The problem this page exists to solve

Most published pricing for AI tools is wrong. Not maliciously: it was accurate when written and nobody went back. Vendors in this category restructure plans every few months, change billing units, and rename tiers, and the articles ranking for their names were written before the last change.

So the first thing we can offer isn’t better opinions. It’s a date.

Every figure carries a check date

Every pricing table on this site sits under a strip showing when we last read the number and where we read it. If that date looks old to you, treat the number as suspect and follow the link. We’d rather you checked the source than trusted us.

We re-check the tools that appear in our most-read pages every quarter, and immediately whenever a vendor announces a change.

Three levels of use, stated on every page

We tell you how well we know a tool rather than implying we know it well.

Used in production — we pay for it and use it for our own work. We can tell you what breaks on a Tuesday.

Hands-on — we’ve run it on real tasks, enough to judge output and workflow, but it isn’t part of how we work.

Researched — we’ve read the documentation, the pricing pages, the terms, and what users report, and we haven’t run a paid account ourselves.

Most pages on a new site are researched, and saying so is the point. A site claiming hands-on testing of forty tools is claiming something implausible.

What we do when sources disagree

Often. Five guides will give five prices for the same plan.

We don’t average them and we don’t publish a sixth number. We go to the vendor’s own page. If the vendor’s own documents contradict each other, which happens, we say that they do and explain the likely mechanism rather than picking the reading that suits the argument.

What we do when we can’t verify

We leave it blank and label it.

You’ll see tables here with a “price checked” column where some rows have a date and some don’t. That’s deliberate. A figure we haven’t opened ourselves is either marked as reported, with its source and date, or it isn’t there.

An incomplete table that tells you what it doesn’t know is worth more than a complete one that quietly guesses.

How we price things

We price a finished job rather than a plan, because plan prices tell you very little.

One 10-minute video with retakes counted. One year of weekly publishing. One team of ten at the tier that includes what they actually need. Then we divide, and we show the arithmetic so you can check it or substitute your own volumes.

Where a number is our calculation rather than a vendor figure, the page says so.

What our scores mean

Some review pages carry an editorial score across use-case fit, output quality, ease of use, pricing, reliability and support. It’s an editorial judgement, reproducible from the reasoning on the page. It is not a measurement, and we’d rather you read the trade-offs than the number.

Getting it wrong

We will. When we do, we correct the page and say what changed rather than editing quietly.

One example already: an early article here recommended PlayHT. PlayHT shut down permanently on 31 December 2025 after Meta acquired the team, and we hadn’t checked. We removed it and the correction is noted in that article. Several pages elsewhere still recommend it, which is exactly the failure this methodology exists to avoid.

If you find something wrong, email us and we’ll fix it.

Last updated: September 2026

3. About

Slug: about Título SEO: About AiStackCorner Meta: Who runs AiStackCorner, what the site is for, and why it covers four categories instead of four hundred.

What this is

AiStackCorner works out what AI tools cost once you use them for real work.

Not what the pricing page says. What a finished video costs after retakes, what a team of ten pays when the seat count multiplies, what happens to the credits you didn’t use, and where the upgrade path stops buying you anything.

Why four categories

There are directories with thousands of AI tools in them. They’re useful for discovering that something exists and useless for deciding whether to pay for it, because nobody can keep three thousand pricing pages current.

We cover four categories: AI voice and audio, AI video, AI meetings and transcription, AI writing and content. That’s few enough that the numbers on this site can be true on the day you read them.

When one of the four has enough depth to stand on its own, we’ll consider a fifth. Not before.

Who writes it

Andreu. I run a small portfolio of software review sites, and this one exists because the AI tools space has the worst signal-to-noise ratio of anything I’ve written about: opaque credit systems, plans that change quarterly, and a lot of published advice that was true in March.

I’m not a full-time reviewer with a lab. Where I haven’t used a tool, the page says researched rather than implying otherwise. That’s set out in full on the methodology page.

How it pays for itself

Affiliate commission, disclosed on every page that carries a link and explained on the affiliate disclosure page. Every article also names at least one option that earns us nothing, because that’s the only thing that makes the rest credible.

Getting in touch

Corrections, disagreements and vendors who think we’ve got something wrong: hello@aistackcorner.com. If a price here is out of date, telling us is genuinely useful and we’d rather hear it from you than leave it wrong.