How to Spot an AI-Written Resume (and What to Do About It)

The core tells: what AI-generated resumes look like

AI resumes follow a template even when the candidate tries to customize them. The tells cluster into three categories: language, structure, and specificity.

First, the language. AI defaults to high-register, formal phrasing that sounds impressive but says nothing. You will see verbs like 'spearheaded,' 'optimized,' 'facilitated,' and 'leveraged' in every bullet point. Real resumes use some of this language, but they mix it with plain talk and vary the sentence structure. An AI resume maintains the same tone and complexity from the summary straight through to the skills section. If every sentence could appear in a corporate press release, that is a red flag.

Second, the structure. AI-generated resumes often use identical formatting across different jobs: three to four bullet points per role, each starting with an action verb, each roughly the same length. Real career progression is messy. One job might have two bullets because the role was short or narrow; another might have six because it was the candidate's biggest responsibility. Perfect uniformity suggests a template, not a career.

Third, and most important, specificity. AI has no work history, so it cannot name the CRM you used, the exact revenue increase you delivered, or the specific problem you solved. Instead, it generates placeholders: 'improved efficiency,' 'enhanced customer satisfaction,' 'drove measurable results.' A real candidate will tell you they cut onboarding time from two weeks to five days by creating video walkthroughs and a checklist in Notion, or that they increased demo-to-close rate from 18% to 31% by rewriting the sales deck and adding a live ROI calculator. The difference is not subtle.

Buzzword density and the absence of tools

AI loves buzzwords because they sound professional and pass keyword filters. A resume heavy on 'synergy,' 'stakeholder alignment,' 'cross-functional collaboration,' and 'strategic initiatives' but light on named tools, platforms, or methodologies is almost certainly AI-generated or AI-assisted without editing.

Real workers name their tools. A customer support candidate will mention Zendesk, Intercom, or Freshdesk. A marketer will list HubSpot, Google Analytics, or Mailchimp. A project manager will reference Asana, Jira, or Monday.com. These names anchor the resume in actual work. AI cannot invent them because it has no idea which tools the candidate used, so it substitutes generic phrases like 'utilized industry-leading platforms' or 'applied best-in-class methodologies.'

The same pattern shows up in achievements. An AI resume says 'increased sales by a significant margin' or 'reduced costs through process optimization.' A real resume says 'increased Q3 sales from $47k to $68k by cold-calling 30 leads per day and offering a limited-time bundle' or 'cut monthly software spend from $1,200 to $850 by consolidating three tools into one and negotiating an annual contract.' The second versions prove the candidate did the work because only someone who lived it can supply that level of detail.

Run this test: highlight every concrete noun (tool names, metrics, project names, client types) in the resume. If you can count them on one hand across a multi-year career, the resume is either AI-generated or so vague it tells you nothing useful either way.

Why you should not auto-reject AI resumes

Spotting AI is useful. Automatically rejecting every flagged resume is a mistake, for three reasons.

First, false positives. Non-native English speakers often write in the same formal, slightly stilted register that AI uses. They avoid contractions, stick to textbook grammar, and lean on professional-sounding phrases because that is how they learned English. A resume from a qualified immigrant candidate can look identical to a ChatGPT output even though every word is true. Auto-rejecting on AI detection alone means you lose those candidates.

Second, AI-assisted is not the same as AI-written. Many job seekers use AI to clean up grammar, reformat sections, or generate a first draft that they then edit heavily. If the final version includes real metrics, named tools, and specific outcomes, the candidate has done the work. They just used a tool to present it more clearly. Penalizing that is like penalizing someone for using spellcheck.

Third, discrimination risk. If your AI-detection filter disproportionately flags resumes from non-native speakers, older workers unfamiliar with modern resume norms, or candidates from less-polished educational backgrounds, you have built a filter that violates fair hiring principles even if that was never your intent. US employment law does not care whether your bias was intentional. It cares about disparate impact, and an auto-reject policy based on writing style creates exactly that.

Use AI detection as a flag, not a verdict. If a resume looks AI-generated, note it, then look at the substance. Are there specific metrics? Named tools? Outcomes you can verify? If yes, move the candidate forward and ask clarifying questions in the interview. If no, the resume was useless whether a human or a bot wrote it, so screen them out on lack of substance, not on detection alone.

How to verify substance in the interview

If you suspect a resume is AI-generated but it includes some real details, the interview is where you separate real experience from fiction. Ask questions that force the candidate to go one level deeper than the resume.

Instead of 'Tell me about a time you improved a process,' ask 'You mentioned cutting onboarding time by half. Walk me through exactly what the process looked like before, what you changed, and how you measured the result.' A candidate who did the work can answer this in detail: they will name the steps, the bottleneck, the specific fix, and the metric. A candidate who copy-pasted an AI-generated bullet will stumble, generalize, or pivot to a different example.

For tool-heavy roles, ask them to describe their workflow. 'You listed Salesforce on your resume. Walk me through how you used it in your last role. What reports did you run? What fields did you customize? What was the most annoying thing about the setup?' Real users have opinions and war stories. AI-resume candidates give you textbook definitions.

For metrics, ask for context. 'You said you increased conversion by 22%. What was the baseline? Over what time period? What else changed during that window that might have influenced the number?' A candidate who lived the work knows the context, the caveats, and the confounding factors. A candidate who invented the number or borrowed it from AI will give you a vague answer or defensively insist the number was accurate without being able to explain how they got it.

The goal is not to catch people in lies. The goal is to confirm that the resume reflects real experience. If the candidate can back up their claims with specifics, it does not matter whether they used AI to draft the initial version. If they cannot, you have saved yourself a bad hire.

Using tools to flag AI content (and their limits)

Several free tools now claim to detect AI-generated text, including dedicated resume scanners. These tools analyze phrasing patterns, sentence structure, and vocabulary to estimate the likelihood that a given resume was written by a language model. SiftFirst includes a free AI resume detector that flags likely AI content as one signal in its evidence-backed screening process. You upload the resume, and the tool highlights sections that match known AI patterns.

These tools are useful as a first pass, but they are not foolproof. AI detection models produce false positives (flagging human-written text as AI) and false negatives (missing AI-generated content that was lightly edited). The accuracy improves when the text is long and unedited, which is why they work better on cover letters than on resumes. A candidate who runs their AI draft through a paraphrasing tool or adds a few personal details can often evade detection entirely.

Treat detection tools the same way you treat keyword filters: they help you prioritize your time, but they do not make the hiring decision for you. If a tool flags a resume as likely AI-generated, read it yourself. Do you see specific metrics, named tools, and concrete outcomes? If yes, the candidate is worth a conversation regardless of what the detector says. If no, screen them out for vagueness, not for failing an automated test.

The best use of these tools is not to police AI use but to surface the resumes that need a closer look. In a pile of 200 applications, an AI detector helps you quickly separate the candidates who put in the effort from the ones who hit 'generate' and submit. That is a time-saver, not a substitute for judgment.

What to do when the whole pile looks AI-generated

If you are screening for a remote role, a high-volume position, or anything posted on a major job board, you may find that 60% or more of your applications look AI-generated. At that scale, flagging individual resumes does not solve the problem. You need a different approach.

First, tighten your job description. AI-generated resumes flood generic postings because candidates (or the bots submitting on their behalf) spray applications everywhere. If your posting is vague or uses only common keywords, you will attract spray-and-pray applicants. Add specific requirements: name the tools you use, describe a real scenario the candidate will face, and ask for a concrete work sample or a one-paragraph answer to a role-specific question in the application. AI can generate a resume, but it cannot fake a real answer to 'Describe the last time you had to choose between two competing customer requests and explain how you decided.'

Second, screen on evidence, not polish. A resume that says 'increased engagement by 40%' with no context tells you nothing. A resume that says 'ran weekly email campaigns in Mailchimp, A/B tested subject lines, and grew average open rate from 19% to 27% over four months' tells you the candidate did the work. Build your screening criteria around specificity, not around formal tone or keyword density. This is where tools like SiftFirst help: you define the criteria (specific tools, measurable outcomes, relevant experience), and the system scores each candidate on evidence, not on how impressive the phrasing sounds.

Third, accept that some AI use is inevitable and focus on what matters. If a candidate used AI to rewrite their bullets for clarity but the substance is real, you have not been deceived. You have received a readable resume. Your job is to verify the substance, not to audit the drafting process. Save your energy for the candidates who fabricate experience entirely, not the ones who used a tool to present their real work more clearly.

FAQ

Related

• AI Resume Detector (free tool) →
• How to screen customer support candidates →
• How to screen sales representatives →
• Screening guides by role →
• Try the free screening demo →

More guides

• How to Screen 200+ Resumes Fast Without Missing the Best Candidates →
• How to Hire Without an ATS: A Small Business Guide →
• How to Write Hiring Criteria That Actually Rank Candidates →
• The Resume Screening Checklist for Small Teams →