The short version: nobody can give you an exact count because new AI tools launch faster than anyone can catalogue them, but the biggest directories now track somewhere between 25,000 and 30,000 tools, and that number is out of date by the time you read it. The number itself doesn’t matter. What matters is that you don’t need to compare all of them, you need a filter that throws out 99% of them in the first ten minutes.
So how many AI tools are there, really?
I checked three of the biggest AI directories while writing this. There’s An AI For That, one of the oldest and most-cited tool directories, listed just under 13,000 tools in early 2024. By late 2025 that figure had pushed past 25,000. Futurepedia, a smaller but well-curated directory, sits at around 4,000 to 5,000 listed tools because it’s more selective about what gets in. Product Hunt’s AI category alone launches somewhere between 15 and 40 new AI products every single day.
None of these numbers agree with each other because they’re counting different things. Some count every ChatGPT wrapper someone built over a weekend. Some only count tools with paying customers. Some strip out anything that hasn’t been updated in six months. If you want a single honest answer: there are tens of thousands of AI tools, the real number is unknowable, and it grows by hundreds a week. Treat any site that gives you a precise figure with a pinch of salt, because that number was already wrong the moment it was published.
Why the count keeps exploding
Most of what gets called “an AI tool” isn’t a new piece of technology. It’s a thin interface wrapped around someone else’s model, usually OpenAI’s GPT-4 or GPT-5, Anthropic’s Claude, or Google’s Gemini, with a logo, a pricing page, and a landing page written by the same model it’s selling access to. Building one of these takes a solopreneur a weekend and about £50 in API credits. That’s why the market is flooded. It’s not that AI research is producing 30,000 breakthroughs. It’s that 30,000 people found the same three APIs and built 30,000 different front doors to them.
This is the part most “best AI tools” listicles won’t say plainly: a huge share of the tools cluttering these directories are functionally identical to each other, running on the same three underlying models, differentiated mainly by their onboarding flow and their price. Once you know that, tool comparison gets a lot less overwhelming, because you’re not choosing between 30,000 unique technologies. You’re choosing between a handful of underlying models and a much smaller number of different workflows built on top of them.
My £4,000 lesson in tool shopping
Two years ago a client hired me to sort out their content operation. Six-person marketing team, drowning in blog posts, social captions, and email copy. Before I got there, they’d already bought Jasper (£99 a month), Copy.ai (£49 a month), and a lesser-known tool called Anyword (£79 a month), and someone on the team had a personal ChatGPT Plus subscription they were using for the actual work, because it turned out to be simpler than any of the three paid tools they’d bought.
That’s £2,556 a year spent on tools that were mostly sitting unused, on top of the ChatGPT subscription that was doing the real work. When I asked why they’d bought all three, the answer was the same one I hear constantly: each tool had a great demo, a persuasive sales page, and a “free trial we thought we’d cancel.” Nobody had sat down and asked what the actual workflow needed before they bought anything. We cancelled two of the three tools within a week, kept Anyword for its headline-testing feature because nothing else did that specific job as well, and redirected the rest of the budget into training the team on prompting. Output quality went up. Spend went down by roughly £1,800 a year.
The uncomfortable bit nobody puts in the headline
Here’s the part that sits badly with a lot of AI content: most businesses don’t have an AI tool problem, they have a decision-making problem, and buying another tool is often how people avoid making the decision. It’s easier to sign up for a new £29-a-month subscription than it is to sit down for an hour and map out exactly what your team does, where the bottleneck sits, and what “good” looks like. The tool becomes a substitute for the thinking. I’ve watched agencies stack four separate AI writing tools on top of each other while the actual bottleneck was that nobody had written a decent brief in eighteen months. No tool fixes a missing brief.
If you take one thing from this post, take that: before you open a single comparison chart, write down the specific task, in one sentence, that you want AI to do. “Help with marketing” is not a task. “Turn a 40-minute client call recording into a first-draft blog post” is a task. The second sentence can be shopped for. The first one will send you down a rabbit hole of 30,000 tools with no way out.
A step-by-step way to choose one
This is the process I run with clients now, and it takes about 90 minutes, not weeks.
- Step 1: Write the task, not the category. Not “an AI tool for social media” but “generate three LinkedIn post drafts from one long-form article every Tuesday morning.”
- Step 2: Decide who has to use it. A tool your most technical person loves and everyone else avoids will be dead within a month. I’ve seen this kill more AI rollouts than any pricing problem.
- Step 3: Set a real budget ceiling before you look at anything. Most solid business tools sit between £15 and £100 a month per seat. If a shortlist option is charging £400+ a month, it needs to justify that against a tool at a tenth of the price, not just against “doing nothing.”
- Step 4: Shortlist three, never more. Search “[your exact task] tool” rather than “best AI tools 2026,” and stop after three relevant results.
- Step 5: Run a free trial with a real piece of your own work, not the demo content. A tool that looks brilliant on the vendor’s sample data can fall apart on your actual, messier data.
- Step 6: Check what happens when you leave. Can you export your data? Is your content locked in their format? This kills more good decisions later than people expect.
- Step 7: Give it 30 days, then measure, don’t just feel. Time saved, output produced, or revenue touched. If you can’t point to a number after a month, that’s your answer.
Match the tool to the exact category, not the umbrella term “AI”
Once you’ve narrowed the task down, the tool question usually gets far smaller than “how many AI tools are there,” because you’re now shopping inside one category, not thirty thousand tools. If the task is images for a product launch, the real question is narrower and already answered here, in how to choose an AI image generator for commercial use, because commercial licensing terms rule out most of the flashy options before you’ve even compared quality. If the task involves customer messaging, pricing for that category is wildly inconsistent and worth checking against what WhatsApp Business automation tools cost in 2026 before you sign a 12-month contract on a demo call. If you’re trying to make a website feel less static, the AI-powered widgets worth adding are covered in what interactive tools you should add to your website, and most of them are cheaper and simpler than people assume.
This narrowing-down problem isn’t unique to AI, by the way. It’s the same pattern I wrote about when someone asked me how many affiliate marketing platforms exist: the honest total is a big scary number, and the useful answer is always “far fewer than that, once you know your niche and your budget.” The same logic applies to social media competition tools, where the market looks enormous until you realise most businesses only need one or two features done well.
Red flags that mean skip it, regardless of the reviews
- No visible pricing page, only “book a demo.” Usually means the price is negotiated per lead’s budget, not per value.
- The demo video is all AI-generated voiceover and screenshots, with zero real customer shown using it.
- Reviews are all five stars and all posted within the same six-week window, which usually means an incentivised review push, not organic trust.
- The company has fewer than three people on LinkedIn claiming to work there, on a tool that’s asking for your card details and your customer data.
- It requires a 12-month contract with no monthly option at all, on a category (AI tools) that changes every quarter.
When to stop shopping and get help instead
If you’ve been through the step-by-step above and you’re still stuck between two or three tools, or if the honest problem is that you don’t yet know which tasks in your business are even worth automating, that’s the point where paying for an hour of outside judgement is worth more than another week of comparison charts. This is exactly what an AI implementation coach does: not sell you a specific tool, but sit down with your actual workflow and tell you which of the thirty thousand options is worth your money and which are noise. I’d rather charge a client for two hours of that conversation than watch them sign a £1,200 annual contract for a tool that duplicates something they already own.
One more practical point, since it comes up constantly with clients trying to measure whether a new tool is even working: if the tool touches your website or your marketing funnel, make sure you can see the impact in your analytics before you commit long-term. It’s worth knowing the real setup limits, which I covered in how many Google Analytics accounts you can set up, so you’re tracking the new tool’s effect rather than guessing at the end of the month.
The number that matters
Not 25,000. Not 30,000. The number that matters is one: the single task you’re trying to solve, measured before and after. Everything in this post is designed to get you to that one number faster, because the tool market isn’t going to slow down or get more organised. If anything, by this time next year the directories will be listing 40,000-plus tools, and the honest advice will still be the same: write the task down first, shortlist three, test with your own work, measure after 30 days, and ignore the other 29,997.
Frequently asked questions
How many AI tools exist in 2026?
The largest public directories, like There’s An AI For That, now list somewhere between 25,000 and 30,000 AI tools, but the true number is unknowable because dozens of new tools launch every day and nobody tracks tools that quietly shut down.
What’s the fastest way to choose an AI tool without wasting weeks comparing options?
Write down the exact one-sentence task you want the tool to do, shortlist no more than three tools that claim to do that specific task, test each with your own real work rather than the vendor’s demo content, and measure results after 30 days rather than after a demo call.
Are most AI tools just built on the same underlying models?
Yes, a large share of AI tools on the market are interfaces built on top of OpenAI’s, Anthropic’s, or Google’s models rather than original technology, which is why so many products in the same category feel nearly identical once you use them.
Is it worth paying for multiple AI tools that do similar things?
Usually not. Most businesses end up paying for two or three overlapping subscriptions because each one was bought separately without checking what the team already had, and a short audit of current tools against actual tasks typically cuts that spend by a third or more.
