The ATS is keyword theatre

Somewhere right now, a person who has run payroll migrations for a decade is being filtered out of a payroll role because their CV says 'compensation systems' and the posting says 'payroll operations'. No human decided this. A string comparison did.

The applicant tracking system was sold as a way to handle volume, and volume is real. But matching keywords is not reading, and everyone on both sides has adapted to the pretence. Applicants paste the job description into their CV in white text, or more respectably, let a model rewrite their history in the posting's vocabulary. The companies respond by adding more keywords. The theatre expands; the information content does not.

What the filter optimises for

A keyword filter selects for people who are good at describing work in the language the filter expects. That is a real skill. It is a writing skill. If the role is writing, fine. For every other role, the filter is measuring the wrong thing with perfect efficiency, at scale, with no appeals process.

The quiet scandal is that everyone senior enough to fix this got their own jobs through networks, not filters. The ATS is for other people's careers.

A different first pass

The first pass should look at work, not vocabulary. A tagged, published project either demonstrates workflow design or it does not, and a capability search over evidence returns people because of what they built, not because of which synonyms they guessed. That is slower to populate than a keyword index and we accept that, because a small amount of true signal beats any quantity of matched strings.