Why most AI tool reviews are useless (and what this blog does instead)

Here is how most AI tool reviews get written.

Someone with a content site and a few affiliate programmes decides AI tools are a good niche. They sign up for a free trial — sometimes not even that. They read the product's own feature page. They rephrase it. They add a star rating and a button that says 'Try [Tool] Free.' They collect their commission.

This is not reviewing. It's repackaging. And the internet is absolutely full of it.

If you've spent any time looking for honest guidance on whether a particular AI tool is worth subscribing to, you've probably noticed that most articles tell you everything except what you actually want to know: does it work, is it worth the money, and what are the catches?

What goes wrong

There are a few structural reasons why AI tool content is so consistently poor.

The first is incentive misalignment. The people writing these reviews are paid when you click and buy. They are not paid when you find the review useful. That's a small distinction that produces enormous consequences. A useful review might conclude that a tool isn't worth your money. A monetised review almost never does.

The second is that AI tools have very aggressive affiliate programmes. Many pay 20–30% recurring commission — meaning the reviewer earns every month you stay subscribed. That's a powerful financial incentive to tell you everything is great.

The third is the pace of the space. New AI tools launch constantly. Keeping up requires actually using things, which takes time. It's far faster to write about a tool than to use it seriously. Most review sites choose speed.

The result is a landscape where you can read twenty articles about the same tool and come away knowing less than when you started. You've consumed a lot of feature lists. You haven't received any honest signal.

What an honest review actually requires

I run several publishing projects simultaneously — newsletters, blogs, a software product. I depend on these tools. I pay for them out of my own pocket, which means I have a genuine interest in knowing whether they're worth it before I subscribe and whether they remain worth it over time.

That's the only qualification that matters for this kind of writing. Not a background in tech journalism. Not a partnership with a tool company. Just actual use, over actual time, under actual conditions.

An honest review needs to answer a handful of questions that most reviews don't touch.

Does it actually work the way it claims to?

Tool landing pages are works of optimistic fiction. The gap between what a tool promises and what it delivers in daily use is often significant. The review should close that gap.

What are the real costs?

'Starting from £X/month' is almost never what you'll pay. What tier do you actually need to use the tool properly? What happens when you hit usage limits? Are there hidden costs — extra seats, export fees, API charges?

What breaks or frustrates?

Every tool has failure modes. Things it handles badly. Edge cases that matter. A review that doesn't address these is incomplete. I'm not looking to be cruel about software — I understand that building things is hard. But if something reliably produces mediocre output in a particular situation, you should know before you subscribe.

What's the verdict?

Every review on this site ends with one of three verdicts: Recommended, Conditional, or Not recommended. No hedging. If I think a tool is worth your money, I'll say so and explain why. If I think it's overpriced, underbuilt, or outclassed by something cheaper, I'll say that too.

What this blog is

The Practical AI covers AI tools for writers, newsletter publishers, and one-person businesses. Not developers. Not enterprise teams. People who need to produce things — content, products, software — and want to know which tools actually help them do that.

I publish twice a week. Wednesdays are longer pieces — full reviews, comparisons, or in-depth workflow posts. Fridays are shorter: a single sharp observation, a quick take, something worth reading in three minutes.

The newsletter goes out on Wednesdays with the main piece. It's free. You can subscribe on the site.

What this blog isn't

It isn't comprehensive. I can't review every tool — there are too many and new ones arrive constantly. I review what I use or have a genuine reason to test. If a tool isn't here, it usually means I haven't used it seriously enough to say anything worth reading about it.

It isn't impartial in the sense of having no views. I have views. I'll share them. What it is, is honest — I have no incentive to tell you a tool is good if I don't think it is, and I won't.

Some links on this site are affiliate links. I'll note when that's the case. It doesn't change the verdict.

Where to start

If you're new here, the best place to start is the post I'm publishing this Friday: the six AI tools I currently pay for and why. After that, the Claude review goes up next Wednesday — that's the tool I use most, and the obvious place to start if you're trying to understand what the current generation of AI writing tools is actually capable of.

More soon.

— Ellis

Some links on The Practical AI are affiliate links. This post contains none. Full disclosure policy at /disclosure/