What does the AI tax on entry level jobs actually mean?
The AI tax on entry level jobs is the extra burden junior candidates now pay to get hired. You are competing for fewer true training roles, employers expect you to work alongside AI on day one, and a generic CV gets dismissed faster than it did a few years ago. The tax is not just fewer openings. It is also higher proof requirements: clearer skills, better examples, and less tolerance for vagueness. That is why the entry level job squeeze feels harsher even when headlines talk about resilience or productivity gains. ([siepr.stanford.edu](https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial))
The numbers back that up. The New York Fed said recent college graduates started 2026 with unemployment around 5.7 percent and underemployment at 41.5 percent. In mid-2025, iCIMS found entry-level openings were drawing 36 applicants per opening on average. Stanford researchers also found that workers ages 22 to 25 in the most AI-exposed occupations saw a 13 percent relative decline in employment after the spread of generative AI. That is not a vibe shift. It is a measurable squeeze on the first rung. ([newyorkfed.org](https://www.newyorkfed.org/research/college-labor-market?os=app))
The nuance matters. Handshake’s 2026 student research says the evidence for direct AI displacement of early talent is still mixed, not uniform across every field. At the same time, the share of full-time job descriptions on Handshake that mention generative AI has risen nearly fivefold since 2023. My read is simple: the tax is real, but it falls hardest on generic candidates chasing generic roles. If your CV makes you look interchangeable, AI-heavy hiring systems and overloaded recruiters will treat you that way. ([joinhandshake.com](https://joinhandshake.com/research/economic-research/workforce-outlook-class-of-2026-in-the-ai-economy/))
Why are junior roles getting squeezed first?
Junior work has always included routine tasks: first drafts, basic research, data cleanup, QA passes, scheduling, note taking, status updates, and templated communication. Those tasks are exactly where modern AI is strongest. Anthropic’s January 2026 Economic Index says some occupations, including data entry keyers, show broad task exposure, and Stanford found losses were concentrated in roles where AI is more likely to automate work rather than augment it. That is junior roles automation in practice. Companies are not deleting every starter job. They are deleting the most teachable, repetitive pieces of those jobs. ([anthropic.com](https://www.anthropic.com/research/anthropic-economic-index-january-2026-report?_bhlid=76e855ebb03f5ec3fce386d27a4fe1063b11f59c))
That changes how gen z hiring works. Employers still need early-career talent, but they increasingly want a junior employee who can review AI output, fix errors, escalate edge cases, and communicate with humans. iCIMS reported that 96 percent of recruiters think entry-level workers will manage AI agents within two years. Handshake also found AI language appearing in job descriptions much more often than it did in 2023. So the market has not stopped hiring beginners. It has started expecting beginners to operate like AI-assisted professionals from week one. ([icims.com](https://www.icims.com/company/newsroom/juneinsights2025/))
Most resume advice on this is wrong. Learning a few prompts is not enough. Employers do not pay you because you can ask ChatGPT for a summary. They pay you because you can tell when the summary is wrong, incomplete, risky, or tone-deaf. In other words, your value is moving up the stack. Judgment, prioritization, customer context, and accountability now matter earlier in your career than they used to. If you pitch yourself as a cheaper pair of hands, AI will undercut you. If you pitch yourself as a reliable reviewer and problem solver, you stay in the game. ([siepr.stanford.edu](https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial))
How should your resume change when hiring is AI-first?
Start by dropping the fantasy that you need to beat the ATS with tricks. Modern hiring platforms are not just keyword counters. Greenhouse says its AI is embedded across job setup, application review, interviewing, and reporting, with features like keyword filtering, scorecard summaries, and resume anonymization. Workday has rolled out a Recruiting Agent inside its AI stack. That means your CV needs to be clear, literal, and easy to summarize. Fancy layouts, vague headlines, and clever jargon are liabilities now, not differentiators. ([greenhouse.com](https://www.greenhouse.com/ai-recruiting))
Your resume should show AI skills as a work method tied to outcomes, not as a badge. A 2026 hiring experiment with 1,725 recruiters found that AI skills increased interview invitation rates by roughly 8 to 15 percentage points across office assistance, software engineering, and graphic design scenarios. Handshake also shows employers are mentioning generative AI far more often in job descriptions than they were in 2023. So do not write ChatGPT, Claude, Gemini, or Copilot in a dead skills list and hope for magic. Show what you used, why you used it, and what improved. ([arxiv.org](https://arxiv.org/abs/2601.13286))
Here is what that looks like in practice. A junior marketer should write that they used Gemini to cluster customer questions, then turned that into a five-email onboarding sequence that lifted activation. A support analyst should say they used Claude to draft macros, then manually tested them against real tickets and reduced response time. A junior backend engineer should say they used ChatGPT to generate test cases, then fixed edge-case failures before release. The pattern is simple: tool, task, judgment, result. That sequence reads like work, not hype.
How should you use ChatGPT, Claude, and Gemini without sounding fake?
Use AI as a sparring partner, not a ghostwriter. Handshake found that most rising seniors who use generative AI treat it more like a brainstorming and self-teaching tool than a pure content machine. That is the right instinct. When a model writes your whole CV from scratch, it smooths out your real experience and replaces it with polished mush. You end up sounding like every other applicant who asked for a professional resume rewrite. Recruiters can feel that flattening immediately, even when they cannot explain why. ([joinhandshake.com](https://joinhandshake.com/research/economic-research/workforce-outlook-class-of-2026-in-the-ai-economy/))
A better pattern is brutally specific. Paste the job description and ask the model to extract the exact nouns, tools, and business outcomes the employer cares about. Then ask it to compare those against your current CV and identify missing evidence, not missing adjectives. Next, ask for three bullet rewrites that preserve your facts, numbers, dates, and scope. Then stop. Pick the version that sounds like you. If the model invents a metric, a certification, or a tool you never used, delete it instantly. Speed is useful. Fiction is fatal.
My preferred workflow is simple. First, write your own rough CV and cover letter. Second, use ChatGPT, Claude, or Gemini to pressure-test it against one target role. Third, run the draft through HRLens for an ATS and clarity check. Fourth, read the final version out loud. If a sentence sounds like nobody would say it in a meeting, cut it. The goal is not to hide AI use. The goal is to make AI invisible because your application still feels concrete, human, and verifiable.
What makes you stand out when everyone can generate a CV?
Proof beats polish. That is the clearest shift in this market. iCIMS found that 44 percent of Gen Z candidates would welcome job simulations to show what they can do, while only 30 percent think employers truly value their skills. Read that again. Candidates want a chance to demonstrate ability because they know a resume alone is not enough. So build a proof pack: one project link, one short case study, one GitHub repo, one dashboard screenshot, one writing sample, or one Loom walkthrough. In an AI-first market, evidence travels farther than adjectives. ([icims.com](https://www.icims.com/company/newsroom/juneinsights2025/))
The most AI-resistant early-career skills are not mystical. They are the skills that sit around the model: framing the problem, choosing the right input, checking quality, spotting exceptions, understanding users, and explaining tradeoffs. Stanford’s paper distinguishes between work AI automates and work AI augments. Anthropic’s research makes the same broader point from a task perspective. So if you want to stand out, do not just say you are AI-savvy. Show that you can supervise messy reality, not just generate neat text. ([siepr.stanford.edu](https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial))
Networking matters more, not less, when the top of the funnel is flooded. Do not send a cold note that says you are passionate and attaching your resume. Send a small signal of relevance. For a sales ops role, point out one broken handoff in the company’s funnel and how you would track it. For a customer success role, rewrite one weak help-center article. For a junior product analyst role, sketch the metric tree you would start with. A real observation from you is harder to ignore than another AI-polished application.
Which career moves reduce the entry level job squeeze?
Aim for roles where AI acts like a copilot, not a replacement. Good examples include implementation specialist, revenue operations analyst, technical support engineer, customer education associate, QA analyst, IT operations coordinator, solutions consultant, and healthcare operations roles with real process complexity. These jobs still include repetition, but they also require context, escalation judgment, and live interaction with systems or customers. That is exactly where entry-level candidates can still compound fast. You want exposure to work that teaches decisions, not just output. ([siepr.stanford.edu](https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial))
Be more selective than the internet tells you to be. Greenhouse’s own 2026 messaging says candidates are using AI to apply to more jobs than ever, and trust is eroding on both sides. iCIMS also reported strong application growth while openings stayed flat and hires lagged. Mass-applying into that noise is usually a losing trade for junior talent. A smaller list of well-matched roles, each with a tailored CV, a tight evidence pack, and one human touchpoint, will outperform a hundred Easy Apply clicks you barely remember sending. ([greenhouse.com](https://www.greenhouse.com/newsroom/greenhouse-launches-ai-principles-framework-setting-the-standard-for-responsible-hiring-in-the-ai-era?utm_source=openai))
If you feel the AI tax on entry level jobs right now, do one thing this week. Pick one role title, one industry, and one business problem you can help solve. Then rebuild your CV around that triangle. Not around your classes, not around your intentions, and not around every tool you have touched once. The market is punishing vague beginners. It still rewards useful beginners. Make it easy for an employer to see which one you are.