A keyword filter looks for terms such as 'Python', 'CSCS card' or 'account management' and ranks CVs by how many appear, sometimes weighted by frequency. It cannot tell whether a skill was used last month or ten years ago, or whether 'managed' means leading a team or sitting near one. Synonyms are often missed, so a strong candidate who writes 'client relationships' instead of 'account management' can score low.
The weaknesses grow at volume. Applicants learn to copy phrases from the advert, and AI writing tools make this easy, so keyword scores increasingly reflect wording rather than ability. Approaches that read a CV in context and score it against defined criteria, with the evidence shown, give recruiters something they can check. Keyword search still has a place in sourcing, where it is used to find profiles rather than rank them.