A client once told us about every stakeholder he managed. Product, marketing, sales, engineering, support, all of it, for two years. None of it was on his CV. In his head the title already said it, so why spell it out.
I told him to write it down anyway. The AI reading his profile does not know he does that just because of his title. Neither does a recruiter eighteen months into the job. Something falls out in the translation every time, and nobody tells you it happened.
That conversation happened before any of what I'm about to show you existed. I think about it every time a client assumes a recruiter, or the software sitting underneath a recruiter, will simply infer what they mean.
Because since November, five companies that build the software recruiters actually use have told the market, in their own words, exactly what changed.
We've pulled up Greenhouse and Lever live in coaching sessions to show clients what a recruiter sees on their own screen. Those two, plus Workday, Ashby and SmartRecruiters, are the five platforms running underneath a large share of tech job ads in Australia and New Zealand right now. In 2026, four of the five shipped the same idea within months of each other: let AI write more of what the recruiter reads, before the recruiter reads anything you actually wrote yourself.
Lever, in November, shipped an AI Interview Companion that drafts the interview questions a recruiter asks you, built from your CV. Ashby, at its May conference, announced Scheduling Agents that run your entire booking process with no human step, plus a notetaker that summarises what you said in the room automatically. SmartRecruiters, owned by SAP since last September, built an agent called Winston that returns a sub-score on your education, your skills and your experience before anyone opens your file, alongside a fraud-detection prototype that flags applications on behavioural and device signals. Workday's HiredScore grades you against a role and can resurface you for a job you never applied to, pulled straight out of a database you forgot you were in.
Greenhouse tells a more careful version of the same story. Its own AI wave landed in June, including a notetaker that maps interview notes straight onto the scorecard the hiring team reads, same category of feature as the other four. What's different is the sell. "We built these capabilities to strengthen the decisions people make, not to replace them," its chief product officer, Meredith Johnson, said. "The hiring decision is always yours." Its VP of product, Robby Perdue, put it more bluntly: most AI in hiring adds speed without adding clarity, and at Greenhouse, people still make the call.
That's a real split in a market of five, not a marketing line we're reading generously. Four companies are racing to remove a human step. One is building the same technology and choosing to frame it as decision support, not decision-making. Which platform sits under the job you're applying to changes how much of your first impression a person actually forms themselves, and almost nobody thinks to check which one it is, even though the ATS is usually sitting right there in the application URL.
None of this means a robot rejects you. That panic is a few years out of date, and I've written about where it actually comes from. What changed here is upstream of any decision. More of what a recruiter reads about you, before they form their own view, was written by something else. The questions you're asked. The summary of what you said in the room. The sub-score sitting next to your name. Whether you resurface for a role you didn't even apply to.
Call it the Ghostwriter. It only knows what you actually gave it: a transcript, a CV, an application form. It cannot infer the thing you assumed was obvious. It cannot hear your tone and decide you meant more than you said. And unlike a real ghostwriter, you never get to read the draft before it goes out.
Here's the useful part. If the Ghostwriter can only draw on what you explicitly gave it, explicit is worth more now than it was a year ago. Not implied. Not assumed. Not "obviously I would have."
Two places to fix this today.
In your CV, put your level, your scope and your most relevant work in the first five lines, in plain words a stranger would use, not the shorthand your last team used internally. A summarising model guesses at your seniority when the words aren't there, and it guesses wrong more often than a person would.
In your next interview, if there's an outcome, say the number out loud. Don't say "I grew sales." Say "I grew sales 40 percent in six months." A transcript-based summary drops whatever a generous human listener might have filled in for you. It can only write down what actually landed in the room.
You can check your own draft right now, for free. Paste your CV into any AI chatbot and ask it to describe your seniority and scope in one sentence. If what comes back is vaguer than what you meant, that gap is close to what a recruiter's assistant would also miss.
That free check tells you the gap exists. It can't tell you what to do about it for a specific role, because a generic chatbot has no idea what the recruiter on the other side is searching for. CV Canvas, inside Careersy AI, reads your CV against the exact job ad the way a recruiter does, requirement by requirement: what they scan for, how relevant you actually are, and how to make the match obvious before anything summarises it for you. That is what we built it for. See your CV the way a recruiter's software does.
You've spent months writing for an audience of one recruiter. Write for the one that actually reads first.
Not by itself, at any of the major ATS platforms. Greenhouse, Ashby and the others state plainly that a human still makes the final call. What changed is how much of what that human sees, before they decide, was written by AI: the interview questions, the interview summary, a fit sub-score. The decision is still theirs. The draft they're reading often isn't yours.
A piece of AI built into the recruiting software (the ATS) a company uses, that takes an action rather than just storing data. In 2026 that ranges from drafting interview questions (Lever) to running the entire interview-scheduling process with no human step (Ashby) to sub-scoring your CV against a role (SmartRecruiters' Winston).
Workday, Lever, Ashby and SmartRecruiters have all shipped or announced agentic AI features in 2026 that draft, summarise, score or resurface candidates automatically. Greenhouse is the outlier, publicly building its 2026 AI features around keeping every hiring decision with a person.
Paste it into any free AI chatbot and ask for a one-sentence summary of your seniority and scope. Compare that to what you actually meant. Wherever the two don't match, that's language you need to make explicit rather than implied.
No, and the difference matters. The old myth, that a robot auto-rejects CVs on keywords, is mostly false. Auto-rejection is almost always a screening question you answered on the form, not the CV itself. What's true and new in 2026 is different: AI increasingly writes the summary, the questions and the score a human reads before they judge you, even though a human still makes the actual call.
Everyone gives you job-search tactics. Nobody warns you what three months of silence does to your head. Three concrete moves, from someone who read the applications, to take back the part the market works hardest to steal: a bit of control.
5 min read
Deep Research that remembers your prep and forgets your stale facts. Careersy AI Lite, an entry point for everyone. And the week's smaller fixes. The Debrief points at the room, not the company.
5 min read
