I Tested 20 AI Tools for 30 Days — Here Are the Only Ones Worth Using
Thirty days ago I made a decision that my wife thought was ridiculous.
I told her I was going to spend the entire month testing AI tools. Not casually — properly. Every tool I'd been hearing about, reading about, or seeing recommended in newsletters and YouTube videos and Reddit threads. I was going to actually use them for real work, track what happened, and report back honestly.
She said "that sounds like a lot of work for a blog post." She wasn't wrong. But here we are.
I tested 20 tools over 30 days. I kept a running document with notes after each session — what I used it for, what it produced, whether I'd actually use it again if nobody was watching. Some tools I tested for three days before giving up. Some I ended up using every single day by week two. A few surprised me completely.
This is the honest version of what happened.
The Tools That Didn't Make It — And Why
Let me start here because most comparison articles skip this part.
I tested four different AI writing assistants that I won't name specifically because the point isn't to embarrass anyone. What I'll say is this — three of them produced content that was technically correct, grammatically fine, and completely soulless. Reading it felt like eating something that looked like food but had no taste. You couldn't find anything wrong with it. You also couldn't find anything right with it.
One of them had an interface so confusing that I spent twenty minutes trying to figure out where the output actually appeared on screen. I'm not exaggerating. Twenty minutes. I closed it and never went back.
Two AI image tools made the cut onto my list based on reputation and then immediately fell off based on results. The images they produced looked like what happens when you describe a concept to someone who has never seen the real thing. Technically it had the right elements. Emotionally it had nothing.
I also tested three productivity and scheduling AI tools that promised to organize my entire work life. One of them reorganized my calendar in a way that I genuinely needed three days to untangle. That was a memorable experience in the wrong direction.
The Ones That Actually Survived
ChatGPT — I know, everyone says this. But I tested it alongside five other language models during the same period and the gap is real in daily use. Not always in raw capability — some competitors matched or exceeded it on specific tasks. But for general, varied, unpredictable work — the kind where you need something different every hour — nothing matched its consistency.
What I used it for most: first drafts of everything, thinking through problems out loud, explaining things I didn't understand, and writing emails I didn't want to write. By week three I had stopped noticing I was using it. It just became the first place I went when I hit friction.
Perplexity AI — This one genuinely surprised me. I expected to use it occasionally for research. I ended up using it more than any other tool on this list except ChatGPT. The combination of real-time web search and direct answers with citations changed how I handle any question that requires current information. I stopped opening Google for most things by day twelve. That's not something I expected to write.
Midjourney — The learning curve here is real and I want to be honest about that. The first three days were frustrating. The Discord interface feels dated. Getting prompts right takes practice that nobody warns you about properly.
But by day eight something clicked. I started understanding the prompt logic. The outputs went from inconsistent to genuinely impressive. By week three I was producing images for this blog that I would have paid a designer for six months ago. The quality ceiling is higher than anything else I tested in the image category. It earned its place on this list the hard way.
Grammarly — I almost didn't include this because it feels too obvious. But I tested three Grammarly alternatives during the month and none of them touched it for the combination of accuracy, integration, and how unobtrusively it sits in your workflow. It's in my browser, it's in Google Docs, it catches things silently and moves on. The alternatives either missed too much or interrupted too often. Grammarly stayed.
Notion AI — I was a Notion user before this experiment. Adding the AI layer changed it significantly. The meeting summary feature alone — paste rough notes, get clean action items — saved me real time every week. I tested two other AI-powered workspace tools during the month and they both had features Notion AI lacked, but neither of them had the combination of solid core product plus useful AI additions that made Notion AI worth staying with.
Otter.ai — I tested four transcription tools. Otter won on accuracy for normal conversational speech and on the quality of its AI summaries. The others were close enough on raw transcription that the difference wasn't decisive. But the summaries Otter produced were consistently more useful — they pulled out the actual decisions and action items rather than just highlighting random sentences. For anyone in regular meetings, this one earns its place.
ElevenLabs — I tested this skeptically because I'd heard the voice quality claims before and been disappointed. I was not disappointed this time. I used it to produce narration for two short videos during the month. Nobody who watched those videos asked whether the voice was AI-generated. One person asked who the narrator was. That's the bar I was testing against and it cleared it.
The One That Surprised Me Most
Leonardo AI — I hadn't planned to test this one. Someone mentioned it in a comment on an earlier article and I added it to the list almost as an afterthought.
It became one of my most used tools by the end of the month.
The free tier is more generous than almost anything else I tested. The interface is cleaner than Midjourney and more powerful than DALL-E for stylized work. For blog visuals, social content, and any image that needs a specific illustrated or artistic style — it produced results faster and with less prompting frustration than the more famous options.
It's not better than Midjourney at the very top end of quality. But for the volume of work most people actually need to produce, it's closer than the reputation gap suggests. And the free tier means most people can use it properly without paying anything.
What I Learned That Nobody Writes About
Here's the thing about AI tools that a month of daily use made very clear.
The tools that stuck weren't the ones with the most features. They were the ones that removed friction from something I was already doing. ChatGPT removed friction from starting. Perplexity removed friction from researching. Grammarly removed friction from finishing. Otter removed friction from remembering.
Every tool I dropped — and I dropped eleven of them — had one thing in common. They added steps to my workflow instead of removing them. They made me think about the tool instead of the work. That's the test worth applying to any AI tool you're considering. Does using it make the work easier or does it make the tool management harder?
The ones on this list passed that test. Clearly and consistently over thirty days of real use.
Should You Test 20 AI Tools?
Honestly — no. Don't do what I did.
Pick one problem you have right now. Find the tool on this list that directly addresses it. Use it every day for two weeks before you try anything else. The people who get the least out of AI tools are almost always the ones who signed up for everything at once, got overwhelmed, and quietly went back to doing things the old way.
My wife was right that this was a lot of work for a blog post. But if it saves you from spending thirty days figuring out what I already figured out — it was probably worth it.
Even she admitted that. Eventually.
0 Comments