If AI Writes Every Resume, What Are Recruiters Actually Evaluating?

Open any resume writing group online right now and you will see the same advice repeated constantly: paste your work history into ChatGPT, ask it to rewrite your bullet points, ask it to add metrics, ask it to match the job description keywords. Candidates are doing this by the thousands, and the results look impressively similar. Strong action verbs. Clean formatting. Quantified achievements that sound polished no matter how vague the original input was. So here is the question every recruiter should be asking right now: if most resumes are written or heavily edited by AI, what exactly are we evaluating when we read one?

The resume was already an imperfect signal

Before AI entered the picture, resumes were never a perfect measure of ability. They measured writing skill, self promotion, and how well someone understood what a hiring manager wanted to hear. Someone could be excellent at their job and terrible at describing it on paper, and someone else could be mediocre at their job but great at making it sound impressive. AI has not created this problem. It has made it much easier to fake the second half.

That changes the math for recruiters. A well written resume used to be a weak but real signal that the candidate had strong communication skills, at least in writing. Now it mostly signals that the candidate knows how to write a good prompt, or paid someone else to run their resume through a tool. The gap between "sounds impressive" and "is actually capable" has never been wider.

What recruiters should stop trusting

Quantified achievements need a second look. "Increased revenue by 34 percent" sounds concrete, but AI tools are very good at generating plausible sounding numbers when asked to add metrics to a vague bullet point. That does not mean every number on a resume is fabricated. It means numbers alone are no longer proof of anything without a source or a follow up question to confirm them.

Tone and polish also need to be discounted. A resume that reads smoothly, uses varied vocabulary, and avoids awkward phrasing used to suggest a certain baseline of competence and effort. Now it mostly suggests the candidate has access to AI tools, which almost everyone does. Polish has become the floor, not a differentiator.

"Polish has become the floor, not a differentiator. The gap between 'sounds impressive' and 'is actually capable' has never been wider."

What recruiters should actually be evaluating

If the resume itself is less trustworthy as a standalone signal, the evaluation has to move somewhere else. A few things still hold up.

• Deep Specificity: A candidate who can describe the exact tools they used, the exact obstacles they ran into, and the exact decisions they made in a project is showing real experience. Vague AI polish tends to stay vague under questioning.
• Conversations Around Failure: In an actual conversation, someone who can talk honestly about a project that did not work and what they learned from it is showing self awareness that a resume cannot fake as easily.
• Live Problem Solving: Watching how someone thinks through a real scenario, in real time, tells a recruiter far more than any bullet point. This is one of the reasons structured interviews are becoming a bigger part of the hiring process instead of a formality after the resume screen.
• Pipeline Consistency: A resume claim that does not hold up when a recruiter asks a follow up question, or when the candidate contradicts something they said earlier in the interview, tells you more than the resume ever could.

Why keyword matching is getting worse, not better

There is a strange feedback loop happening right now between AI resume writers and applicant tracking systems. Candidates use AI to match their resume language more closely to a job posting. Applicant tracking systems reward that closer language match with a higher score. Recruiters trust that score, at least partly, when deciding who to move forward.

The end result is a system where the candidates who score highest are often the ones who used AI most effectively, not necessarily the ones best suited to the role. Recruiters who understand this loop are starting to treat a high match score as a much weaker signal than they used to, and some are starting to distrust it almost entirely.

The candidates who lose out in this shift

It is worth naming who gets hurt by a resume screening process built around AI-optimized language. Career changers, who describe transferable skills in terms that do not exactly match a job posting written for someone already in the field, tend to score lower even when their underlying abilities transfer well.

Candidates who are less comfortable with AI tools, whether due to access, language barriers, or simply not knowing the tricks, end up competing against resumes that have been heavily optimized, without the same advantage. And genuinely excellent candidates who write plainly and honestly, without inflating their achievements, can look less impressive next to resumes that have been polished by a tool built specifically to sound impressive.

What good AI-assisted resumes still cannot fake

It would be a mistake to treat every AI-polished resume as suspicious by default. Most candidates using these tools are still describing real experience, just with better phrasing. The line between legitimate assistance and fabrication usually shows up in specifics. AI tools are excellent at making vague input sound impressive, but they are much worse at inventing convincing detail from nothing.

This is a useful test recruiters can apply directly during screening. Pick one bullet point from the resume and ask the candidate to explain it in more depth. Someone who actually did the work described will have layers of detail ready. Someone who leaned entirely on AI to generate an achievement they barely remember will often struggle to go beyond what is already written on the page.

The role of AI-assisted evaluation

There is a certain irony in the fact that as AI makes resumes harder to trust, AI is also becoming part of the solution on the hiring side. We built APTI Careers around this exact shift. Instead of relying on a resume alone, our platform runs AI-conducted interviews with video proctoring, gaze tracking, and integrity verification, so hiring teams get a much clearer read on whether a candidate's real time answers match the story their resume tells.

This is not about catching people out or treating every candidate as a suspect. Most people are honest, and most resumes are a genuine reflection of someone's background, even with AI assistance in the writing. But when the volume of applications is high and the polish level is uniformly high too, hiring teams need a better way to separate signal from noise, and that has to happen somewhere beyond the page.

Building interview questions for an AI-assisted world

The old approach to interview prep, reading the resume the night before and asking about the highlights, does not hold up well anymore. If a bullet point was heavily shaped by AI, asking the candidate to simply restate it will not reveal much, because they can restate it convincingly even with limited direct involvement in the original work.

A better approach is to ask questions the resume cannot answer for the candidate in advance: What was the hardest part of that project? What would you do differently if you started it again today? Who disagreed with your approach and why? These questions require the candidate to think in the moment, and thinking in the moment is exactly the thing AI-polished resumes cannot fake on someone's behalf.

Resumes are not disappearing, but their job is changing

None of this means resumes stop mattering. They are still the entry point, the first filter, the thing that gets someone into the conversation. What is changing is how much weight a recruiter should put on the resume alone, and how early in the process a real conversation needs to happen to verify what is written there.

The recruiters who adapt fastest to this shift will be the ones who stop treating a polished resume as proof of anything and start treating it as an invitation to ask better questions.

Separate genuine experience from well-written prompts.

If you want to see how AI-conducted interviews can help your team separate genuine experience from well written prompts, we would be glad to walk you through what we have built at APTI Careers.