Before the shortlist, before the interview, before the test — the wording of a job advertisement has already done selection work.
A job advertisement is not a neutral description of a role. It is a signal, read by a prospective applicant who is simultaneously asking a simpler question than the employer intends: is this for someone like me? That assessment happens fast, it happens before the application form is opened, and it is shaped by things that have nothing to do with whether the applicant could do the job.
- Pre-1975British computing adverts could specify sex, age and marital status directly
- 1975Sex Discrimination Act made explicit exclusions in job advertising unlawful
- Late 1960s–70sreclassification of programming attached formal degree requirements to roles previously filled without them
The mechanisms are reasonably well understood. Requirements lists that are longer than the role demands raise the cost of applying unevenly. Research into application behaviour has consistently found that women are more likely than men to treat a long requirements list as a genuine threshold — applying only when they meet most criteria — while men apply at lower rates of match. The asymmetry is not a personality trait; it is a rational response to workplaces where women in technical roles have historically faced steeper penalties for being seen as unqualified. The advert did not create that pattern, but it activates it.
The explicit markers disappeared; the implicit ones took longer, and some have not gone.
Language compounds the effect. Adjectives clustered around aggression and individual dominance — ninja, rockstar, crushing it, killer instinct, dominate — are not just stylistic choices; they function as cultural cues that signal who the existing team imagines itself to be. Computational linguistics work on job-posting corpora has demonstrated that masculine-coded language is associated with lower female application rates even when the role itself is not inherently gendered. Technology job postings, particularly in software engineering and infrastructure, carry a heavier concentration of such language than comparable postings in finance or life sciences.
The framing of flexibility matters too, and in a specific way. An advert that mentions flexible working as a concession — we may consider part-time, subject to business needs — signals something different from one that treats varied working patterns as structural. The first implies that flexibility is an exception requiring negotiation; it warns anyone who needs it that they will be marked as needing it. Mid-career women, who account for the largest share of departures from technical roles, read this signal accurately, because many of them have already negotiated it once and know what followed.
- Application-rate asymmetry
- documented gap between men's and women's threshold for applying relative to stated requirements; reflects rational response to differential penalties, not personality
- Masculine-coded language
- measured via computational linguistics analysis of job-posting corpora; associated with reduced female application rates independent of role type
- Flexibility framing
- distinction between flexibility as structural feature vs. negotiated exception; signals downstream cost to applicant
The historical accumulation
None of this is new behaviour. British computing advertisements from the 1960s and 1970s named not just experience but marital status, age brackets and — explicitly in some cases — sex, until the Sex Discrimination Act 1975 made direct exclusions unlawful. The explicit markers disappeared; the implicit ones took longer, and some have not gone. Advert language as a field of analysis is partly archaeology: tracing how the filtering function migrated from overt statement to coded vocabulary to structural framing.
The reclassification of programming in the late 1960s and 1970s also reshaped what adverts asked for. As the occupation moved upward through pay grades, formal qualification requirements attached themselves to roles that had previously been filled by trained clerical workers, many of them women. An advert that requires a computer science degree for a role that a capable analyst-programmer without one could fill is doing the same filtering work as the older adverts — it just looks like a quality standard rather than a preference. The credential requirement and the exclusion are functionally identical in outcome even when they differ in intent.
The practical implication for anyone reading published workforce data — what the surveys count, pipeline figures, application ratios — is that none of those numbers captures the attrition that happened before the count started. Every woman who read an advert, judged it not for her, and closed the tab is absent from every subsequent dataset. The filter is silent, it leaves no record, and it runs continuously.