Does ATS read synonyms or exact keywords only?

Ethan Reynolds

Career Strategist at JobHire.AI & Former Tech Recruiter

🗓 Updated: April 2026 ✓ Industry Expert ⏱ 6 min read

I spent 4 years on the recruiting side of the table before moving over to career strategy, and I want to talk about something nobody really says out loud. 120 resumes a week for 4 years. AI-written ones have a tell.

Quick background: 4 years tech recruiting, mostly mid-level engineering and product roles, probably looked at somewhere north of 20k resumes total. the last 18 months of that was peak ChatGPT-on-resume era and it got genuinely funny by the end. our team had running jokes about it

People ask me sometimes "did you actually notice when stuff was AI-generated" and yeah, pretty much immediately. not because we ran detectors, nobody at our company was paying for those. it was just pattern matching after the 400th time you've seen the same phrase

But before I get into the tells, I want to explain what we're actually looking for in the first place. Because the tells only matter once you understand the job a recruiter is doing when they open your file.

What we're actually doing when we read your resume

So the ATS part, the keyword scan thing, that's already done by the time I'm looking at your file. It's the easy part. You pass it by mirroring the language in the job posting and most people figure that out.

When an actual person opens your resume we're doing something different. We're basically trying to figure out if you're real. Three things we're looking for, in roughly this order:

1. 01 specificity. "Managed a team" tells me nothing. "Managed 6 people through a product launch, ticket volume dropped 30%" tells me you were there. Real experience has numbers and context attached to it.
2. 02 does the career actually make sense. Not "is it perfect" because nobody's career is perfect. But does the arc track? AI output reads polished sentence by sentence and then gets weirdly incoherent if you zoom out and look at the whole thing. You notice it after a couple minutes.
3. 03 a voice. Hardest one to describe but basically: does it sound like a human wrote it, or does it sound like a tool. After a few hundred resumes you just know. It's consistent and it's real.

" The ATS lets you in. The recruiter decides if you stay.

The actual tells

Here's the stuff we actually flag. In no particular order, just what comes to mind:

1. 01 em dashes. I know this one gets mentioned a lot but seriously, ChatGPT can't stop. One of our coordinators started actually counting them at one point. Five em dashes on the first page of a resume and we'd all just look at each other.
2. 02 "results-driven [thing] with a passion for [other thing]". Or any variation. My eyes just slide off that opening sentence at this point. I genuinely cannot remember the last time I read one of those summaries. It's like the word "the", registers as nothing.
3. 03 every bullet is the same shape. "Accomplished X by doing Y resulting in Z" copy-pasted across 8 jobs. Real careers don't read like that. Some of my best wins took 3 sentences to explain, some of them are barely a phrase. When the rhythm of every bullet is identical you know something's up.
4. 04 brand examples that don't make sense. We had a stretch where everyone applying for a marketing role would mention Duolingo or Liquid Death as examples of brands they admire. Nobody worked at either. They just got pulled out of whatever the model thinks "good marketing brand" looks like. When 3 people in the same week reference the same company you didn't ask about, it stops being coincidence.
5. 05 the "too clean" career. Hardest one to articulate but maybe the strongest signal. Real careers are messy. Someone took a sideways move for personal reasons they don't quite explain. There's a 7 month gap somewhere. A role title that doesn't quite fit what they were actually doing. When everything connects neatly and the language is uniformly elevated for ten years straight, my brain starts asking why.

" It's not that anything is wrong. It's that nothing is interesting.

What people online call "AI smell"

That last one (the "too clean" career thing) is what people online call "AI smell" and I think it's a great term for it. You finish the page and you couldn't describe the person to a coworker if you tried.

The reason it happens is that AI optimizes for what a resume is supposed to sound like on average. Not what's true about you specifically. So what comes out is this weirdly hollow version of professional. It comes from three places really:

Recruiters won't sit there and analyze any of this. They'll just feel like nobody's home and move on. That's the whole problem.

A few practical things if you're job hunting

You don't need to start over with your resume. You just need to actually be in it. Here's what I'd do:

1. 01 read your resume out loud before you send it. If you wouldn't say "spearheaded cross-functional initiatives" to a friend at a bar (and you wouldn't) then a recruiter is going to notice when they read it. It just sounds wrong out loud and our brains pick it up even when reading silently.
2. 02 every bullet needs a number. Not "improved response time", give me "from 4 hours to 45 mins". Even rough estimates are fine. "Around 30%" is still data. "Improved efficiency" is not.
3. 03 write your own summary section. This is the first thing we read and the part most likely to sound generated. Two or three sentences in your own words. Doesn't need to be polished. It just needs to actually be about you.
4. 04 Kill the em dashes.

That summary one is the biggest deal. Here's what I mean by the difference:

The second one isn't polished. It also sounds like an actual human typed it in 3 minutes based on real things they actually do. That's the entire point. You can work with rough content that's real. You can't fix something that's technically perfect and empty.

The bigger issue with how people use AI

Most people use AI as a ghostwriter. Paste the job posting, hit generate, send what comes back. That doesn't work and it's why your resume looks like the other 60 in the stack that week.

The candidates who actually got phone screens used AI completely differently. They wrote their experience out first in rough, ugly, first-draft form. Then used AI to clean up the language. Their facts. Their numbers. The AI just made it readable. That's the whole game.

The second approach takes more time. Most people won't bother. That's the opening.

That's mostly what I've been telling people who ask me about their resumes lately. Take what's useful, ignore the rest.