33.3%of open senior B2B technology sales and marketing roles require AI skills of the candidate
The same postings, counted other ways
- 40.6%mention AI inside the requirements
- 74.3%mention AI anywhere in the posting
- 17.1%of those in-requirement mentions describe the employer, not a requirement
- 16.1%disclose that the employer screens applications with AI
Most AI language sits outside the skills requirements
A keyword search would report the loosest of those figures. The gap to the headline splits in two, and the first part is much the larger:
- 33.7 ptsAI language sitting outside any requirement, in company descriptions, benefits and legal text
- 7.3 ptsrequirements-section mentions that still ask nothing of the candidate
So a naive count is inflated mainly by where the words sit. Within the requirements themselves most AI mentions are real: 82.9% of the 281 in-requirement mentions genuinely ask something of the candidate.
The same postings, counted three ways
How a figure like this is built decides what it means. All three readings below come from the same postings; they differ only in what is allowed to count.
Separately: employers disclosing they use AI on applications
16.1% (95% CI: 13.6–19.1%, n = 700) of the postings analyzed state that the employer uses AI or automated tools to review or screen applications. This is a different claim from the one above and points the other way: it is not demand for AI skills in the role, it is AI in the hiring process itself, and it usually appears in an equal-opportunity, privacy or how-to-apply block rather than in the requirements.
It matters here for one specific reason. A keyword search for AI across the whole advertisement picks these disclaimers up and counts them as evidence of demand, which they are not. Removing them changes the loosest reading materially:
Which roles were counted
Postings had to be open at the time of collection and no more than 30 days old, because the claim is about roles employers are hiring for now rather than demand accumulated over a year. Each one then had to clear three tests: the role owns a function or leads leaders at director level or above; its main accountability is sales, marketing or revenue-facing go-to-market; and the employer primarily sells technology to other businesses. Recruiting firms, staffing agencies posting for an undisclosed employer, consumer technology companies and non-technology employers were excluded.
Where the postings came from
Boards in the study design that could not be collected
| Board | Reason |
|---|---|
| Glassdoor | Not collected. Glassdoor's job search returned no results at any level of location detail, failing with an error from its own location lookup, so no postings could be retrieved from it. The collection library's Glassdoor support appears to be out of date against the current site. |
| ZipRecruiter | Not collected. ZipRecruiter returned no results for any search issued, so no postings could be retrieved from it. |
| Google Jobs | Not collected. Google's jobs search returned no results for any search issued, so no postings could be retrieved from it. |
What this does and does not measure
It measures what employers write down. A job advertisement is a marketing document, often not written by the hiring manager. It can name AI skills that are never tested at interview, and it can omit expectations that decide the hire. That gap is real and this study cannot see across it.
Source concentration. Although postings were collected from 2 boards, 87.6% of the analyzed postings came from LinkedIn alone (613 of 700). That is not the mix that was collected: the boards contributed far more evenly, and the imbalance arose because postings from the other source were rejected at a much higher rate by the business-to-business technology test. Read this as substantially a LinkedIn reading rather than an evenly weighted one.
Coverage is limited to the 2 boards named above. Roles advertised only on a company careers page, or filled through an executive search firm, are invisible here, and senior hiring runs through search firms more often than junior hiring does. Postings written in French are under-represented, so Canadian coverage is weaker than the raw counts suggest.
A director title also means different things at a 40-person company and a 40,000-person one. Judging function ownership rather than title alone reduces that effect without removing it, so the population mixes scope levels.
In 5.4% of analyzed postings the advertisement had no section headings the analysis could recognize. Those were tested against the whole text, a looser standard than the rest. Their rate was 36.8% against 33.1% where headings were clear.
Finally, this is one reading at one date. AI language in job advertisements has been moving quickly, so treat 33.3% as a measurement of collected on 2026-08-06 rather than a stable rate.
How to check this
Every posting counted as requiring AI skills carries a quotation from the advertisement, and each quotation was verified word for word against the source text before the posting was counted. A classification whose quotation did not verify was discarded rather than counted. The full methodology note records the search terms, the matched vocabulary, the section rules, the confidence threshold, the collection route, and the complete list of limitations.