Stanford Data: Junior Hires in AI-Exposed Jobs Down 19%

By James Calloway

Employment of workers aged 22 to 25 in AI-exposed occupations stands 19% below where it would be had it kept pace with less-exposed peers, according to a Stanford Digital Economy Lab working paper revised Aug. 12, 2026. The gap comes from hiring freezes, not layoffs — and that distinction should reshape how CMOs write their 2027 marketing budgets.

The paper, by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, draws on anonymized ADP payroll records covering a balanced panel of firms with 3.5 to 5 million employees per month through June 2026. In raw terms, the two most-exposed quintiles of young workers fell about 11% since November 2022, while the three least-exposed quintiles grew about 10%. Every older age group grew over the same period — by as much as 11% for workers aged 35 to 40. In short: AI is thinning the junior ranks before they ever join.

What does the Stanford paper actually prove?

The authors call the facts "early, descriptive indicators" and say the work does not estimate a causal impact of AI. The headline gap shrinks from about -0.18 to -0.09 when they control for education, and the smaller figure is significant only at the 10% level. Earlier versions of the paper reported regression estimates of 13% and 16%, so those figures should not be lined up against the 19% headline. The paper also measures occupations broadly and reports no SEO or marketing figures.

The mechanism matters more than the number. Separations among young workers fell at least as much in the most-exposed occupations as in the least exposed — the opposite of what displacement would imply. Where AI substitutes for human tasks, junior hiring dries up; where AI complements workers, employment is flat or rising, and the adjustment shows up in employment, not pay.

What does that mean for the 2027 marketing budget?

The answer starts with a column by Greg Jarboe, published via Search Engine Journal, that reconciles the Stanford data with his earlier advice. In July, he argued CMOs should fund five functions instead of legacy channel buckets, citing Gartner data showing labor rising from a mean 21.9% to 24.5% of marketing budgets even though 43% of CMOs expected to cut labor spending. His read now: the extra labor share likely pays experienced people, while seats that would have gone to 22- to 25-year-olds sit empty.

"A budget that raises labor share while freezing entry-level hiring has not told the CFO what it is funding," Jarboe wrote, adding that this omission "will be the first thing a sharp finance chief finds." Gartner doesn't split marketing labor by seniority and Stanford doesn't break out marketing, so the inference is unproven — but it's testable inside any department.

His September column also needs a sharper test. He had argued that SEO and paid media managers become directors by proving AI pays for itself while keeping their teams intact. Stanford's hiring-not-layoffs finding undercuts "intact" as a metric: "'Intact' tells you who is still there. It says nothing about who was NOT hired." If the entry-level pipeline narrows, the pool for the next generation of managers narrows with it. Note the reconciliations: a Harvard Business School study cited in September tracked job postings through March 2025 — a demand signal — while Stanford measures realized employment through June 2026. They point the same direction, but Jarboe warns against adding the 13% and the 19% together.

Where does Kevin Indig's planning memo fit?

Kevin Indig's Growth Memo on planning 2027 starts from a fixed-cost premise: salaries, tools and infrastructure are fixed, so the real question is which capabilities to fund. Proven work — technical SEO, content refreshes, digital PR that earns citations — gets full funding. The memo adds a capped bucket for learning work with one owner and a decision date, and allocates roughly 5% of the marketing budget to community, sometimes 10%.

His three measurement layers — "exposure (AI citations, mentions, share of voice), behavior (branded search, direct visits), and business outcomes (pipeline)" — close a gap in the September scorecard, which tracked only hours saved, hours returned, outcomes and cost. If exposure rises while branded search stays flat, Indig says, you have a distribution problem.

The tasks Jarboe flagged in September as most exposed — recurring reports, bulk title tags, bid adjustments — are the ones juniors used to learn on. Community work, digital PR and citation tracking are harder to hand to a model. His position: a junior seat built around that work belongs in Indig's learning bucket, with a named owner, set hours and a decision date. That's a stronger CFO case than keeping a seat because it has always existed.

How to use this in your 2027 plan

Jarboe's playbook has three steps:

• Sort every junior task by what AI does to it. Track what 22- to 25-year-olds actually do each week for a month, and mark each task as substitutable or judgment-dependent.
• Write the entry-level role into the learning bucket. Give it a named owner, fixed weekly hours for community and citation tracking, a quarterly review and a decision date.
• Put a hiring line on the scorecard. Show the seats you chose not to fill and why, and report Indig's three layers monthly beside it.

"An intact team is the wrong approach," Jarboe concludes. "A team you are still training is the better measure." A budget approved this fall will help decide who is in the room when the next directors are chosen — and Stanford itself calls its data early.