18% of job ads mention AI. Only 2% are AI jobs — and two thirds of the mentions come from 22 companies that say it in every ad.

FACTANKER read 25,967 complete job descriptions from 224 employers across four hiring platforms. The gap between what companies say about AI and what they hire for is measurable — and every number below links to the postings it came from.

Cite as: FACTANKER, https://factanker.com/record/ai-in-job-ads-vs-ai-jobs

What we measured — and what we did not

“How many jobs are AI jobs?” is usually answered with a survey or a keyword count on a job board’s search page. We did neither. FACTANKER observes employer career pages directly: on 13 September 2026 the registry held 46,155 currently listed postings from 224 employers across four applicant-tracking platforms (Greenhouse, SmartRecruiters, Lever, Ashby), plus 5,371 postings observed earlier and since withdrawn. For 25,967 of the listed postings we hold the complete description text, not just the title.

That let us separate two questions that are usually collapsed into one: how many postings are for AI work (the job title says so), and how many postings talk about AI (the description mentions it anywhere). The answers are five percentage points and one large caveat apart.

The result

1,059 of 46,155 listed postings — 2.29% — carry an AI term in the job title. That is the count of roles a company is actually hiring for: AI engineers, machine-learning scientists, computer-vision leads, MLOps specialists, data scientists.

Among the 25,967 postings where we hold the full text, 4,664 — 18.0% — mention AI somewhere in the description. Eight times the title rate. That is the number most often read as “AI is coming for every job.”

It does not survive contact with the source text. 3,093 of those 4,664 mentions — 66% — come from just 22 employers who mention AI in 90% or more of every posting they publish, regardless of the role. Pinterest opens each ad with the same sentence about AI being “a powerful partner.” Every Okta, Elastic, GitLab, Figma and Anaplan posting in the registry carries such a line. It is company self-description, repeated verbatim — not a requirement of the job.

Strip those 22 employers’ house sentences out and AI is mentioned in roughly 1,571 postings — 6.1% of the ones we can read in full.

The boilerplate: 22 companies, 3,093 mentions

These are the employers whose postings mention AI at a rate of 90% or higher. The right-hand column is what actually distinguishes them: how many of those same postings are for AI roles.

Two further employers in this group are omitted from the table because we found, while preparing this piece, that our registry has attached their hiring board to the wrong legal entity. Their postings are counted in the totals above; we will not print a company name we have reason to doubt. The correction is queued.

Elastic mentions AI in all 356 of its postings and is hiring for 12 AI roles. That ratio — 30 mentions per role — is what “AI everywhere” looks like when you read the source instead of counting keywords.

Where the actual AI roles are

By title share, among employers with at least 100 listed postings:

ServiceNow is hiring for more AI roles than any other employer in the registry in absolute terms. Reddit has the highest concentration: more than one in four of its open positions names AI, machine learning or data science in the title.

The platform a company hires on predicts the rate better than its industry does. AI title share by platform: Ashby 7.0% (625 postings), Lever 7.1% (1,003), Greenhouse 5.9% (12,448), SmartRecruiters 1.4% (32,079). SmartRecruiters carries the large service and retail employers — the single biggest posting volume in the registry is a pizza delivery chain with over 24,000 open positions — and it is where the AI hiring boom is, by this measure, almost absent.

AI roles are not more disposable than other roles

Of the 5,371 postings we observed and later found withdrawn, 2.09% carried an AI term in the title — against 2.26% among the postings still listed. Within the precision this observation window allows, AI positions come down off career pages at about the same rate as everything else. Whatever else is true about AI hiring, these roles are not being posted and pulled faster than the rest.

What this does not say

We hold complete description text for 56% of listed postings; coverage is 100% on Greenhouse and Ashby, 99% on Lever, and 37% on SmartRecruiters. Every text-based figure above is therefore computed on the subset we can read, and that subset under-represents the large service employers. The title-based figures cover all 46,155.

224 employers is not the US labour market. It is the set of employers whose career pages FACTANKER currently observes, weighted toward listed technology companies. Read these numbers as what these 224 companies are doing, not as a national rate.

“AI in the title” is a proxy for “AI role” and it is imperfect in both directions: it counts a data-science position that predates the current wave, and it misses a backend engineering role that will spend its days on model infrastructure. We publish the term list below so the count can be checked and disagreed with.

Method

Source: employer career pages, observed directly through their applicant-tracking systems, as of 13 September 2026. A posting counts as an AI role if its title contains, as a whole word, any of: AI, ML, LLM, NLP, MLOps (case-sensitive, to exclude ordinary words in other languages), or, case-insensitively: machine learning, artificial intelligence, deep learning, large language model, computer vision, generative AI, neural network, data science, data scientist, prompt engineer. A posting mentions AI if the same terms appear anywhere in the description. An employer is counted as boilerplate if at least 90% of its postings mention AI and it has at least 20 postings with full text.

Every posting behind these counts has its own page in the registry, with the employer, the platform, the date we first observed it, and a link to the original ad. Start at the jobs registry.

The records behind this article