Candidates have found a way to 'trick' recruitment AI tools

A recruiter in Britain was recently reviewing applications for an engineering role when he noticed something odd at the foot of a résumé: a hidden line of text, addressed not to him but to a chatbot, instructing the AI to declare the applicant exceptionally well qualified. He only caught it because he had recoloured the document to read it properly. As reported by The New York Times, that moment captures a fast-spreading phenomenon: job hunters embedding secret prompts in their applications to trick the AI systems which stand between them and a human being.

It would be easy to file this under “desperate candidates behaving badly.” We think it points to something more important about fundamental flaws in modern hiring practices.

An arms race nobody is winning

The scale is striking. Roughly 90% of employers now use AI to filter or rank résumés, according to the World Economic Forum (Leonid remains firmly in the other 10%!). In response, candidates have begun hiding instructions in white text, in font metadata, even inside the file data of a headshot. The hiring platform Greenhouse, which processes around 300 million applications a year, estimated that 1% of the résumés it reviewed in the first half of 2025 contained such a trick. ManpowerGroup, the largest staffing firm in the United States, says it now detects hidden text in about 100,000 résumés a year — roughly 10% of everything it scans with AI.

The tactics spread through TikTok and Reddit, and they build on an older habit of stuffing résumés with invisible keywords. Every time the screening software adapts, the workarounds evolve. It is a classic cat-and-mouse game; and like most arms races, it produces a lot of effort and very little trust.

Candidates only try to fool the bots because they know a bot is all that stands in the way. When applicants believe no human will ever read their words, gaming the filter starts to feel somehow rational. One applicant put it plainly: “recruiters use AI to do their screening, so why shouldn't candidates use AI to get past it?”

An automated filter creates two failure modes at once. It waves through the people who have learned to manipulate it, and it silently discards genuinely strong candidates whose experience doesn't match a keyword pattern. The result is a shortlist optimised for gaming the system rather than for talent. No amount of smarter detection software fixes that underlying gap; it only escalates the contest.

Integrity is a judgment call, not a keyword

Recruiters themselves are divided on what a hidden prompt even means. Some see a resourceful, out-of-the-box thinker; others see plain dishonesty and reject the candidate on the spot. As one recruiter told the Times, “I want candidates who are presenting themselves honestly.” That tension - is this initiative or deceit? - is precisely the kind of question software cannot answer. It calls for context, conversation and judgment.

The Leonid view: keep people at the heart of the process

None of this is an argument against technology. AI can be a genuinely useful tool for handling volume, or creating efficiency - and it isn't going away. But it should never have any influence over a hiring decision. The moment automation becomes the gatekeeper rather than the assistant, you invite exactly the behaviour that is now filling recruiters' inboxes.

This is why our model at Leonid keeps people at the absolute heart of everything we do. With LeonidLive, we produce short video interviews with candidates - using questions set by the client - so hiring teams can see a person's communication, passion and authenticity long before an offer is on the table. You cannot inject a hidden prompt into a genuine conversation. You cannot fake, to an experienced human, the way someone thinks about a governance problem or explains their work. What an algorithm reduces to a ranking, a conversation reveals in full.

The résumé arms race is a warning worth heeding. In corporate governance and R&D tax recruitment, where the stakes of a senior hire are high and character matters as much as credentials, the answer to gamed screening isn't a better filter, it's a better process. One with human judgment firmly built in.

Leonid's vision is 'keeping people at the heart of everything we do' - and no more so is this relevant than in today's AI-driven world.