AI Jobs: What 2,400 Company Boards Actually Show

Search "AI jobs" and you get two kinds of result. Aggregator pages listing whatever they've syndicated, and career-advice posts telling you the ten hottest AI roles of the year. Neither tells you what the market actually looks like.

We watch about 2,400 companies publish roles directly to their own applicant tracking systems, hourly. That gives us a view of AI hiring that isn't filtered through what companies chose to advertise. Some of what we see contradicts the standard story.

The market is a long tail, not a handful of labs

Ask most people to name AI employers and you'll get OpenAI, Anthropic, maybe DeepMind or Scale. Those companies are real and they're hiring. But they are a rounding error in the actual volume.

Half of all live AI roles come from 289 different companies. The median AI employer has exactly two open AI roles. Not two hundred, not twenty. Two.

This is the single most important thing to understand if you're looking for this kind of work, and it explains why the search feels so frustrating. The market isn't concentrated somewhere you can go and browse. It's spread across thousands of companies posting one or two roles each, most of which you've never heard of and none of which are running recruitment marketing campaigns.

It also explains the aggregator problem. A company with two open roles and no talent brand isn't paying to syndicate listings. The role goes up on their Greenhouse or Ashby board, it sits there, and it may never appear anywhere else. The big job sites are excellent at surfacing roles from companies that want to be found. They are structurally bad at the long tail, because the long tail doesn't advertise.

Almost nothing is junior

Of all live AI roles, roughly 30% carry a senior, staff, principal, lead, or director signal in the title. About 6% carry a junior, graduate, or intern signal.

That five-to-one ratio is worth sitting with, because it contradicts most of what's written about breaking into AI. The volume of "learn these skills and get an AI job" content implies an entry-level market. The postings suggest otherwise: companies are overwhelmingly hiring people who have already done the work somewhere else.

If you're early in your career, the practical read is that the direct route is narrow. The roles that exist are titled things like machine learning engineer, applied AI engineer, and AI product manager, and they mostly want prior shipped experience. The indirect route — joining a company that's adopting AI rather than building it, and moving into the work internally — is where most of the actual entry points are, and none of those jobs are titled "AI engineer".

The most common exact titles we see, in order, are machine learning engineer, AI engineer, senior machine learning engineer, applied AI engineer, and senior AI engineer. Three of the top five are explicitly senior.

AI listings sit around longer than average

Here's a finding that surprised us. About 55% of currently live AI roles were posted within the last 30 days. Across our whole catalog, that figure is 63%.

So AI listings are, on average, slightly older than everything else. That's the opposite of the "AI hiring is exploding" framing, and it's a useful corrective if you're applying to things.

We'd be careful about over-reading it — a role staying open longer can mean weak demand, or it can mean the requirement is genuinely hard to fill and the company is being picky. Both produce the same number. But it does mean the panic framing is wrong: these aren't roles that vanish in 48 hours. Being fast still helps, and being first still helps more, but the AI market is not obviously faster-moving than any other engineering market right now.

One caveat we'll state plainly rather than bury: we don't yet have enough history to publish reliable numbers on how quickly AI roles close after posting. We'll write that up when we have a clean 90 days of data, and not before.

AI companies use different hiring software, and that tells you something

Across our whole catalog, roughly 74% of companies post through Greenhouse and 25% through Ashby. Among companies posting AI roles, Ashby's share rises to 37%.

That sounds like trivia. It isn't, quite. Ashby is a newer applicant tracking system that skews toward well-funded, recently-founded startups — it's what a company that raised a Series A in the last few years tends to adopt, where an older or larger organisation is more likely to already be on Greenhouse. The ATS a company uses is a rough proxy for its vintage and stage.

So the ATS skew is a measurable signal that AI hiring is concentrated in younger companies. Which lines up with the long-tail finding: lots of small, recently-funded teams, each hiring a couple of people, none of them with a recruitment marketing budget.

For you as an applicant, the practical version is that a meaningful slice of AI roles live on boards that never syndicate anywhere, at companies whose names you'd have to already know to search for.

Remote is more common, but not as common as you'd think

About 17% of AI roles are remote, against roughly 13% across our full catalog. Genuinely higher, but not the fully-distributed picture the discourse suggests.

Geographically the concentration is stark. San Francisco leads by a wide margin, followed by London, New York, Singapore, and Berlin. If you're in Europe, London and Berlin are where the density is, though the long tail means individual roles turn up in a lot of smaller cities.

We should be honest about a limitation here: location data on job postings is free text that employers write themselves, so it's messy. "Remote", "Remote (US)", and "Remote - EMEA" are three different strings meaning three different things. Treat the geography as directional.

What this means if you're looking

Putting the five findings together, the shape of the advice is fairly clear.

Assume the roles you want are not on the big job sites. Not as a conspiracy — just as a structural consequence of a market made of small companies posting two roles each with no reason to advertise. If you only look where advertising happens, you're seeing a filtered slice.

Target companies before you target postings. Because the median AI employer has two open roles, company-level targeting beats keyword-level searching. Work out which twenty or fifty companies you'd actually want to work at, and watch their boards directly.

Adjust for the seniority reality. If you're not already senior, the direct AI-titled route is narrow, and the honest path is usually adjacent — a role at a company doing AI work, in a function you can already do, and moving toward the work from inside.

Don't panic about speed, but don't be slow either. The 55% figure says AI roles aren't uniquely fast-closing. Applying in the first few days still matters, because the first applications get read most carefully, but you're not in a race measured in hours.

What's live right now

These are the ten most recently posted AI roles in our catalog, straight from company boards.

How we got these numbers

Because you should know what you're reading.

We poll the applicant tracking systems of roughly 2,400 companies every hour — Ashby, Greenhouse, Lever, Workable, Recruitee, SmartRecruiters, Teamtailor, and Personio — and record roles when they appear. There's no syndication step, so we see a posting at roughly the same time the employer's own careers page does.

An "AI role" here means a role whose title matches terms like AI, machine learning, ML, LLM, NLP, deep learning, computer vision, generative AI, MLOps, or prompt engineer. That's a title-level definition, so it captures AI roles anywhere rather than all roles at AI companies. A backend engineer at an AI lab won't be counted; an AI product manager at a bank will be.

We exclude a category of bulk listings that would otherwise distort everything: research-panel, annotation, and AI-training gig work, where a handful of employers post thousands of near-identical listings. Left in, those six employers alone would account for more than a quarter of the total and would make the market look far more concentrated than it is.

Our catalog is European-focused, meaning companies posting roles in European cities rather than only companies headquartered in Europe. That shapes the geography above and is worth knowing when you read the city list.

Frequently asked questions

Are AI jobs actually growing, or is this hype?

Both things are true at once. The volume of open AI roles is genuinely large and spread across thousands of companies rather than a few labs. But our data also shows AI listings sitting open slightly longer than the catalog average, which is not what a red-hot market looks like. Treat individual claims about explosive growth with some scepticism, including ours.

Can I get an AI job without an AI background?

Directly, it's hard. Roughly 6% of AI-titled roles carry a junior or graduate signal against 30% at senior level and above, so the entry-level door is narrow. The more common route is joining a company that's building or adopting AI in a function you already have experience in, and moving toward the work internally. Those roles are rarely titled as AI jobs, which is exactly why they're less contested.

Why can't I find these jobs on the big job sites?

Because the median AI employer has two open roles and no recruitment marketing budget. Syndicating listings to aggregators costs money and effort that a company hiring two people often doesn't spend. The role goes live on their own applicant tracking system and frequently never travels further. Aggregators are good at surfacing roles from companies that want to be found and structurally weak on the long tail.

What is a prompt engineer job now?

Far rarer than the 2023 discourse suggested. Prompt engineer barely registers as a distinct title in our data. The skills got absorbed into machine learning engineer, applied AI engineer, and AI product roles rather than becoming a standalone career. If you're targeting that title specifically, widen your search.

Where are most AI jobs located?

San Francisco leads by a substantial margin, then London, New York, Singapore, and Berlin. About 17% are remote, against roughly 13% across our full catalog. Location fields on job postings are free text written by employers, so treat any geographic breakdown as directional rather than precise.

Watching for AI roles as they go up

If the long-tail finding is right, the practical problem isn't knowing that AI jobs exist. It's that they're distributed across thousands of company boards, and checking them by hand is a job in itself.

That's what we built ReqHunt for. Set your criteria once, and every hour we check new postings from company boards against them.