A Third of Bot-Beaters Hid Resume Text. Most of It Was Copied Keywords

In Greenhouse's new candidate survey, 81% of US respondents were rejected without speaking to anyone. And 62% tried at least one tactic to get past the screening software. A third of those hid text in their resume for an AI reviewer.

If you're one of them, or you're thinking about it, the largest study of real resumes has bad news. Most hidden text is keywords you could have written where people can see them, and it shows up in exactly the places recruiters read.

Does white text on a resume beat the ATS?

Almost never. At ManpowerGroup, the largest staffing firm in the US, it does the opposite: candidates caught with white text are not moved forward.

The white text trick means typing keywords in a font color that matches the page, so a person can't see them but software can read them. Prompt injection is a variant where the hidden text is an instruction to an AI screener, such as telling it to rate you highly. Here's what the evidence says about whether either one works:

• ManpowerGroup finds hidden text in about 100,000 resumes a year, roughly 10% of the resumes it scans with AI, according to its data analytics lead Max Leaming. Those candidates don't advance.
• Tomas Chamorro-Premuzic of ManpowerGroup gave the most generous expert estimate I've found: hidden text "might contribute 10% or 15% of the variability." That's the optimistic case, and it assumes nobody catches you.
• An ACL 2026 study found injection can move rankings, but "as more applicants stuff their resumes with hidden instructions, the advantage erodes." The more people do it, the less it's worth.
• Farah Sharghi, a recruiter with more than 10,000 interviews behind her, put it simply: "It's not going to work."

How many job seekers actually hide text in their resume?

About 21% of US job seekers say they've hidden text in a resume. When researchers measure real resumes, only about 1% contain it. That roughly 20x gap tells you more about the trick than either number does on its own.

Greenhouse surveyed 2,832 job seekers, including 1,200 in the US. The 21% figure is my own calculation from its numbers: 62% tried a tactic and a third of them used hidden text, which comes to about 248 US respondents. The same math puts about 25% of all US respondents as claiming skills they don't have, since 40% of tactic users admitted to doing it.

Two caveats before you quote these numbers:

1. Greenhouse also estimates that about 1% of roughly 300 million resumes contain hidden text. Sources disagree on whether that covers the first half of 2025 or a full year, so treat that total as approximate.
2. Greenhouse's November 2025 report said 41% admitted to prompt injection. That survey asked the question differently, so the drop from 41% to 21% isn't a real decline. They measure different things.

Why are the survey and the measurements so far apart? I see two plausible reasons. Either many attempts are so crude that detectors don't count them, or plenty of people call ordinary keyword tailoring "hacking the ATS." Either way, most people who say they hid text aren't doing what they think they're doing.

What does hidden resume text actually say?

Mostly copied keywords, not clever AI commands. In the USENIX study of real resumes, more than 90% of injected text had no explicit instruction at all. It was skills lists and job requirements copied from the posting.

Tianlong Chen (UNC, formerly hireEZ) and Neil Gong (Duke) led the study. They found hidden injection in 1.19% of resumes in one dataset and 0.91% in the other, and the rate rose sevenfold between July 2024 and November 2025. Chen pointed to how the idea spreads: "There are TikTok and YouTube videos teaching people how to do it and free templates to generate hidden prompts."

The researchers deliberately did not test whether hidden text changed hiring outcomes. So the biggest real-resume study can't tell you whether the trick helps. What the evidence does show is a rising rate, an advantage that shrinks as more people use it, and detectors that are already running in production.

Ya'el Courtney, a Stanford postdoc, found hidden prompts in 2.25-point white text across several resumes. That's typical. It doesn't need to be clever, because it's written for software. The trouble is that software doesn't keep secrets.

Can ATS detect hidden keywords?

Yes, and it barely has to try. Color is a property of the pixels on the rendered page, not of the text string, so any parser that pulls text out of your file reads your white keywords exactly like your black ones.

That's how the trick works against you. The parser does what you hoped and pulls the text out. Then it shows that text to a person, without the formatting that hid it.

Where hidden text shows up

• The parsed candidate profile. Parsers drop formatting, so your hidden block shows up as plain text in the profile recruiters actually read.
• Keyword search snippets. If a recruiter searches for a skill and the match comes from your hidden block, the snippet shows them a raw list of the posting's requirements.
• Reformatted documents. Louis Taylor, a UK recruiter, found a hidden ChatGPT command that stayed invisible until he reformatted the document.
• Purpose-built detectors. hireEZ has put both of its detectors from the USENIX work into production. One uses vision-language models to compare what a person would see on the page with what the machine extracts.

That last one is an automated version of pasting your resume into Notepad, and employers are already using it.

Why hidden keywords backfire with recruiters

Recruiters find candidates by searching for keywords, and a search matches visible text just as well as hidden text. Hiding a keyword adds risk and gets you no extra reach.

Scale makes this matter. In Refolk's index of professional profiles there are 145,411 US recruiters and talent-acquisition people (Recruiter, Technical Recruiter, TA Specialist, TA Partner). The UK has 11,693 with the same titles, so the US pool is 12.4 times larger. That's a lot of people typing skills into search boxes.

Only 1,651 of those US recruiters, about 1.1%, list "Boolean Searching" as a skill. Listing a skill isn't the same as using it, but this suggests most recruiters aren't writing complex search strings to dig up buried terms. They type a skill name and scan the results.

A skill in a visible bullet matches that search. The same skill in hidden text matches too, but the snippet now shows the recruiter something that looks like manipulation. Hiding it didn't get you a single extra match. It only changed what the recruiter thinks of you once they find you.

Run the plain text resume test before an employer does

The plain text resume test takes a few minutes. Strip your resume down to raw text and read every word, because whatever shows up there is what the ATS sees.

1. Select all. Open the final PDF or DOCX, press Ctrl+A (Cmd+A on a Mac), and look at what gets highlighted. If blank space gets highlighted, there's text there you can't see.
2. Paste into Notepad. Copy everything into Notepad or any plain-text editor and read it top to bottom. Any word you can't find on the visible page is a problem.
3. Check the reading order. Columns, text boxes, and tables can scramble the order. If a job title ends up separated from its dates, a parser may pair them wrong.
4. Look in headers, footers, and behind images. Templates often hide text boxes in these spots.
5. Change every font color to black. Select all, set the color to black, and set the text to a readable size. Anything new that appears was hidden.
6. Re-run after every edit. Stray text tends to sneak back in while you're tailoring a resume to a posting.

If you used one of the free templates Chen describes from TikTok and YouTube tutorials, assume it has a hidden block until this test proves otherwise.

Sort every hidden keyword: make it visible or delete it

Every hidden keyword is either true or false. True ones become visible bullets with evidence behind them. False ones get deleted. Nothing belongs in between.

This is the uncomfortable part of the data. More than 90% of injections are copied skill lists, and Greenhouse found that 40% of people using tactics admitted to adding skills they lack. So the hidden block is often where the false claims live. Hiding them doesn't make them less false. It just lets you avoid looking at them.

The two-way sort

• True, and you have proof. Write a visible bullet that names the skill and shows how you used it: what you built, for whom, and with what result.
• True, but thin. Put it in the skills section with honest wording ("working knowledge," or a coursework or side-project line) and be ready to talk about it.
• False. Delete it. If the posting lists it as required, that tells you something about your fit. Hiding it won't fix that.

A visible line matches the same search the hidden word would have. The difference is that it holds up when a recruiter reads it, and it gives the interviewer something real to ask you about.

Doing this sort by hand for every application is slow, and that's why people reach for the hidden block in the first place. It's also the work Refolk does for you: it builds your resume from your own history, so every keyword it adds is tied to something you actually did, and anything it can't support stays out.

Where resume keywords from the job description belong

Keywords from the job description belong where a reader will see them: your headline, summary, skills section, and bullets. Use the posting's exact wording wherever it honestly describes your work.

This is legitimate tailoring, the honest version of what the 54% who used AI to work out what screening systems look for were attempting. If the posting says "stakeholder management" and your resume says "worked with partners," you lose a match you had every right to.

• Headline. Use the posting's title if it fairly describes what you do.
• Summary. Two or three lines that name the top must-haves in the posting's own words.
• Skills section. Use the exact terms, and spell out each acronym once, since recruiters search either form.
• Bullets. Put each priority keyword inside an achievement, not on its own.

The harder question is whether you match the job at all. Refolk gives you a fit score for each posting that shows which requirements you genuinely meet and which you'd be stretching on. That's the question the hidden block was dodging.

FAQ

Will white text on my resume get me rejected?

At some employers, yes. ManpowerGroup, the largest US staffing firm, doesn't move forward any candidate it catches with white text, and it finds hidden text in about 10% of the resumes it scans with AI. If the hidden text claims skills you don't have, you're also misrepresenting yourself, and that becomes obvious the moment an interviewer asks about it.

Does a tiny font count as hidden text?

Yes. Ya'el Courtney, a Stanford postdoc, found hidden prompts set in 2.25-point white text. hireEZ's production detector compares what a person would see on the page with what the machine extracts. Text set too small to read is as easy for it to catch as white text, text behind images, or text boxes off the page.

Is it okay to copy wording directly from the job description?

Yes, as long as it's true and visible. Matching the posting's exact terms in your skills section and bullets is how keyword search is supposed to work. Problems start only when the words are hidden, or when they describe experience you don't have.