I Reviewed 47 AI Tools and Here's What Nobody Tells You About This Space

Ash K — Wed, 08 Apr 2026 08:34:52 +0000

This article was originally published on Medium. I'm cross-posting here for the Dev.to community.

I spent the last few months doing something slightly obsessive. I signed up for every major AI tool I could find — writing tools, image generators, code editors, video makers, chatbots — and used each one for at least 20 minutes before writing a single word about it.

Not skimming the features page. Not reading someone else’s review. Actually sitting down, creating an account, and trying to do real work with each one.

That became RawPickAI, an independent review site I built because I couldn’t find one that did this basic thing.

Here’s what I learned along the way. Some of it might save you money. Some of it might just confirm what you already suspected.

“Free tier” is the most abused term in SaaS

This drove me crazy during testing. Nearly every AI tool advertises a free tier. But the range of what “free” actually means is absurd.

ChatGPT’s free tier gives you GPT-4o mini and limited access to GPT-4o. That’s genuinely usable. You can get real work done without paying a rupee.

Then you have tools where “free” means you can log in, look at the dashboard, and do absolutely nothing useful without upgrading. I’m not naming names, but if you’ve tried a few AI tools, you know exactly what I’m talking about.

My rule of thumb after testing 47 tools: if a free tier doesn’t let you complete at least one real task end-to-end, it’s not a free tier. It’s a demo with extra steps.

The pricing problem nobody talks about

I’m based in India. When I see “$20/month” on a SaaS pricing page, my first thought isn’t “that’s reasonable” — it’s “that’s ₹1,860 a month.”

For context, a decent coworking space in Delhi costs about ₹5,000/month. A single AI code editor subscription is a third of that. Stack three or four AI tools and you’re paying more for software than for a place to sit and use it.

This is why I started showing pricing in both USD and INR on every review. Not because the conversion math is hard — anyone can Google it — but because seeing ₹1,860/mo next to $20/mo changes how the price feels. And how a price feels determines whether you’ll actually keep paying for it after the first month.

No other AI review site I found does this. Which is strange, because India has the second-largest internet population in the world.

Most review sites have never opened the tool they’re reviewing

This is the uncomfortable truth about the AI review space. A huge number of “reviews” online are rewritten feature lists from the tool’s own marketing page. The reviewer signed up, maybe, took a screenshot of the dashboard, and then wrote 2,000 words about features they read about but never tested.

How can I tell? Because they never mention the friction.

They don’t mention that Cursor’s autocomplete is genuinely the best I’ve ever used, but the Pro plan at $20/month feels steep when VS Code with Copilot does 80% of the job. They don’t mention that Midjourney produces the most beautiful images of any generator, but the Discord-based workflow is actively hostile to beginners. They don’t mention that Claude is brilliant at long-form analysis but occasionally refuses to help with completely reasonable requests.
Real testing surfaces these things. Reading a features page doesn’t.

What I actually learned about the tools

After 47 reviews, some patterns emerged that surprised me.

The best tool isn’t always the most expensive one. Perplexity’s free tier is genuinely more useful for research than ChatGPT Plus at $20/month. Google Gemini gives you a lot of capability for free that would cost you money elsewhere. The value-for-money winners often weren’t the tools with the biggest marketing budgets.

Code editors are the most competitive category right now. Cursor, Windsurf, GitHub Copilot, and Tabnine are all fighting for developers, and the gap between them is smaller than the marketing suggests. I scored Cursor highest, but Windsurf is catching up fast and costs less.

AI video is still too expensive for most individuals. Runway, HeyGen, and Descript all produce impressive results. But the pricing starts at $12–24/month and the free tiers are extremely limited. If you’re not making money from video content, it’s hard to justify.

Writing tools have a ceiling. I tested ChatGPT, Claude, Jasper, Copy.ai, Writesonic, and Grammarly for writing. Here’s the thing — the AI-generated draft is never the finished product. Every single one of these tools produces output that needs human editing. The question isn’t “which one writes the best?” but “which one gets me closest to done?” And that answer depends entirely on what you’re writing.

Why I built a scoring system (and why it matters)

Every tool on RawPickAI gets scored on five dimensions: Ease of Use, Output Quality, Value for Money, Feature Depth, and Free Tier. Each one rated 0–100.

I did this because star ratings are useless. A tool with 4.5 stars tells you nothing. Is it easy to use but expensive? Powerful but confusing? Great features but no free tier?

Breaking it into five scores means you can see where a tool is strong and where it falls short. Cursor scores 93 on Output Quality but 72 on Value for Money — which tells you it’s excellent but pricey. That’s more useful than “4.5 stars.”

The uncomfortable business model question

Let me be transparent about something most review sites won’t say out loud: affiliate links create a conflict of interest.

If I earn a commission when you sign up for Tool X through my link, I have a financial incentive to rank Tool X higher. Every review site that uses affiliate links faces this tension. Including mine.

Here’s how I handle it: the scores are set before any affiliate consideration. The methodology is public. If a tool scores poorly, it scores poorly regardless of whether they have an affiliate program. I’d rather have a site people trust than one that makes a quick buck from a dishonest recommendation.

Is it a perfect system? No. But it’s more honest than pretending the incentive doesn’t exist.

What’s next

I’m currently at 47 reviews with 10 comparisons and 8 best-of lists. The site is self-funded, I run it solo from India, and I test every tool myself.

If you’re evaluating AI tools — especially if you care about what things actually cost in rupees — check out RawPickAI. The methodology page explains exactly how I score things, and the comparison pages are probably the most useful if you’re deciding between two specific tools.

And if there’s a tool you want me to review, I’m all ears.

I write about AI tools, honest reviews, and building things from India. If you found this useful, follow me here or check out rawpickai.com.

DEV Community: Ash K