AI credentials improved interview chances for underrepresented women, but gaps remained

Anglia Ruskin University reported on 8 October that an AI qualification improved interview prospects for underrepresented women in an experiment involving London vacancies, but substantial recruitment gaps remained. Applications highlighting the credential received invitations at a rate of 14%, compared with 11% without it, according to the university’s account of research conducted in 2025.

The study, led by professor Nick Drydakis, is titled Artificial Intelligence-Related Digital Skills and Employment Outcomes for Underrepresented Women. Wiley’s journal listing dates its online publication in Industrial Relations: A Journal of Economy and Society to 29 September. The October announcement describes employers’ responses to fictional applicants, rather than the employment outcomes of people who completed training.

How AI credentials changed interview invitations

According to the university’s report, researchers submitted more than 3,000 applications for real private-sector vacancies across ten workplace fields in London. The roles did not require a college degree. That setting matters: the findings concern this part of the recruitment market, rather than every occupation or employers in other countries.

The fictional female candidates were highly comparable, with applications varying by race, age, sexual orientation or autism-spectrum disclosure, and by whether they highlighted an AI qualification. The credential described six months of AI-related professional development through a vocational training provider.

Adding the AI credential increased the latter rate to 14%. Using those rounded figures, that is a three-percentage-point increase, or approximately 27% relative to the 11% baseline. The pooled result does not establish that each of the four underrepresented groups experienced an identical improvement.

The 14% rate still sat eleven percentage points below the 25% majority-group benchmark without an AI credential. However, the study did not measure whether majority-group applicants would receive a similar boost from the qualification. It therefore cannot establish the remaining gap if both groups highlighted AI training.

What the experiment says about skills and pay

Because the candidates were fictional, the experiment tested employers’ responses to a credential on an application. It did not measure anyone’s learning, independently tested AI competence or workplace productivity. The available account also leaves unresolved whether employers responded to the AI content specifically or to a broader signal of additional training.

The university also describes evidence of ‘wage sorting’: underrepresented applicants were more likely to receive invitations for lower-paid vacancies among the jobs they sought. AI credentials improved access to better-paid vacancies, but gaps persisted. Those comparisons concern advertised job opportunities, not wages actually earned after someone was hired.

The full paper was unavailable for examination. The university’s account supplies the overall invitation rates, but exact subgroup estimates, confidence intervals and detailed statistical specifications could not be checked. The reported percentages should consequently be read as rounded summary findings attributed to the university.

Earlier research on older women’s hiring prospects

Separate research by David Neumark, Ian Burn and Patrick Button provides historical context for recruitment barriers. Their NBER working paper first appeared in October 2015, was revised in November 2017 and subsequently appeared in the Journal of Political Economy in 2019.

Using more than 40,000 applications, those researchers found robust evidence of hiring discrimination against older women, especially women approaching retirement age. Their abstract reports considerably less evidence of age discrimination against men after adjustments for potential methodological biases.

The earlier study examined how comparisons could be distorted by giving older and younger applicants similar experience. It also used richer job profiles for older workers, including moves into less demanding roles. These design questions illustrate why the construction of fictional applications matters when interpreting recruitment experiments.

That older research did not test AI qualifications. It establishes relevant precedent for studying hiring barriers through applications, rather than independently confirming the new study’s credential-related findings.

Drydakis calls for fair recruitment alongside training

“AI training can help open doors, but it should not be viewed as a substitute for fair recruitment practices,” said Drydakis, who directs Anglia Ruskin University’s Centre for Inclusive Societies and Economies.

He called for investment in digital skills to be accompanied by efforts to tackle discrimination and assess qualifications consistently across applicant groups. That is the researcher’s recommendation; the report does not announce an adopted employer policy or government programme.

Whether the additional invitations translated into job offers, sustained employment or higher earnings remains unanswered. The report gives no verified timetable for research following applicants into work.