documentation in ways that were slightly out of date.

Perplexity was the surprise — it pulled in current Stack Overflow threads and documentation links alongside its answer, which gave the response more credibility and made it easier to verify.

Winner for coding: ChatGPT for speed and accuracy; Claude if you want to understand what's actually going on.

Task 4: Brainstorming

Fifty names for a small-batch hot sauce company targeting foodies who take themselves slightly too seriously. Go.

Every single tool had fun with this one. ChatGPT gave us a solid, varied list with a good mix of punny names and genuinely clever ones. Claude leaned into the absurdist end of the spectrum in a way that made us laugh out loud — "Capsaicin Diplomacy" is a real suggestion it made, and we kind of love it. Gemini played it safer. Perplexity kept trying to tell us about existing hot sauce brands, which, not what we asked for, buddy.

Winner for brainstorming: Claude for pure creativity; ChatGPT for reliable volume.

The Hallucination Problem

We need to talk about this, because it's still a real issue. When we asked all four tools about specific recent events and niche technical topics, every single one made something up at least once. Perplexity was the most transparent about uncertainty and the most likely to cite sources. ChatGPT and Gemini confidently stated incorrect things more often than we'd like. Claude tended to flag uncertainty more clearly, but it still slipped up.

The takeaway: never use these tools as a source of truth for anything you can't verify independently. They are writing assistants, not encyclopedias.

So Are They Worth Paying For?

Honest answer? It depends entirely on how you're using them.

If you write a lot — emails, reports, creative projects — Claude Pro and ChatGPT Plus are both genuinely worth $20/month. Claude feels more like a thoughtful collaborator; ChatGPT feels more like a fast, capable assistant. Neither is universally better.

If you're primarily doing research, Perplexity's Pro plan makes a lot of sense, especially with its source-citation approach.

If you're already paying for Google One, Gemini Advanced is included and it's a solid all-rounder — not the best at any single thing, but good enough at most things.

Free tiers are legitimately useful for casual use. But if you're hitting the limits regularly, the paid upgrades are hard to argue with at this price point.

Bottom Line

The AI writing tool space is genuinely good right now — better than it was even a year ago. None of these tools are magic, and none of them are going to replace the judgment and voice that a real human brings to writing. But as productivity tools? As brainstorming partners? As a first-draft generator when you're staring at a blank page at 11pm? They're legitimately useful.

Just, you know. Read what they produce before you hit send.