How Recruiters Spot AI-Written CVs in 2026: 7 Red Flags That Get You Auto-Rejected

There's a new sixth sense in recruitment: the ability to spot a ChatGPT CV in under 10 seconds. After reading hundreds of AI-assisted applications per week, recruiters aren't fooled anymore. And the consequences of getting flagged are worse than you think. You don't just get rejected for this role. You get mentally filed under “didn't write their own CV” for every future role at that company.

The Scale of the AI CV Problem

Before we get to the red flags, the numbers explain why recruiters have become so aggressive about this:

Recruiters aren't being precious. They're drowning. And when someone's drowning in templated applications, the fastest way to stay afloat is to reject anything that smells like a template.

The 7 Red Flags Recruiters Look For

Here are the specific patterns that get your CV flagged, based on recruiter interviews, hiring-platform data, and linguistic research from Stanford on AI-generated text.

Red Flag #1: The “Spearheaded-Leveraged” Verb Pile

What it looks like: Every bullet point on your CV starts with “Spearheaded,” “Leveraged,” “Drove,” or “Orchestrated.”

“Spearheaded” has been called the absolute favourite verb of every major LLM. “Leveraged” is a classic filler that masks a lack of specific technical detail. When a recruiter sees six bullet points in a row all starting with one of these, it's the equivalent of an author tell.

What to do instead: Start bullets with the specific action you took, not the corporate verb for it. “Rewrote the onboarding email sequence; cut drop-off by 22%.” “Built a Zapier workflow that replaced three manual reports.” Specifics beat synonyms.

Red Flag #2: The Em Dash Epidemic

What it looks like: Three, four, or more em dashes (—) on a single-page CV.

This one is almost comically consistent. Human writers use em dashes roughly once every 500 words. ChatGPT drops them every 50 to 80 words: a rate two to three times higher than typical human writing. Recruiters have learned to scan for the visual pattern before they even read the content.

What to do instead: Do a find-and-replace for — across your CV. If you have more than two, you're in AI territory. Break the thoughts into separate sentences.

Red Flag #3: The “Pivotal-Intricate-Realm” Vocabulary

What it looks like: Words that sound sophisticated but say nothing specific.

Stanford University research on AI-generated text identified four words that appear at dramatically higher rates in LLM output than in human writing: realm, intricate, showcasing, and pivotal. Add delve, robust, synergistic, dynamic, and comprehensive to the list and you have the AI sophistication palette.

The tell isn't any single word. People do say “strategic.” The tell is clustering. If three or more of these land in the same paragraph, recruiters mentally mark you as AI-written before they finish the page.

What to do instead: Replace the adjective with the evidence. Instead of “robust data pipeline,” say “data pipeline handling 40M events/day with 99.9% uptime.”

Red Flag #4: The Three-Bullet Symmetry

What it looks like: Every role has exactly three bullet points, each roughly the same length, each following the same Verb → Action → Metric pattern.

Real careers aren't symmetrical. Some roles you absolutely crushed and have six stories to tell. Others you cruised for two years. When every job in your CV has the same rhythmic, balanced, three-bullet structure, that's a template talking, not a career.

What to do instead: Vary bullet lengths. Have two bullets for a minor role. Five bullets for the job where you built something big. Let the CV breathe like a real career, not a LinkedIn template.

Red Flag #5: No Tool Names, No Niche Obstacles

What it looks like: Your CV describes what you did in generic terms, with no specific tools, vendors, platform versions, or weird real-world problems you hit.

This is the single most reliable AI tell. Large language models are brilliant at describing jobs in the abstract. They're terrible at naming the specific tool stack you actually used, the vendor you fought with, or the Tuesday-afternoon obstacle that derailed the launch.

Notice what the human version includes that AI won't: the specific platforms (HubSpot, Salesforce), the specific event (2024 API deprecation), the specific mechanism (Salesforce Flow), and the specific before/after (4 hours → 12 minutes). None of this sounds “impressive.” All of it sounds real.

What to do instead: For every bullet, ask: “What tool, vendor, platform, or specific circumstance would only someone who actually did this job know?” Add it.

Red Flag #6: Prompt Injection Residue and Placeholder Text

What it looks like: “[Your Name]”, “[CompanyName]”, or worse: literal instructions the candidate forgot to remove.

Greenhouse's 2026 report found that 22% of deception cases involved hidden prompt injections: candidates trying to manipulate ATS screeners by embedding instructions in white text. Recruiters now routinely copy-paste CV content into a text editor to strip formatting and catch this.

Prompt injection is now treated as fraud, not a joke. Greenhouse's Real Talent launch in partnership with CLEAR was built specifically to catch this behaviour. Some ATS platforms now flag candidates who've attempted injection for identity verification before any interview.

What to do instead: Before you send, paste your CV into a plain-text editor. Look at the raw text. If anything looks like it shouldn't be there, remove it. And never, ever use prompt injection. It's a fast-track to a career-wide blacklist.

Red Flag #7: Inflated Metrics with No Context

What it looks like: “Increased revenue by 47%.” “Improved efficiency by 62%.” “Boosted engagement by 3x.” All with no baseline, no timeframe, no scope.

AI loves metrics because it knows humans love metrics. But it doesn't know what metrics were actually available to you. So it invents plausible-sounding percentages. Recruiters in 2026 have seen so many suspiciously round, suspiciously large numbers that metrics without context now actively lose credibility.

The human version is less flattering on paper. “11% churn” sounds worse than “40% improvement.” But it's credible. The AI version is just a number floating in space, and numbers floating in space now read as lies.

What to do instead: Use the PAR method (Problem, Action, Result) and always include the baseline, the scope (team size, budget, customer segment), and the timeframe. If you can't remember the exact number, say so qualitatively (“roughly doubled”) rather than inventing a precise-sounding figure.

Why This Matters More in 2026 Than Ever

This isn't a fad. The AI detection arms race has structural forces behind it that aren't slowing down:

The rational candidate response isn't to stop using AI. It's to use AI the way a good editor uses spellcheck: invisibly, in service of your voice, not replacing it.

How to Use AI on Your CV Without Getting Flagged

AI is still a massive force multiplier for job seekers. The trick is knowing where to deploy it and where to keep it out.

The Hirelytica Advantage

We built Hirelytica specifically because the current system penalises good candidates for the sins of AI spammers.

Frequently Asked Questions