by Anil Raj
I spent the last year testing over 20 AI tools — with my own money, my own time, and honestly, a lot of wasted weekends.
Some tools impressed me on day one and felt useless by day four. Some were so complicated that just watching the tutorial ate up half my afternoon. A few cost me real money before I realized they were doing nothing I couldn’t already do faster without them.
After all of that, here’s the one thing I know for sure — there is an enormous amount of hype in the AI tools space, and very few tools that actually change how you work.
This post is about the three that did. The three I use every single day without fail. But first, let me tell you how I ended up wasting months before I figured this out.
I Was Making the Same Mistake Everyone Makes
When the AI tools explosion happened at the end of 2024, my phone was getting hit with notifications every other day. “This new AI tool just killed ChatGPT.” “This one’s free and ten times better.” “You’re falling behind if you’re not using this.”
I was right there in that crowd, jumping on every new thing.
In just the first month, I had created accounts on Jasper, Copy.ai, Writesonic, Rytr, Notion AI, Otter.ai, Descript, Runway, Midjourney, Leonardo, Perplexity, You.com, Poe, Character.ai, and at least six or seven others I can barely remember now. Free trials for most, paid subscriptions for a few.
The result? I had 30 browser tabs open at any given time. My password manager was full of logins I hadn’t used in weeks. And my actual work output was roughly the same as before — maybe even a little worse because I was constantly switching between things.
A friend asked me one afternoon, “You’re using all these AI tools now — is your work actually faster?” I thought about it honestly for a moment. And the answer was no, not really.
That’s when I understood what was actually happening. I was collecting tools, not using them. And that is probably the most common mistake people make when they first get into this space.
What I Did Differently
I spent one month closing everything down. One by one, I deleted accounts or just stopped logging in. I kept only three things that I genuinely felt were making a difference, and I worked with just those three for a full month.
The results were better than anything I had seen when I was juggling twenty tools at once.
Here’s what those three tools are, why each one made the cut, and exactly how I use them.
Tool #1 — ChatGPT (GPT-4o)
Yes, I know. You’ve heard of it. You’ve probably used it. But if you’re using it like a chatbot — typing in a question and reading the answer — you’re getting maybe twenty percent of its actual value.
When I first used ChatGPT, I asked it to write a poem. It did. I thought, okay, this is just a fancy search engine with a personality. I didn’t take it seriously for months after that.
Then a client came to me with a deadline. They needed a 5,000-word detailed report on a niche topic I had almost no background in. Two days to deliver. I was genuinely stressed.
I decided to treat ChatGPT the way I’d treat a really smart colleague — I gave it full context, explained what the client needed, described the specific angle, told it who the audience was. The output didn’t just help.
It changed my entire understanding of how to use the tool. The difference wasn’t the tool. It was how I was talking to it.
Where ChatGPT actually earns its place in my workflow:
The first is research and outlining. Every time I start a new blog post or content project, I don’t open a blank document first. I open ChatGPT. I ask it to help me find the most useful angle for a specific audience, figure out what questions that audience is actually searching for, and build a structure that doesn’t just repeat what’s already everywhere online. What used to take me two to three hours of background reading now takes fifteen minutes.
The second is rewriting and editing. I write my first draft in my own voice, then I bring specific sections to ChatGPT and ask it to make them more conversational, or cut the parts that drag, or simplify something I’ve made too complicated. It doesn’t always improve things, but it does about seventy percent of the time — and that’s enough to make it worth doing.
The third is client communication. Proposals, follow-ups, delicate conversations where I need to say something carefully — I write my own draft first, then ask ChatGPT to refine it. It’s faster and usually more professional than what I’d send on my own.
One honest warning though: Never paste ChatGPT’s output directly into anything you’re publishing or sending. I made this mistake once. I sent a client an article that was basically ChatGPT’s output with light editing. They spotted it immediately. “This reads like AI wrote it,” they said. They weren’t wrong. That was an embarrassing moment, and it taught me the rule I follow now — use ChatGPT for thinking and structure, but write in your own words.
ChatGPT Plus is $20 per month. The free version still works but gets slow and inconsistent during peak hours. If you use it regularly, the paid plan is worth it.
Tool #2 — Claude (by Anthropic)
This one doesn’t make most people’s lists, and honestly, that might be exactly why it’s so good right now. It’s not overcrowded with hype yet.
I found Claude when I had to write a detailed 8,000-word white paper for a client. I tried ChatGPT first. The output was fine, but it felt thin. Like someone had gathered a lot of surface-level information and arranged it neatly without actually understanding any of it.
Someone I follow online mentioned Claude as an alternative. I tried it the same day, gave it the same brief, and within the first few paragraphs of the output I could tell the difference. There was actual depth.
The reasoning was visible. It felt like talking to someone who had genuinely thought about the subject
Where Claude pulls ahead: For anything over 2,000 words, Claude handles long-form writing better than any other AI tool I’ve used. It holds context longer in a conversation, which matters a lot when you’re building a complex document section by section. ChatGPT tends to lose track of earlier decisions. Claude doesn’t, or at least not as quickly.
On analysis and reasoning tasks, Claude is more nuanced and — this matters — more honest. ChatGPT has a habit of telling you what you probably want to hear. Claude will push back if something in your thinking is off. I’ve had it point out problems with my own content strategy that I hadn’t noticed. That kind of honest response is genuinely useful.
The conversational tone in Claude also feels more natural than most AI tools. It doesn’t feel like you’re querying a database. It feels closer to an actual back-and-forth with someone who’s paying attention.
A real example from my work: I needed to do a competitor analysis for a niche I was entering. I gave Claude five URLs and asked it to identify gaps in their content strategy — places where none of them were covering something the audience clearly wanted. The analysis it gave me was detailed, specific, and actually actionable. I built an entire three-month content plan from that one conversation. That used to be a full day of work. It took about twenty minutes.
Claude has a free version. The paid plan is $20 per month. I use both — ChatGPT for quick tasks and short-form work, Claude for anything that requires depth or extended thinking.
Tool #3 — Perplexity AI
This is the most underrated of the three. Most people either haven’t heard of it or they’ve dismissed it as “just another AI search engine.” That’s a mistake.
My first real moment with Perplexity happened when I was looking for current statistics for a blog post. I went to Google first. Ten tabs open, four of them worth reading, two of those outdated, and by the end I still wasn’t sure what the actual current number was.
I tried Perplexity with the same question. It gave me a direct answer, cited its sources, and the data was current. No hallucination. No made-up statistics that sounded plausible. Actual sourced information.
That’s the thing ChatGPT and Claude genuinely can’t do well — search the live web and give you accurate, up-to-date information. Both of them have knowledge cutoffs. Perplexity doesn’t. It’s pulling from the current web in real time.
How I use Perplexity every day:
Before writing any piece of content, the research phase now happens entirely in Perplexity. I ask it for the latest statistics on a topic, what’s happened recently in a space I’m covering, what the current pricing is for a tool I’m reviewing. It gives me sourced answers I can verify.
For competitor research, it’s faster than any other method I’ve found. I can ask what articles have recently been published on a specific topic and get a clear picture of what already exists.
When a new client comes to me from an industry I know almost nothing about, Perplexity gives me a ten-minute background education that actually holds up to scrutiny. I’m not walking into meetings with confidently wrong information anymore.
An honest limitation: The sources Perplexity cites are not always high quality. It will sometimes pull from mediocre sites. So I don’t use it as my only source for anything important. If a statistic or claim matters, I click through and verify it at the original source. But as a starting point and research accelerator, nothing else comes close.
The free version has limited daily searches. The Pro plan is $20 per month. I use Pro because it comes up in my workflow every single day.
What Happened to the Other 17+ Tools?
Fair question. The honest answer is that some of them are decent — but decent and necessary are two very different things.
Jasper does content writing reasonably well, but it’s hard to justify at $49 per month or more when ChatGPT and Claude are cheaper and better for most tasks.
Midjourney is genuinely impressive for image generation. But my work is text-heavy and images are secondary. If your work is visual, it might belong on your list.
Otter.ai and Fireflies are useful if you’re on a lot of calls and need automatic transcription. I’m a freelancer with limited meetings. If you’re in a corporate environment or you’re doing a lot of interviews or client calls, worth considering.
Notion AI is a good add-on if you’re already deep into the Notion ecosystem. But getting into Notion just for the AI features doesn’t make sense.
Runway and Pika — AI video tools that I think will become more relevant to my work in the next year or two. Right now they don’t fit what I actually do.
The other ten or twelve tools I tested? I genuinely can’t remember most of their names. That’s probably the clearest signal about whether something was useful.
How These Three Tools Work Together
Let me give you a concrete look at my actual workflow for writing a blog post, because the tools are more useful in combination than they are separately.
Research first — Perplexity. Once I have a topic, I go to Perplexity and look for current data, recent developments, and angles that are getting traction. This takes ten to fifteen minutes and gives me something real to work with.
Outline and angle — ChatGPT. I bring what I found into a ChatGPT conversation. I ask what the most useful angle would be for my specific audience, what structure would actually serve the reader rather than just look complete, and what questions people are searching for that haven’t been well answered yet. Another ten minutes.
Writing — me, with Claude nearby. I write the draft myself, in my own voice. When I hit a section I’m not happy with or I get stuck, I bring that specific part to Claude and ask for help. Not the whole piece — just the part that isn’t working.One to two hours depending on the length.
Final edit — ChatGPT. I paste the complete draft and ask what’s unnecessarily complex, what drags, where a reader might lose interest, and what’s unclear. Then I make those edits myself, in my own judgment.
This workflow used to take me five or six hours. Now it takes two to three, and the quality is genuinely better because I’m no longer skipping the research phase when I’m in a hurry.
One Warning I Want to Be Clear About
A lot of people make this mistake, and I made it too.
AI is a thinking partner, not a ghostwriter.
A few months into using ChatGPT, I started thinking I could just have it write things for me and lightly edit the output. I did that with several posts. The traffic was flat. The engagement was flat. One reader commented: “This feels like it was written by a machine — there’s nothing original here.” They were right.
AI tools can give you information. They can’t give you perspective. They don’t know what you’ve tried and failed at. They don’t know the specific thing you noticed that changed how you see a topic. They can’t replicate what makes your experience yours.
My rule is simple: AI handles structure and research, I handle voice and insight. That combination works. One without the other doesn’t.
What This Actually Costs
Here it is straight:
ChatGPT Plus — $20/month
Claude Pro — $20/month
Perplexity Pro — $20/month
That’s $60 per month total. Real money, especially if you’re just starting out.
But if these tools cut thirty to forty percent of the time you spend on work — and in my experience they genuinely do — then what you’re really paying for is the ability to do more in the same hours. If that translates into one extra client, one more piece of content per week, or consistently better output, the math starts looking different.
If budget is tight, my honest recommendation is to start with ChatGPT Plus only. Use it for a month and pay attention to whether it’s actually changing anything for you. If it is, add Claude. Then Perplexity. Or try all three free versions first and see which ones actually stick before you put money into any of them.
If You’re Just Getting Started
One piece of advice if you’re new to all of this and feeling overwhelmed by the options:
Pick one thing you do regularly — writing emails, doing research, drafting content, whatever takes the most of your time. Try doing that exact thing with ChatGPT for one week. Don’t change anything else. Just do that one task differently.
If it makes a real difference, pay for the plan and start exploring more. If it doesn’t, you’ve only lost a week of experimenting with a free tool.
Don’t open ten tools at once. That’s where the overwhelm comes from. That’s the trap I fell into. One tool, one use case, one month. That’s how you actually figure out what works.
The Bottom Line
After testing more than twenty AI tools, the lesson I keep coming back to is this — more tools does not mean more output. Usually it means the opposite. More switching, more confusion, more time managing tools instead of actually working.
Three tools. A clear workflow. Consistent practice. That’s what changed things for me.
ChatGPT for thinking, structuring, and quick tasks. Claude for depth, analysis, and long-form writing. Perplexity for research and anything that requires current, accurate information.
If you give these three a serious month, I think you’ll notice the difference. And if you try the workflow I described and something about it doesn’t work for your situation, drop a comment below — I respond to every one.
— Anil Raj