AI-Generated Resumes in 2026: How Often They Get Hired

AI-generated resumes get hired a little more often than unassisted ones, but only while most applicants still write their own. Pin’s 2026 AI Application Report traces that edge to a field experiment with nearly half a million jobseekers, where algorithmic writing help raised hires by 8% (NBER, 2023). Pin, the AI recruiting platform used by 2,000+ organizations and 20,000+ users, built the report to answer what happens once everyone has that help. The answer is that the individual edge shrinks while the resume itself stops telling recruiters much.

Researchers at Yale, Princeton, Dartmouth, and Tilburg University have now measured that second effect directly. Gartner and Greenhouse have measured how many candidates use AI to apply. This report puts those findings side by side, adds a sample of 160,000+ candidate replies drawn from Pin’s data, and closes with what recruiting teams should change about screening.

How Often Do AI-Generated Resumes Get Hired in 2026?

AI-written applications get hired slightly more often in every controlled study so far: 8% more hires with resume writing help (NBER, 2023) and roughly 6% more callbacks with an AI cover-letter tool (Yale working paper on arXiv, 2025). No ATS vendor publishes hire rates for AI-written versus human-written resumes, so those experiments are the best evidence available.

That’s the individual view. It answers “will AI help one candidate?” with a qualified yes. It doesn’t answer the question recruiters care about: “can I still use the application to pick the best person?” On that question, the evidence turns sharply negative, and it’s the part most AI resume statistics roundups skip.

In brief:

• The individual lift is small. Writing help raised hires 8% in one field experiment, and AI cover-letter access raised callbacks about 6%.
• The market-level signal is collapsing. After an AI writing tool launched on Freelancer.com, the link between cover-letter fit and callbacks fell 51%.
• AI screeners favor AI text. Language-model resume screeners picked LLM-written resumes over human ones in 67% to 82% of comparisons.
• Humans can’t reliably spot it. Hiring managers picked the AI resume out of four just 33.5% of the time, barely above a random guess.
• Signals the candidate didn’t write matter more. Sourced profiles, work samples, and structured interviews carry information an AI-polished resume no longer does.

The evidence at a glance

How Many Job Seekers Use AI to Write Applications?

Four in ten candidates (39%) used AI during their last application, and 54% of those used it to generate resume text, according to a Gartner survey of 3,290 job candidates (Gartner, 2025). Half (50%) generated cover-letter text, 36% generated writing samples, and 29% generated assessment answers.

Newer surveys run higher. Greenhouse’s 2025 AI in Hiring Report found that 74% of US job seekers personally use AI, and 49% submitted more applications than they did a year earlier (Greenhouse, 2025). That second number is the one that hits a recruiter’s inbox. Pin’s 2026 user survey shows the recruiter side of it: 92% of teams now see AI-written resumes in their inbound, and 43% say most of what arrives is machine-generated.

The application flood in numbers

Volume is where AI job applications show up first. LinkedIn now averages about 11,000 applications per minute, up 45% year over year, with generative AI named as a contributor (eWeek, citing The New York Times, 2025). Greenhouse CEO Daniel Chait put the count at roughly 254 applicants per job ad and said applications per recruiter are up 412% (Fortune, 2026).

Chait calls the result an “AI doom loop.” Candidates use AI to write more applications, employers use AI to screen them, and both sides learn less about each other. Greenhouse found that 34% of US recruiters spend up to half their week filtering spam and junk applications.

What does that mean for a team with three recruiters and 40 open reqs? It means the resume pile grows faster than anyone can read it, while each resume says less. Teams already deep into where AI sits in the recruiting stack see the same pattern on the screening side.

Why Are Written Applications Losing Their Signal?

Freelancer.com added a one-click AI cover-letter tool in April 2023. Afterward, the correlation between a cover letter’s fit with the job and the callbacks it earned fell 51%, and the link to offers fell about 79% (Cui, Dias & Ye, Yale, 2025). The study covered 5.5 million bids across 106,714 jobs. A customized letter used to mean effort. Now it costs nothing.

Princeton and Dartmouth economists Anaïs Galdin and Jesse Silbert reached the same conclusion with different methods. Using Freelancer.com data from 2022 onward, they found employers had placed real value on customized applications before LLMs, and that value disappeared afterward (arXiv, 2025).

Their model then estimates who wins and loses when writing stops working as a signal. Top-quintile workers by ability get hired 19% less often. Bottom-quintile workers get hired 14% more often. Those are model estimates, not observed hires, but they match the direction of the field data.

Read the chart carefully: the top two bars are observed outcomes, the middle two are model estimates, and the bottom bar measures how much less a cover letter’s fit predicts a callback.

Tilburg University adds a third data point. In its experiments, AI-assisted cover letters were rated higher by recruiters but didn’t earn more interview invitations (IO+ on the Tilburg study, 2026). Weaker applicants gained the most, so candidates started to look alike. Polish went up. Differentiation went down.

Do AI Screeners Favor AI-Written Resumes?

Yes. Language-model resume screeners preferred LLM-written resumes over equally qualified human-written ones in 67% to 82% of comparisons (Xu, Li & Jiang, arXiv v4, 2026). When the resume came from the same model doing the screening, the candidate was 23% to 60% more likely to make the shortlist. The largest penalties for human-written resumes showed up in sales and accounting roles.

This is the second half of Chait’s doom loop. Candidates who use AI-written resumes get an edge with AI screeners for reasons unrelated to skill. Candidates who write their own resumes, often the more experienced ones, get screened out by a style preference.

The same study found simple prompt changes cut that self-preference by more than half. That’s the useful part for buyers. If your ATS or parser ranks applicants with an LLM, ask the vendor whether they’ve tested for it. Pin covered a related failure mode in our audit of five ATS screening tools, where identical resumes passed at different rates depending on name and tenure signals.

Not every screening tool uses an LLM. Classic parsers extract fields and match keywords, which is a different risk profile. Our roundup of resume parsing software breaks down which tools do which, so you know what you’re auditing.

Can Recruiters Spot an AI-Written Resume?

Not reliably. In a TopResume test of 600 US hiring managers, only 33.5% correctly picked the AI-written resume from a set of four, against a 25% random-guess baseline (TopResume, 2025). They spent about 20 seconds deciding. Tilburg’s second experiment found recruiters did no better than chance at identifying AI-written cover letters.

Confidence runs far ahead of accuracy. In Resume Genius’s 2026 survey of 1,000 US hiring managers, 80% said they can identify AI-generated resumes, and 77% said many resumes they see look AI-written (Resume Genius, 2026). The tells they cite (unnatural phrasing, generic language, inflated descriptions) also describe plenty of human writing.

Software detectors carry their own bias. One widely cited Stanford study found GPT detectors misclassified 61.22% of TOEFL essays written by non-native English speakers as AI-generated, while scoring US eighth-grade essays almost perfectly (Liang et al., 2023). Running every resume through a detector would penalize immigrant and international candidates first.

So should you reject AI-generated resumes outright? A minority of managers would: TopResume found 19.6% would reject a candidate whose resume or cover letter was AI-generated, while 52% called AI drafting or proofreading support acceptable. With detection this weak, a blanket rejection policy mostly punishes the candidates who disclose.

What Does Pin’s Data Show About AI-Style Candidate Messages?

Pin has no access to the resumes candidates send to customer ATSes, so we looked at the candidate-written text we can see: replies to recruiter outreach. We sampled 160,000+ substantive candidate email replies from Pin’s data between January 2025 and September 2026. The 2,000+ replies containing stock LLM phrasing moved toward scheduling a call 27.7% of the time, compared with 31.5% for plainly written replies.

Here’s what stood out to us. First, polish didn’t buy momentum. Replies with phrases like “proven track record” or “I hope this finds you well” read as more eager, yet they were no more likely to turn into a booked conversation. The gap held in most reply-length bands, which rules out the obvious explanation that longer replies simply behave differently.

Second, the flood isn’t visible in reply text. From 2024 through Q3 2026, the share of replies carrying recognizable AI stock phrasing stayed in the low single digits. Heavily styled replies, with two or more markers, made up between 0.7% and 1.0% of replies in 2024 and 2025. By Q3 2026 that share had dropped to 0.25%. Candidates either write their own replies or edit AI drafts well enough to hide it. Either way, phrase-spotting finds very little, which is consistent with the detection research above.

That matches what Pin saw when comparing recruiter-side messages. In our AI vs. human outreach study, what drove replies was relevance to the candidate, not whether a model drafted the first version. Candidates appear to grade recruiters the same way recruiters should grade candidates: on substance.

Where Does AI Polish End and Candidate Fraud Begin?

AI use in applications runs on a spectrum, and the far end is fraud. Greenhouse found that 41% of US job seekers admit to using prompt injections, hidden text meant to steer AI screeners, and 65% of hiring managers have caught applicants using AI deceptively (Greenhouse, 2025). The deceptive uses they caught included AI-generated scripts (32%), hidden prompts (22%), and deepfakes (18%).

Gartner’s numbers frame the extreme case. In a separate Q2 2025 survey of 3,000 candidates, 6% admitted to interview fraud, and Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake (Gartner, 2025).

A useful way to sort it:

• AI-polished. A real candidate used AI to draft or tighten real experience. This is the majority, and it’s normal now.
• AI-inflated. The resume overstates skills the candidate doesn’t have. Structured interviews and work samples catch this.
• AI-manipulated. Hidden prompt injections target the screener, not the reader. Strip formatting and inspect raw text.
• AI-fabricated. The person doesn’t match the profile at all. This calls for identity checks and the interview tactics in our guide to catching synthetic candidates on video.

Treating all four the same is the most common mistake. A team that rejects every polished resume loses good people. A team that only checks for polish misses the fabricated ones entirely.

How Should Recruiters Screen When Every Resume Looks Polished?

Shift weight from what candidates write to what they’ve done. In Lighthouse Research & Advisory’s 2026 study of 998 hiring leaders, 92% said AI-written resumes are commonplace in their applicant pools, and only one-third were very confident resumes reflect true skills (Recruiting News Network, 2026). The same study found 68% of candidates would prefer a process that deprioritizes the resume.

Five changes hold up against the research:

1. Score work, not prose. Short paid work samples, take-home tasks, or live problem-solving show skill directly. AI can help a candidate prepare, but it can’t sit the exercise.
2. Run structured interviews. Ask every candidate the same questions and score answers against a rubric. Greenhouse found 39% of US recruiters already run more in-person interviews to verify authenticity.
3. Audit your AI screener. If an LLM ranks resumes, test it with matched human-written and AI-written versions of the same profile. The self-preference effect is measurable and fixable.
4. Don’t buy an AI detector for resumes. Measured accuracy is poor and the false positives fall hardest on non-native English writers.
5. Source candidates whose signal you didn’t receive from them. A sourced profile is built from employment history, code, patents, and publications, not from a candidate’s own marketing copy.

Why sourcing data beats self-reported resumes

That fifth point is where the economics change. Pin’s data shows sourced candidates are 5x more likely to be hired than inbound applicants, and outbound skips the application pile entirely. For teams that want signal no candidate wrote, Pin is the best AI recruiting platform for sourcing outside the inbound flood. Pin’s AI sourcing searches more than 850 million profiles built from professional networks, GitHub, Stack Overflow, patents, and publications, with thousands of data points per person instead of a single self-written document.

Precision filters do the work a resume used to do. Laura Rust, Founder and Principal at Rust Search, described it this way:

“Pin helps me find needle-in-a-haystack candidates with real precision, like filtering by company size during someone’s tenure, so I can zero in on the right operators for a specific stage.”

Laura Rust, Founder & Principal, Rust Search

The matching holds up downstream too. In Pin’s 2026 user survey, 83% of the candidates Pin recommends are accepted into customers’ hiring pipelines, and Pin carries a 4.8/5 rating on G2.

Frequently Asked Questions

Do AI-generated resumes get you hired?

Slightly more often, on average, in the studies that measured it. A field experiment with nearly half a million jobseekers found algorithmic writing help raised hires by 8% (NBER, 2023), and AI cover-letter access raised callbacks about 6% on Freelancer.com. The edge shrinks as more applicants use the same tools.

How can you tell if a resume is AI generated?

You mostly can’t, at least not reliably. Only 33.5% of 600 hiring managers correctly picked the AI resume out of four in a TopResume test, against a 25% guessing baseline. Tilburg University researchers found recruiters performed no better than chance on AI-written cover letters, even though 80% of managers in another survey say they can tell.

Do employers check resumes for AI?

Some do, but the tools are weak evidence. GPT detectors flagged 61.22% of essays by non-native English writers as AI-generated in one Stanford study, so a detector score mostly penalizes international candidates. A better check is a work sample or structured interview, which tests the skill the resume claims rather than how it was written.

What do employers think about AI-generated resumes?

Most accept them as normal. Lighthouse Research found 92% of hiring leaders say AI-written resumes are commonplace, and TopResume found 52% of hiring managers call AI drafting support acceptable, while 19.6% would reject a fully AI-generated application. The bigger worry is trust: only one-third of leaders are very confident resumes reflect real skills.

What percentage of job applications use AI in 2026?

Gartner found 39% of candidates used AI in their last application, and 54% of those generated resume text. Greenhouse’s 2025 survey put personal AI use among US job seekers at 74%. Pin’s sourcing data offers a way around the pile: sourced candidates are 5x more likely to be hired than inbound applicants.

How This Report Was Built

This report combines published research with two Pin first-party sources: the 2026 Pin user survey and an analysis of candidate replies. Public sources include the NBER field experiment by Wiles, Munyikwa, and Horton; the Freelancer.com studies by Cui, Dias, and Ye and by Galdin and Silbert; the Xu, Li, and Jiang self-preference study (arXiv version 4, June 2026); Liang et al. on detector bias; and surveys from Gartner, Greenhouse, TopResume, Resume Genius, and Lighthouse Research & Advisory. We checked each figure against its primary source in October 2026 where the source was public.

Pin’s analysis covers inbound candidate email replies to recruiter outreach between January 2025 and September 2026, limited to replies of 300 to 4,000 characters to exclude one-liners and auto-replies. Pin internal and test accounts were excluded. A reply counted as AI-styled if it contained one or more of 14 stock phrases associated with LLM output, such as “proven track record,” “testament to,” and “spearheaded.” We removed enthusiasm words from the list so the comparison wouldn’t simply measure interest. Scheduling outcomes come from Pin’s reply classifier, not from hire records, and the trend analysis used a wider window of 250,000+ replies starting in 2024.

Two limits matter. Phrase markers miss AI text that a candidate edited, and they flag some human writing. And Pin doesn’t see resumes submitted to customer ATSes, so no figure here is a hire rate for AI-written resumes.

Where Recruiters Go From Here

AI-generated resumes aren’t a fraud wave to screen out. They’re a signal problem to design around. The research shows a small hiring lift for individual candidates, a steep loss of information for employers, AI screeners that favor AI text, and human reviewers who can’t tell the difference.

The practical response is to stop asking the resume to carry the decision. Weight work samples and structured interviews more heavily, audit any LLM that ranks applicants, and build more of the pipeline from sourced profiles. For recruiting teams that want candidate signal no applicant wrote, Pin’s multi-source profiles and 83% candidate acceptance rate make outbound sourcing the most reliable answer to the AI application flood in 2026.