Résumé bullets: humanize past the eyebrow test.

The résumé-specific tells

AI-polished bullets share a pattern: verb inflation ('spearheaded' for attended, 'orchestrated' for scheduled), adjective padding ('dynamic', 'results-driven', 'passionate'), and suspicious roundness ('improved efficiency by 40%' with no baseline, timeframe or mechanism). Individually plausible; six in a row reads as manufactured — and interviewers now probe exactly these bullets, where inflation collapses in seconds.

The rewrite formula

Plain verb + what + measured result + how. 'Spearheaded revenue optimization initiatives' → 'Rewrote the renewal email sequence; late renewals dropped from 31% to 19% in one quarter.' The plain verb earns trust, the mechanism makes the number believable, and the specificity gives the interviewer a real question to ask — one you can actually answer well. Do this for your top five bullets; leave routine duties plain.

The verb translation table

Test any bullet by asking: would my former manager laugh at this verb? If yes, translate down — credibility compounds across the whole page.

The cover letter: same disease, stronger medicine

Cover letters are where AI inflation is most fatal, because the reader is explicitly evaluating your writing. The humanized structure mirrors the cold-email recipe: one line proving you know this company specifically (a product decision, a post, a job-ad detail — not "your prestigious organization"), two experiences connected to their actual needs with numbers, and a closing sentence a person would say. 150–200 words. Every phrase from the blacklist that survives into a cover letter costs an interview somewhere.

Three ATS myths that cause bad rewrites

"Keywords must appear exactly, many times" — parsers match variants and context; stuffing "project management" five times reads as gaming to the human who opens the file next. "ATS rejects automatically" — in most mid-size companies the ATS ranks and filters views; a human still decides, which is why the eyebrow test matters more than the parser. "Formatting doesn't matter if keywords are right" — tables and multi-column layouts still scramble in older parsers; clean single-column layout protects the content you humanized. Optimize for the machine to parse and the human to believe — in that order, and never the reverse.

The summary section: six lines that decide the skim

Most résumés now open with an AI-written "professional summary" — and most read identically: "Results-driven professional with a proven track record of delivering innovative solutions…". A humanized summary is a claim ledger, not an adjective parade: who you are in one plain clause, your two most defensible numbers, and the thing you want to do next. Example: "Backend engineer, 6 years, mostly payments. Cut settlement failures 38% at Klarna-scale volume; led the 4-person team that rebuilt reconciliation. Looking for a staff role where reliability is the product." Twenty seconds of reading, three follow-up questions planted, zero words from the blacklist. If a line in your summary could appear on a stranger's résumé unchanged, it isn't summarizing you.

Skills sections and the credibility budget

Think of the whole document as spending from one credibility budget. A skills section listing 24 technologies spends it fast — nobody believes deep competence in 24 things, and interviewers enjoy proving it. List what you'd accept an hour of questions on, ordered by confidence, and cut the rest; absence is safer than exposure. The same budget logic settles the classic dilemmas: one suspicious metric taints the real ones next to it (source your numbers or soften them honestly — "roughly a third" survives scrutiny that "34.7%" invites); "familiar with" is an anti-signal (either you can use it or it goes); and the more senior the role, the more the plain-verb rule matters, because the reader has personally done the things you're describing and knows exactly how much orchestration "orchestrated" involves.