A recruiter who used to open 80 applications for a role now opens 400. Some report 300 or more as the routine average. The work did not get four times more valuable, and nobody gave them four times longer to do it.
This is the defining change in hiring right now, and it has a straightforward cause. Writing a tailored application used to cost an hour. It now costs about ninety seconds. When the price of applying collapses, people apply to everything, which is a completely rational response and also a disaster for everyone downstream.
The numbers are worth sitting with:
- Around 67% of HR leaders say reviewing AI generated applications has slowed their hiring process
- 84% of HR teams report heavier workloads as AI tailored applications pile up
- 65% of hiring managers say AI enhanced resumes have made it harder to verify whether a candidate actually has the skills claimed
- Gartner projects that by 2028, one in four candidate profiles worldwide will be entirely fake
- Recruiters now spend around 23 hours screening resumes for a single hire
Here is the part that matters for you. Polished, fluent, keyword aligned writing used to be a differentiator. It is now the baseline, produced by everyone, at zero cost, in seconds. Being well written is no longer evidence of anything.
What This Actually Changed
Not the filters. The filters were always there and they still work roughly the same way, which our guide on writing an ATS friendly resume covers.
What changed is the human reading afterwards, and what they are now reading for. When four hundred resumes all say "results driven professional with a proven track record of driving cross functional initiatives", that sentence stops carrying information. It becomes noise the eye skips.
So attention moves to whatever is left: specifics that could not have been generated without knowing your actual job.
The Things a Generator Cannot Fake
A language model writing about your work only knows what your job title generally involves. It does not know what happened at your job. That gap is the whole opportunity.
Numbers with awkward shapes. "Increased efficiency by 40%" reads as generated because it is round and unsourced. "Cut invoice processing from 9 days to 5 after we dropped the second approval step" reads as remembered, because the numbers are odd and the mechanism is named.
The specific obstacle. Generated bullets describe smooth successes. Real work has friction. Mentioning the constraint you worked around, the system that did not support what you needed, the team of two doing the work of five, is instantly more credible than another clean win.
Nouns nobody else would pick. The tool, the process, the regulation, the vendor, the failure mode. Concrete nouns are the fingerprints of someone who was there.
Scope stated plainly. How many people, how much budget, how many locations, over how long. Generators are vague about scale because they do not know yours.
Does That Mean Do Not Use AI?
No, and anyone telling you not to is fighting the tide. Used well it is a genuinely good editor: tightening a paragraph you already wrote, catching repetition, suggesting a stronger verb, spotting that your third bullet says the same thing as your first.
The failure mode is different. It is asking a model to generate your experience rather than sharpen it, which produces text that is fluent, plausible, and about a generic person doing your job title. The distinction is between using it to sharpen what you wrote and using it to invent what you did.
A workable division of labour: you supply every fact, number, and specific. The model helps with phrasing. The moment it is inventing content rather than rewording yours, stop, because the interview will find it.
The Verification Problem Is Now Yours Too
Because inflated claims are so cheap, honest claims get treated with more suspicion. That is unfair, and it is also the environment.
The practical response is to make your claims checkable. Name the employer, not just the sector. Give dates that line up. Link a portfolio, a repository, a published piece, a certification that can be looked up. Anything a recruiter can verify in ten seconds moves you out of the pile of things they have to take on faith.
And keep your resume consistent with your public profiles. Contradictions between the two used to look like carelessness. In an environment where a quarter of profiles may be fabricated, they look like something worse.
The Short Version
When everyone's application is well written, well written stops being a signal. Specific, checkable, and slightly awkwardly real is what gets read now.
That is good news if you actually did the work. It is only bad news if your plan was to sound like you did.