Deel published a piece last week arguing that AI-augmented workers now earn 62% more for the same job, and cited our count of how rarely UK employers name AI in a job advert at all. We only supplied half that story, so we went back to the data and tested the other half: inside the same occupation, do adverts that ask for AI advertise more money?
They do. The gap is bigger than we expected in one job family, small in another, and untestable in the other nineteen occupations we hold, which is the honest shape of it.
The pay gap, inside the same job
Comparing adverts for the same occupation, one group asking for AI ability and one not, on advertised salary:
Median advertised pay, AI adverts vs the rest
- Product manager
- £82,500 with AI, £56,000 without. 1.47x, 11 adverts against 28.
- Data scientist
- £85,000 with AI, £80,510 without. 1.06x, 50 adverts against 14.
Two occupations. That is all that survived the rule we apply to every comparison, which is that both sides need enough adverts to compare. Everywhere else the AI side was one or two adverts, or zero.
Loosen it to adverts that mention AI anywhere rather than ask for it as a skill, and a third occupation appears: product manager 1.25x, marketing manager 1.09x, software engineer 1.04x. Same direction, smaller gaps.
What this does not tell you is whether AI causes the extra money. A product manager advert asking for AI is likely to be a more senior role at a better funded company, and our only control is the job title. The honest claim is that these adverts advertise more, not that adding AI to your CV moves your salary by 47%.
AI is normal in a few jobs and invisible in the rest
Ten per cent of the adverts in this sample mention AI, and 6.4% ask for it as a skill. Set that next to our hand-read study of frontline work, where 1.8% mentioned AI and 0.39% asked for it, and the two look like they contradict each other. They do not. They are different jobs.
This sample is weighted towards technology and commercial roles. That one is care work, retail, admin, driving and the NHS. AI is a normal requirement in the first group and almost absent from the second.
| Pattern | Adverts | Share | Where it concentrates |
|---|---|---|---|
| Mentions AI at all | 181 | 10.0% | data scientist 94.8%, product manager 47.1%, software engineer 35.5% |
| Asks for AI as a skill | 116 | 6.4% | data scientist 68.8%, product manager 31.4%, software engineer 19.4% |
| Automation, without naming AI | 108 | 6.0% | data scientist 37.5%, software engineer 30.6%, marketing manager 15.3% |
| Names Copilot | 16 | 0.9% | data scientist 8.3%, software engineer 6.5% |
| Names ChatGPT or OpenAI | 10 | 0.6% | data scientist 8.3%, product manager 2.0% |
| Names any other model | 2 | 0.1% | product manager 2.0% |
| Polices AI use in your application | 0 | 0.0% | none |
1,804 fresh deduplicated UK adverts, 21 occupations, collected July 2026. Employer counts are the number of distinct employers behind the matching adverts.
Nobody will tell you which tool
Copilot is named in 0.9% of adverts. ChatGPT or OpenAI in 0.6%. Every other model put together, 0.1%.
So employers are asking for a category, not a product. That is genuinely useful if you are writing a CV, because it means there is no specific tool to list. What lands is a sentence about what you did with it and what changed as a result, which is the same rule that applies to every other skill on the page.
Not one advert banned AI-written applications
We searched all 3,015 adverts in the raw set for any rule about using AI to write your application. Zero. Not a ban, not a disclosure request, not a warning about AI detection.
Take that for what it is. It does not mean nobody minds, and public sector recruiters do sometimes publish acceptable-use wording. It means that on this board, in this month, the rule most candidates think they are breaking was not written down anywhere.
Method
1,804 UK job adverts from two Reed pulls in July 2026: unique records with advert text, posted within 30 days of collection, near-duplicates collapsed on employer plus title plus advert opening, nursery nurses removed from nursing sets, and occupations kept where at least 30 adverts exist. Pay uses the midpoint of the advertised band, with hourly and daily rates annualised at 37.5 hours a week and 52 weeks, which is a convention rather than a fact about any job.
The patterns live in a frozen dictionary file with a version number so every figure can be regenerated exactly. Two of them are matched case-sensitively, because a case-insensitive search for "AI" matches ordinary words and page furniture, and that alone can invent a finding.
Limits, and they matter here more than usual. Advertised pay is not paid pay. Occupation is a coarse control, and within product management the AI adverts are plausibly more senior. Only two occupations had enough adverts on both sides to test, so treat the 1.47x as one job family in one month rather than a market-wide premium. And this is one job board, weighted towards commercial roles, which is exactly why the frontline number is so different.
Our earlier count of AI across frontline occupations is at AI in UK job adverts, and the wider corpus is described in ten findings from 15,291 job adverts. Deel's piece, which prompted this one, is here.
Frequently asked questions
Do UK jobs asking for AI skills pay more?▾
In the two occupations where we could test it properly, yes, and the gap is not small. Product manager adverts asking for AI advertised a median of £82,500 against £56,000 for product manager adverts that did not, which is 1.47 times. Data scientist adverts asking for AI advertised £85,000 against £80,510, which is 1.06 times. Everywhere else there were too few adverts on one side of the comparison to say anything.
How many UK job adverts ask for AI skills?▾
It depends entirely which jobs you look at, which is the real finding. In this sample, weighted towards technology and commercial roles, 6.4% of adverts ask for AI as a skill and 10% mention it at all. In our earlier hand-read sample of frontline occupations, care, retail, admin, driving and NHS roles, it was 0.39% and 1.8%. AI is normal in the jobs a minority of people do and invisible in the jobs most people do.
Which AI tools do job adverts name?▾
Almost none of them. Copilot appears in 0.9% of adverts, ChatGPT or OpenAI in 0.6%, and every other model combined in 0.1%. Employers ask for "AI" as a category rather than for a tool, which means there is no specific product worth listing on your CV to satisfy them.
Do employers ban AI-written applications?▾
Not in this data. Across 3,015 commercial job adverts, not one set a rule about using AI in your application. That does not mean nobody minds. It means it is not being written into adverts on this board, so any rule you are following is one you assumed.
Does an AI advert attract more applicants?▾
Slightly, and not consistently. Across five occupations the median difference was about three and a half more applications, but software engineer went the other way, with 18.5 for AI adverts against 35 for the rest. We are not publishing that as a finding.