Research update · 9 October 2026
An AI-skills certificate won more interviews, but left a large gap
A credential may help an application reach an interview. In a London correspondence experiment, fictitious applications from underrepresented women with a sector-specific AI-related certificate received interviews 14.2% of the time, versus 11.1% with a matched nondigital certificate. Majority-group women without the AI certificate received interviews 25.0% of the time. This is not a count of hires or proof a course will pay for itself.
What the study measured
Nick Drydakis sent three matched fictitious applications to each of 1,016 private-sector, full-time London vacancies from January to June 2025. None required a university degree. Four separate experiments varied signals for race, age, sexual orientation and autism-spectrum disclosure. For each vacancy, the applications represented a majority-group woman without an AI credential, an underrepresented woman with a nondigital certificate, and an underrepresented woman with a six-month AI-related digital-skills certificate.
The pooled 3.1-percentage-point invitation difference rose to about 4 points after adjustment. The application signal was experimentally assigned, so this study can identify the effect of that signal on early-stage responses in this setting. It cannot establish that training created the skills or raised a real applicant's chance by the same amount.
What it cannot answer
The study did not observe offers, completed hires or wages. Its posted-pay analysis compares the advertised pay of vacancies that invited applicants, not salary offers to them. The race-specific AI-certificate difference was not statistically significant even though the pooled difference was. The study did not send a majority-group applicant with the same AI certificate, so it cannot show whether the credential narrowed discrimination in a strict causal comparison.
The control certificate was nondigital. Without a general digital-skills certificate, the experiment cannot tell whether employers valued AI-specific competence or broader digital literacy. A worker deciding whether to pay for training needs evidence on course cost, real skills, offers and pay—not an interview result alone.