The degeneracy spiral
This is a feedback loop. Not a conspiracy - a systemic mechanism that feeds itself.
Step one: candidates find out that ATS rejects CVs that aren't built around the keywords from the listing. They start optimizing. By hand, then with tools, then with AI.
Step two: HR departments start receiving CVs that look more and more similar. Everyone has the same keywords, the same structure, the same level of "fit." The ATS can no longer distinguish one from another - it has to get more stringent to filter anything at all. Thresholds rise.
Step three: higher threshold = even better prompt needed to get through. ChatGPT generates CVs better than most candidates write. The candidate delegates the whole thing.
Step four: the company has its top 20. Everyone looks identical. Because everyone generated their CV using the same algorithm based on the same job listing. The recruiter instead of "top 20 best candidates" gets "top 20 best prompts."
Who benefits? ATS vendors selling increasingly expensive updates. CV optimization tool vendors. External recruiters who know the system and can game it.
Candidates and companies - both lose, because they never truly meet.
The experienced candidate paradox
The more experience you have, the harder it is to fit into a template.
Someone with fifteen years of organizational transformation behind them has projects that lasted years and don't fit under a single title. Roles that shifted from month to month. Achievements that are measurable - but by a measure the algorithm won't recognize, because they're described in the organization's language, not the job listing's language.
ATS sees: "no keyword match." Application rejected.
Meanwhile the system passes through a candidate with two years of experience who spent a week optimizing their CV for that one listing. CV is perfect. Candidate - may or may not be right - but they get noticed.
The paradox: the system most effectively rejects those whose CVs are too original, too complex, or too honest to fit in the template. That's not an accidental flaw - that's an architecture that favors form over substance.
The more valuable the candidate, the less standard their CV, the worse they score. That's why the most experienced people often bypass the ATS through networking - not because they have "connections," but because they know the system won't notice them.
If not the CV - then what?
Open question. And the good news is there's no single answer.
Portfolio and proof of work. Instead of a list of titles - concrete results. Links, case studies, work samples. Harder to fake with AI, harder for ATS to process. Works for roles where results are visible and describable. Limitation: not all roles leave "samples."
LinkedIn as a living document. An actively built profile - with recommendations, comments, visible activity. ATS won't score it, but the human on the other side will. Requires time and consistency. Produces results over months, not weeks.
Referrals and network. Classic. Always worked, always will. A candidate referred by someone trusted bypasses the ATS - goes straight to a conversation. Problem: it requires a network you can't build in a week. And you can't build it only when you're job-hunting.
Conversation before application. Contact first, CV second. Reversing the order. Not possible everywhere, but where it is - it changes the whole dynamic. The CV becomes a formality, not an initial filter. Requires initiative and comfort with "cold contact."
Team-based hiring. If the team decides who joins them - ATS loses its grip. A candidate the team met and wants gets an offer regardless of the score. Requires a process change on the company's side - and the will to do it. But when it works - the results show.
None of these is a complete answer. None replaces everything else. Together they form a strategy that doesn't depend on a single system that can be optimized - or gamed.
Let's come back to the sentence from the opening: the ATS didn't select the best candidate. It selected the best prompt.
That sentence is not a description of some future dystopia. It describes what's happening right now - in many companies, in many hiring processes, every day. The system doesn't evaluate the person. It evaluates how well the person (or the AI acting on their behalf) can fit into a template. Form beat substance. Algorithm beat conversation.
Whether we want it to stay that way - that's a question for people, not systems.
This is the first article in a series on recruiting. Next: on 152 days between application and response - and what that number says about company culture. After that: why I stopped reading CVs before interviews - and why I encourage other hiring managers to do the same.
For now, one question: does your HR department actually know who they're recruiting - a person or a prompt?