The story in 30 seconds

Companies Building AI Are Hiring More Junior Office Workers, Not Fewer.

Companies building AI advertise a larger share of entry-level office roles than their peers, not a smaller one: 13.3% vs. 7.8%. At the same time, the entry-level job-ad mix is tilting toward AI-exposed office work. These are two findings the headlines often miss.

Every figure here is a share of job ads, measured directly from publicly available web data, not headcounts or a survey. Like-for-like Jan to May 2025 vs 2026, on 2.6 million cleaned U.S. entry-level ads.

1 · Contrarian · This study
13.3% vs 7.8%

Firms posting AI-builder roles advertise a larger share of junior office positions than peers matched by industry and size. The relationship is associational, but it runs counter to the narrative that AI is replacing junior workers.

2 · Job-ad mix · This study
+1.5 pts

AI-exposed office roles account for a growing share of entry-level job ads. The increase is noted across all firms (+3.2 percentage points). Meanwhile, the share of in-person and manual roles declined.

3 · Payroll · Stanford / ADP
−13% to −16%

An independent Stanford study found declines in entry-level employment within AI-exposed occupations, reaching a similar conclusion through a different measurement approach.

"The common narrative is that AI is eliminating junior office jobs. Our data shows a more nuanced pattern: companies building AI advertise a larger share of entry-level office roles, not a smaller one. Whatever AI is doing to this market, "replacing the junior hire" is not the full story."

Karolis Didziulis
Product Director, Coresignal

Why now: the AI-and-jobs debate is loud but mostly reactionary. This is a direct read of 2.6 million real ads that can be used in addition to the Stanford University payroll study.

What it means: for new graduates, most first jobs are office and knowledge roles where AI tools are already in use; fewer are in-person or manual.

What the job ads show, measured directly

What 2.6 Million Job Ads Show About How Entry-Level Hiring Is Changing.

We categorized 2.6 million U.S. entry-level job ads by the type of work they involve. The mix of junior hiring is shifting: AI-exposed office roles make up a growing share of postings, while in-person and manual roles make up a shrinking share. The analysis measures the composition of active job ads, not employment levels.

Coresignal study · Published June 2026 · Core comparison: Jan–May 2025 vs Jan–May 2026
Free to republish with credit and a link to Coresignal

Key findings

What we found

In short, the composition of entry-level hiring is changing. Entry-level roles account for a smaller share of all hiring than they did a year ago, and the openings that remain are increasingly concentrated in AI-exposed office roles rather than in-person and manual jobs.

  • The common belief is that AI is replacing junior workers. Yet companies building AI advertise a larger share of entry-level office roles than similar firms (13.3% vs. 7.8%). The relationship is associational rather than causal, but the pattern is the opposite of what that narrative would suggest.
  • AI-exposed office roles are a growing share of entry-level ads. The increase remains visible among employers present in both years (+1.5 points) and across all firms (+3.2), so the result is unlikely to be driven by changes in the employers included in the analysis.
  • In-person and manual roles generally fell (−6 points, same employers). The larger market-wide drop (−9.5) is mostly due to a shift into generic job titles, not evidence of those roles leaving the market.
  • Looking at all seniority levels, entry-level roles are a smaller share of all advertised hiring than a year ago: 34.7% to 30.6%, a 4.1 point drop, in 39 out of 51 states.
  • The share of AI-exposed roles is highest in Washington DC, New Jersey, and New York, with many of the fastest-growing markets concentrated in the same technology and business hubs.
  • These figures show the share of job ads, not the number of people hired. "AI-exposed" refers to the type of work involved, not whether a job will be replaced by AI.
The short version

Inside entry-level hiring, the mix of job ads is shifting

There is a lot of debate about how AI is affecting entry-level jobs. To measure what is actually happening, we analyzed 2.6 million U.S. entry-level job ads, classified them by the type of work involved, and compared the same months one year apart.

Here is what the ads show: among entry-level postings, the roles where AI tools can already do part of the work are growing. Those are software, data, analyst, office admin, accounting, marketing, and recruiting. The roles making up a smaller share are the in-person and manual ones, like food service, retail, warehouse, trades, and driving.

In plain terms, the entry-level job-ad mix is tilting toward AI-exposed office and knowledge roles and away from in-person and manual roles. This describes the composition of advertised roles, not how many people are employed.

What surprised us

We expected software and tech titles to drive the AI-exposed rise. They contribute, but the highest AI-exposed shares sit in administrative, finance, and research roles (43 to 49% of their entry-level ads). And the contrarian pattern above, AI-building firms advertising more junior office roles rather than fewer, contradicts the popular belief.

Why it matters

What this means for the people reading

For new graduates

A larger share of entry-level openings are office and knowledge roles. Skills like data, writing, and using AI tools feature prominently in these ads.

For employers and schools

Demand is moving toward AI-ready roles faster in some states and job areas than others. The tables below show exactly where.

For the AI debate

Job-ad data and payroll data can point in different directions. This study adds one specific, measurable piece of the puzzle: how entry-level hiring is changing.

+1.5
Point rise in the AI-exposed share of entry-level ads, holding the same employers constant (the increase across all firms was +3.2 points).
-6.0
Point drop in the in-person and manual share of job ads, same employers (the raw industry-wide drop is −9.5, but much of it is a shift into generic titles, not jobs reducing).
2.6M
Cleaned U.S. entry-level job ads in the study; the core comparison uses the 1.25M posted in Jan–May.
Same months, one year apart

One market, two trends

AI-exposed jobs
▲ +1.5 points
2025
11%
2026
12.5%
In-person / manual roles
▼ −6.0 points
2025
43.1%
2026
37.1%

The chart shows employers that posted entry-level roles in both years, making it the most conservative view of the trend. Even when comparing the same employers, the share of AI-exposed roles increased while the share of in-person roles declined. The changes are larger when all firms are included (AI-exposed: 10.8%→14.0%; 45.0%→35.6% in-person), although part of that increase reflects changes in the mix of employers and industries. The comparison covers January to May 2025 and January to May 2026. A third category, consisting of generic roles and a small number of unclassified postings, is not shown, so the two lines do not add up to 100%.

Validation: is this just which ads we collected?

Our 2026 dataset contains more ads than our 2025 dataset, so a fair question is whether the shift reflects a real change or simply differences in the employers and industries represented in the data. To test this, we repeated the analysis using a fixed set of employers and controlled for industry mix. In both cases, the direction of the trend remained the same.

‍Same employers: among the companies that posted in both years, the AI-exposed share still rose (+1.5 points) and in-person still fell (−6.0). Same industry mix: after reweighting 2026 to 2025's industry composition, the AI-exposed share still rose (+1.1). The headline direction does not depend on new firms or sectors entering the data. About half of the larger all-firms move (+3.2 / −9.5) is compositional, which is why we lead with the conservative numbers above. The finding is not solely driven by changes in employer composition or industry mix.

What "AI-exposed" means here

An "AI-exposed" role is one where AI tools can already do part of the work. These are mostly office and knowledge roles. Concrete entry-level examples from the data:

‍Administrative: administrative assistant, receptionist, data-entry clerk, scheduler.  Finance: bookkeeper, accounts-payable/receivable clerk, junior accountant.  Legal: paralegal, legal assistant.  
‍
‍Research & analysis: data analyst, business analyst.  Tech: software developer, IT support, marketing coordinator."AI-exposed" describes the kind of work, not a prediction that the job will be replaced. What changed is the share of entry-level ads these roles represent: larger in 2026 than in 2025.

What the study does and does not say

It is about the mix of job ads, not the total number of jobs.

We measured the mix of entry-level job ads, not the total number of jobs or people hired. So this is about which kinds of roles make up a larger or smaller share of advertised entry-level work.

"AI-exposed" means a role could involve AI tools. It does not prove anyone was hired or let go because of AI.

Stated plainly: the entry-level job-ad mix is tilting toward office and knowledge roles and away from in-person ones. We are careful not to extend that to claims about total employment. See how this fits with payroll data below.

Putting it in context

How this fits with the Stanford payroll study

A widely cited Stanford study found that entry-level headcount in the most AI-exposed occupations fell roughly 13–16% for young workers. That finding and ours measure two different things.

Stanford / ADP: counts people actually on payroll, meaning how many young workers are employed in exposed roles, from monthly payroll records.

This study: counts and classifies active job ads, the changing mix of roles employers are advertising, in a like-for-like Jan–May window.

A drop in payroll alongside a shift in the job-ad mix is not a contradiction. It is what a labor market in transition looks like. Employers can advertise a larger share of AI-exposed junior roles even while headcount in some exposed occupations contracts, for example as teams hire differently or restructure existing roles. Our data speaks to what companies need; the payroll data speaks to what companies have. Read together, they are more informative than either alone.

The timeframes also differ: the payroll work leans on monthly snapshots, while ours is a continuous read of postings across 2025 and 2026, compared month-for-month. We present this study as a complement to the payroll evidence, not a rebuttal of it.

Zooming out

Entry-level is a smaller slice of all hiring than a year ago

Everything in our study is measured within entry-level. But when compared to every job ad at all seniority levels, a second pattern appears: entry-level roles are a shrinking share of all advertised hiring. In the like-for-like Jan–May window, they fell from 34.7% of ads in 2025 to 30.6% in 2026, a 4.1 point drop, and a broad one, with 39 out of 51 states moving in the same direction.

Entry-level as a share of all advertised hiring, like-for-like Jan–May:

▼ −4.1 points
2025
34.7%
2026
30.6%

Combine the two layers and the shift shows up across the whole market, not just inside entry-level:

Share of all advertised hiring (Jan to May)
AI-exposed entry-level roles:
▲ +0.6%
2025
3.7%
2026
4.3%
In-person and manual entry-level roles
▼ −4.7%
2025
15.6%
2026
10.9%

These two results are the within-entry shares multiplied by entry-level's share of all hiring, so read them as directional rather than exact.

The two readings agree: entry-level is losing ground as a share of all advertised demand, and the entry-level work that remains leans further toward AI-exposed roles. When read alongside the payroll picture above, two readings of the same ad data move together and a separate payroll study points in a compatible direction: a labor market in transition, not a single dramatic event. These are shares of job ads, not number of people hired.

The most surprising finding

Companies building AI advertise more junior office roles, not fewer

Companies that post AI-builder roles also post a larger share of entry-level office and analyst ads than similar firms matched on industry and size. This is an associational pattern, not a causal effect, but the AI industry, which is believed to be reducing the number of some jobs, is increasing its own headcount.

AI-building firms
13.3%
matched peers
7.8%

"The common narrative is that AI is eliminating junior office jobs. Our data shows a more nuanced pattern: companies building AI advertise a larger share of entry-level office roles, not a smaller one. Whatever AI is doing to this market, "replacing the junior hire" is not the full story."

Karolis Didziulis
Product Director, Coresignal
Where it is happening · by state

Highest and fastest rising in the big metros

Two things are worth separating: where the AI-exposed share is highest (the level), and where it is rising fastest (the change). The genuine leaders on level are Washington DC, New Jersey, New York, Connecticut, and California, all well above the national average. Most of those are also among the fastest risers, so the shift is concentrating in metro and tech-hub states. The few states moving the other way are small, rural ones with fewer ads.

One caution for reporters: a handful of states (Alabama, Ohio, North Carolina) are rising fast but from a low base, and still sit below the national average. They are catching up, not leading. Rank by the 2026 column, not the change column, before calling a state an "AI hiring hub." All 51 states cleared the volume floor, and we screened out cases where a single employer drove a number.

Highest AI-exposed share · 2026 (Jan to May), national average 13.3%
Washington DC
26.9%
New Jersey
18.1%
New York
17.3%
Connecticut
17.2%
California
16.5%
State ⇅ 2025 ⇅ 2026 ⇅ Change ⇅

Share of entry-level ads that are AI-exposed, same months one year apart (Jan to May). Sorted by the 2026 level. ⚠ marks low-base risers that are still below the 13.3% national average. Showing the 10 highest-share and the 10 lowest-share states.

Where it is happening · by job area

Engineering is moving toward code fastest. Customer service is moving off the phones

We group ads by the department employers tag them with, then measure the AI-exposed share within each department from the job titles. Two columns: AI-exposed change is the point change in that share (Jan–May 2025 to 2026); in-person change is the point change in the manual-role share over the same window.

The main point: in Engineering & Technical, the AI-exposed (software/data) share rose 17.6% → 22.5% (+4.9) as the in-person share fell −8.8, so engineering postings are tilting toward code. In Customer Service, the in-person/call-center share dropped −13.2 as those roles moved into generic customer-service titles we hold out.

Why some high-AI areas show a small minus: Research, Finance, and Administrative already run very high on AI-exposed share (43–49%). Their small declines are not a move to manual work: the in-person column barely moves. They happen because a slightly larger share of 2026 titles fall into our held-out generic/unclassified buckets (for Research, the unclassified share rose +3.3 points). It is a labeling drift at the margin, not a real shift toward in-person work.

State ⇅ 2025 ⇅ 2026 ⇅ Change ⇅

Share of entry-level ads that are AI-exposed, same months one year apart (Jan to May). Sorted by the 2026 level. ⚠ marks low-base risers that are still below the 13.3% national average. Showing the 10 highest-share and the 10 lowest-share states.

Job area ⇅ 2025 ⇅ 2026 ⇅ AI-exposed change ⇅ In-person change ⇅

Same months one year apart (Jan to May). Shares are within each job area. AI exposure is about the kind of work, not proof anyone was hired or let go because of AI.

How we did it

Our research methodology

We wanted this to be easy to check. Here is exactly what we did, what data we used, and what each source was for.

We started with real job ads

We used Coresignal's U.S. entry-level job postings dataset, built from publicly available, continuously updated web data. It spans 2021–2026 and cleans down to 2.6 million job ads (2,748,270 initial → 2,626,430 after removing off-topic and mis-tagged senior roles). Because collection ramped up over time, the core comparison uses only the matched Jan–May windows: 346,532 ads in 2025 and 905,658 in 2026. The full 2.6M is the cleaned pool the comparison is drawn from, not the comparison itself. Each ad includes the job title, company, company size, industry, location, and seniority.

We cleaned out the noise

Our data already offer deduplicated job ads (same company, same title, same city, same day cleaned), so one job posted many times counts once. We removed senior jobs that were wrongly tagged as entry-level. We removed off-topic ads that were mislabeled.

We sorted each job by its title

We read each job title and sorted the ad into one of four groups: AI-exposed (AI can already do part of the work), in-person/manual, generic/ambiguous (sales, customer-service and advisory titles we deliberately left out), or unclassified.

We measured shares, not raw counts

For each time period we report each group as a share of all ads in that period. We do this on purpose. Raw counts mostly track how many ads were collected, not real changes in hiring. Shares are the honest measure.

We compared the same months

We compare January to May 2025 with January to May 2026. Matching the months removes the effect of seasonal hiring, so we are comparing like with like.

We checked our labels against outside research

We compared our "AI-exposed" group to two independent studies. We used these only to check our labels, not to claim AI changed hiring.

We stress-tested the result

We re-ran the numbers several ways, including moving every non-software engineer into the AI-exposed group. The direction held every time: the AI-exposed share rose and the in-person/manual share fell. We also checked it is not an artifact of which firms we collected: it holds within a fixed set of employers present in both years (+1.5 points AI-exposed) and after reweighting 2026 to 2025's industry mix. About half of the larger market move is compositional, which is why we lead with the conservative, stable-frame numbers.

Data sources, and what each was used for

Coresignal U.S. entry-level job-postings dataset. The main data: real-time, publicly available job-ad records. These are the job ads we counted and sorted. For how this multi-source data is collected and de-duplicated, see Coresignal's guide to multi-source jobs data.

Limits to keep in mind: we collected publicly accessible job-ad data. While we de-duplicate and screen out cases where a single employer or reposter drove a number, we cannot detect company intent when posting the ad, so read these as patterns in advertised roles, not a perfect census of hiring. Small rural states and tiny job boards are noisier; data from 2023 and 2024 does not properly showcase AI impact, so we lead with the 2025 to 2026 comparison. Most shares in this study are within entry-level postings. The one exception is the estimate of entry-level hiring as a share of all advertised hiring. The exact level should be interpreted with caution, but the year-over-year decline remains clear. These figures describe advertised job openings, not how many people were actually employed. ‍

What would change our findings: if the second half of 2026 reversed the trend, or if the rise in AI-exposed roles disappeared when comparing the same employers and accounting for differences in industry mix. Neither has happened so far.

Common questions

Quick answers.

Use this study

Put the chart on your site in one click

Copy this code into your article. The chart shows the headline finding and links back to this study, so your readers can see the full method.

AI-exposed vs in-person/manual entry-level job ads, U.S. (Jan to May)
2025
AI 11.0%
2026
AI 12.5%
2025
In-person 43.1%
2026
In-person 37.1%

Source: Coresignal, study of 2.6M U.S. entry-level job ads

<div class="cs-hiring-chart-embed">
  <p class="cs-title">AI-exposed vs in-person/manual entry-level job ads, U.S. (Jan to May)</p>

  <div class="cs-chart-group">
    <div class="cs-row">
      <span class="cs-year">2025</span>
      <div class="cs-bar-track">
        <div class="cs-bar-fill is-blue" style="width: 25.5%;">
          <span class="cs-bar-label">AI 11.0%</span>
        </div>
      </div>
    </div>
    <div class="cs-row">
      <span class="cs-year">2026</span>
      <div class="cs-bar-track">
        <div class="cs-bar-fill is-blue" style="width: 29%;">
          <span class="cs-bar-label">AI 12.5%</span>
        </div>
      </div>
    </div>
  </div>

  <div class="cs-chart-group">
    <div class="cs-row">
      <span class="cs-year">2025</span>
      <div class="cs-bar-track">
        <div class="cs-bar-fill is-red" style="width: 100%;">
          <span class="cs-bar-label">In-person 43.1%</span>
        </div>
      </div>
    </div>
    <div class="cs-row">
      <span class="cs-year">2026</span>
      <div class="cs-bar-track">
        <div class="cs-bar-fill is-red" style="width: 86.1%;">
          <span class="cs-bar-label">In-person 37.1%</span>
        </div>
      </div>
    </div>
  </div>

  <p class="cs-source">Source: <a href="https://coresignal.com/ai-entry-level-hiring" target="_blank" rel="noopener">Coresignal</a>, study of 2.6M U.S. entry-level job ads</p>
</div>
Table of contents