AI Academy/Module 7
Can You Tell If a Resume Was Written by AI, and Does It Matter?
AI detectors are unreliable and wrongly flag non native English writers, so do not base hiring decisions on them. Here is when an AI written resume matters, how to find hidden prompts aimed at AI screeners, and how to check claims instead.
- 7module
- 3 of 5lesson
- 10 minread
The short answer
Not reliably, and for most roles it does not matter much. AI detectors make too many mistakes to base a hiring decision on, and they wrongly flag non native English writers far more often than others. What matters is whether the resume is true and whether the person can do the job, so check the claims, not the author.
The one thing worth checking for is hidden text aimed at AI screeners. That is a deliberate attempt to game your process, and it is easy to find.
Why AI detectors do not work for hiring
AI detectors guess whether text was written by a machine by measuring how predictable it is. Plain, simple writing looks predictable, so it gets flagged.
A Stanford study tested seven detectors. They were nearly perfect on essays by US born students, but classified 61% of essays by non native English speakers as AI written. At least one detector flagged 97% of those essays. The researchers also showed the detectors are easy to fool by asking a chatbot to rewrite the text in fancier language.
OpenAI, the maker of ChatGPT, withdrew its own AI text classifier in 2023 because of its low accuracy.
In hiring, that is a serious problem. A tool that rejects people for writing in a second language can lead to discrimination on national origin, which US equal employment law protects. Module 4, Lesson 1 explains how bias gets into tools like this.
Does an AI written resume matter?
For most jobs, no. A resume has always been a marketing document, and many were written with help from friends, templates or paid writers. AI is the newest helper. Lesson 1 showed that many candidates use AI mainly to check and polish.
It matters in three cases:
| Case | Why it matters | What to do |
|---|---|---|
| The resume claims things that are not true | AI makes it easy to add skills the candidate does not have | Ask about each key claim in the first call |
| Writing is a core skill of the job | A polished resume tells you nothing about their own writing | Use a short writing task done in a setting you control |
| The resume hides instructions for AI | This is an attempt to manipulate your screening | Check for hidden text, see below |
What is hidden text in a resume?
Some candidates hide instructions in their resume for AI screening tools, such as white text on a white background saying the candidate is highly qualified. People cannot see it, but an AI tool reads it. This is called prompt injection.
It is not rare. Researchers from Duke University and others studied 200,000 resumes sent to the recruiting platform hireEZ and found that about 1% contained hidden instructions. The rate rose sevenfold between July 2024 and November 2025, helped by online videos and free templates. The researchers noted some candidates may have used templates without knowing the text was there.
How do you check for hidden text?
| Check | How |
|---|---|
| Select all | Open the file and press Ctrl+A (Cmd+A on a Mac). Hidden white text shows up highlighted |
| Change the color | In a Word file, set all text to black and look for new lines |
| Paste as plain text | Copy everything into a plain text editor and read the whole thing |
| Ask your vendor | Ask whether your screening tool detects and ignores hidden instructions |
If you find hidden instructions, treat it like any other honesty issue. You can ask the candidate about it. Some did it on purpose, some copied a template. Your policy should say how you handle it.
How do you check whether a resume is true?
Move your checking from the page to the conversation.
Here is a resume and the job description: [paste, with personal details removed]. List the five claims in this resume that matter most for the job. For each, write one follow up question that would show whether the candidate really has that experience.
Then ask those questions in the first call. People who did the work can tell you the details: the team size, the tool, what went wrong. People who borrowed the words usually cannot. Check what candidate data you can put into a chatbot before you paste anything.
Quick answers
Should I use an AI detector at all?
Not for decisions. The error rates are too high and the bias risk is real.
Can I tell AI writing by words like “delve”?
Not reliably. Plenty of people write that way, and candidates can easily edit those words out.
Should I ban AI written resumes?
You cannot enforce it, and it would punish honest candidates who told you. Ask for truthful applications instead.
Do AI screeners fall for hidden text?
Modern tools are harder to fool than early ones, but ask your vendor. The Duke researchers did not test whether the hidden prompts worked.
Try it on a live req this week
For three applicants on a live req, test one specific claim from each resume on the screening call instead of judging how the resume was written.
Explore in the directory
Module 7: When candidates use AI too
- How are candidates using AI to apply for jobs?
- What do you do when auto apply bots flood your job posts?
- Can you tell if a resume was written by AI, and does it matter?
- How should you handle candidates who use AI in interviews and assessments?
- Will AI replace recruiters, and which skills keep you valuable?