The AI Categories Founders Keep Building In (and the Ones Nobody Touches)

"The AI market is crowded" is not a finding. It is a mood.

It gets repeated in every launch thread and every investor memo, and almost nobody attaches a number to it. Crowded compared to what? Crowded in which part of it? Crowded with products that are still shipping, or crowded with landing pages?

I can attach a number to a small slice of it, so here it is.

I run Nick Launches, a weekly launch directory, and I benefit if you decide to submit something to it. That bias is worth naming at the top rather than leaving it for you to notice halfway down. What follows is a count of what founders actually submitted, not a recommendation about what you should build.

The reason it might be useful to readers here is that it points the opposite way to most of the data on this site. AITrustList measures where users go. This measures where builders go. Those are two different questions, and the categories where the answers disagree are the interesting ones.

Key facts

1,442 products are live on Nick Launches as of 23 August 2026, spread across 38 categories.
613 of them carry the AI tag. That is 42.5 percent of the entire directory.
The five largest categories account for 55.1 percent of all category tags. The nineteen smallest, half the categories, account for 10.8 percent.
The median category holds 34 products. The largest holds 613.
The smallest category, Sustainability, holds 3.

How the count was made, and what it cannot tell you

The numbers come from the category index on Nick Launches, read on 23 August 2026. Every category page publishes its own count, so this is a single read of a public page rather than an estimate.

Four limits, stated up front, because a count like this is easy to over-read.

It is one launch feed, not the market. Plenty of products never submit to a directory at all, and the ones that do skew toward founders who are actively looking for distribution.

Categories are self-selected. Founders pick their own tags at submission. "Other" holds 217 products, which is partly a real bucket and partly a tagging failure.

Products can carry more than one tag. The 38 categories hold 3,449 tags between them across 1,442 products, roughly 2.4 each. So a category's share of the tag pool is not its share of the directory.

Counts are cumulative, not a rate. A category with 100 products might have collected them over sixteen weeks or in one month. The total does not say which.

AI is not a category any more, it is the default setting

613 of 1,442 live products carry the AI tag. Nothing else is close except Productivity at 562, and the overlap between those two is heavy.

But the more useful reading is what sits underneath. The AI category is 31 pages deep, and scrolling it does not feel like scrolling one category. It feels like scrolling a directory that happens to have an AI tag applied to it.

Developer Tools has 255 products. Content Creation has 126. Marketing has 253. Read the descriptions and most of those are AI products too. They just tagged themselves by what the product does rather than by what it is built on.

Which means "the AI market is crowded" is close to meaningless as a statement about competition. You are not competing with 613 products. You are competing with the ten or twenty that solve the same job for the same person, and the AI tag will not help you find them.

That is the practical version of the point this site keeps making with traffic data. The label is not the market. The job is.

And the other end of the same table.

The drop is steep and then it flattens. Rank 20 holds 34 products and nothing below it reaches that number, which is how nineteen categories end up sharing 10.8 percent of the tags between them.

Empty categories are usually empty for a reason

The tempting read of that second table is that Legal, Music and Sustainability are wide open.

Sometimes that is true. Usually it is not, and the reasons are boring.

I cannot prove why any given category is thin, and neither can a count. But the obvious candidate explanations are unglamorous ones. Legal buyers are slow, regulated and reference-driven, and none of that is fixed by shipping faster. Sustainability tooling is usually sold to a compliance function on a procurement cycle measured in quarters. Music has a rights problem that predates anyone reading this.

Low supply is a signal, but the signal is ambiguous. It can mean nobody has looked. It can also mean plenty of people looked, did the maths on the sales cycle, and went to build a Chrome extension instead.

This is where this site's data does something a supply count cannot. If a category is thin on the builder side and there are products in it pulling real search traffic, that gap is worth investigating. If it is thin on both sides, the market is probably telling you something.

Reality check: an empty category is a hypothesis, not an opportunity. The cheapest way to test it is to find the two or three products that are already there and work out whether they are growing or just present.

Where the two lists disagree

Put a supply count next to a traffic ranking and you get four quadrants. Three of them are useful.

High supply, high growth. Crowded and still expanding. Hard, but the demand is real and you can see who is capturing it. Most of AI and Productivity sit here.

High supply, flat traffic. The trap. Lots of founders arrived, nobody is being rewarded, and the crowding is a lagging indicator of hype rather than a leading indicator of demand. Worth checking before you commit six months.

Low supply, real traffic. The one worth digging into. Someone is getting search demand in a category almost nobody is building for.

Low supply, no traffic. Leave it.

You cannot get those quadrants from either dataset alone. A supply count tells you how many people had the same idea. It says nothing about whether any of them are getting users. Traffic rankings tell you who is getting users, and nothing about how many people are lining up behind them.

What I would actually do with this

Find your category in both places. Count the products on the supply side. Then check whether the ones with traffic are the biggest ones or the newest ones, because the answer changes what you should copy.

If your category is crowded and growing, your problem is positioning and distribution, not the idea.

If your category is crowded and flat, get a second opinion before you build.

If your category is nearly empty, spend an afternoon finding out why before you decide you found something. The reason is usually written on the pricing page of whoever is already there.

Nick Launches is a weekly launch directory. The category counts above were read from its public category pages on 23 August 2026, and the raw capture is available to anyone who wants to check the arithmetic.