Junior hiring is the most-cited measurable effect of AI. I am tracking the reported numbers: posting indices from Indeed, payroll data from ADP, the NACE employer survey, layoff trackers. The direction is unambiguous across all of them.
-28%
Entry-level graduate roles vs 2022 peak
Cannot compare
-36%
US tech job postings vs pre-pandemic
Confirmed
148K
Tech redundancies, Jan to May 2026
Traceable, broad scope
Disclaimer: This is personal research, not professional advice. I am a technologist, not an analyst, economist, or forecaster. Nothing here constitutes financial, investment, career, or legal advice. The data comes from third-party sources I believe to be reliable, but I make no representations or warranties as to its accuracy or completeness. My views may be wrong. Past trends do not guarantee future outcomes. If you make decisions based on this, that is your call, not mine. Consult a qualified professional before acting on any of it.
About this report
Why I built it, and why it will get better
I am trying to track what the real measurable effects of AI are being reported. I trace every number to its primary source. Where the data is confirmed, I say so. Where it is directionally accurate but imprecisely sourced, I say that too. Where it cannot be verified, I say that as well. This tracker is the result.
I am not an analyst or forecaster. I trace every number to its primary source. Where a number is confirmed, I say so. Where it is directionally accurate but imprecisely sourced, I say that too. Where it cannot be verified, I say that as well. Where credible sources disagree, I present both positions. This means my dashboard will sometimes show fewer data points than a typical industry report. I think that is a feature, not a bug.
I will update the tracker every month. Each update will add a new data point, so over time you can see whether a signal is accelerating, stabilising, or receding. Historical data will never be overwritten. After three or four updates, the trends will become the most important part of the dashboard. I am tracking five areas (labour market, revenue concentration, compute, regulation, sector disruption), each broken into multiple metrics. This page covers the labour market signal, which has four metrics. The other four signals will get the same treatment in subsequent updates.
Methodology
Source selection: I search for primary datasets from government sources (BLS, FRED, FERC), academic research (Stanford Digital Economy Lab, Burning Glass Institute, Revelio Labs), platform operators (Indeed Hiring Lab, Handshake), and independent trackers (TrueUp, Layoffs.fyi, NACE). I exclude any source where the underlying dataset cannot be located or where the methodology is undisclosed.
Confidence ratings: VERY HIGH = published government dataset or peer-reviewed study with large sample. HIGH = primary dataset with transparent methodology. MEDIUM-HIGH = primary aggregator with broad scope. MEDIUM = secondary source or different methodology. MODERATE = contextual only, not directly comparable. LOW = qualitative or untraceable.
Limitations: Margins of error and confidence intervals are not included in the figures shown. The Stanford/ADP study reports regression coefficients with significance levels; the visualisations here show the central estimate only. Time-series charts use point values without error bands. Where data points cannot be traced to a single publication, I note this explicitly. My goal is to present what the data shows, not to conduct original statistical analysis.
Date of access: All sources were last accessed in June 2026. Source URLs were verified live during research. Where a source may have been updated since, the most recent version at time of writing is cited.
Built with Odokai. The initial research synthesis ran in a single session, with each source then individually verified. The brand styling came from a reusable Skill. The data, charts, and HTML live in a persistent cloud workspace, ready for next month's update with no setup. The platform makes the process repeatable: each cycle is a matter of re-running the Skill against fresh data, not starting from scratch.
Metric 1 of 4
Entry-level graduate roles vs 2022 peak
The original claim: entry-level software engineering postings are down 28% from 2022 peaks. I traced this to FinalRound AI, a commercial interview-prep company, not a primary data source. But multiple rigorous primary sources confirm declines in the 20 to 45% range.
Verdict: Imprecisely sourced, cannot compare to verified range
The 28% figure traces to a commercial blog aggregator, not a primary dataset. However, the underlying trend is well-supported.
Stanford/ADP payroll data (50M workers, regression-controlled) shows a 20% employment decline for software developers aged 22 to 25. Indeed/FRED shows software development postings down 33% from Feb 2020 and 45% from the mid-2022 peak. Burning Glass Institute finds the decline is concentrated in occupations with high AI-exposure scores, consistent with, but not proof of, a causal AI effect. The most defensible range is 20% to 45%, depending on metric and time period.
Note: The Stanford/ADP figure comes from a working paper, not a peer-reviewed publication. Methodology is rigorous, but the paper has not yet undergone formal peer review.
Source
Figure
Scope
Confidence
Primary?
Indeed / FRED
-33%
Software development postings vs Feb 2020
HIGH
YES
Indeed / FRED
-45%
Software engineering postings from mid-2022 peak
HIGH
YES
Stanford / ADP
-20%
Employment of software developers aged 22 to 25 from late 2022 peak
HIGH
YES
Revelio Labs / Bloomberg
-35%+
Entry-level job openings (all sectors) vs Jan 2023
HIGH
YES
Burning Glass Institute
Structural
Decline concentrated in AI-exposed occupations
HIGH
YES
Handshake
#1 to #9
SE fell from most-posted to 9th on platform
HIGH
YES
Handshake
-30%
Tech internship postings since 2023
HIGH
YES
Software development job posting index (Indeed / FRED)
Indexed to Feb 2020 = 100. The mid-2022 peak reached 148. As of mid-2025, the index stands at approximately 67.
Overall index
The real story is the range, not a single number. Verified primary sources show declines of 20% to 45% depending on metric (postings vs employment), baseline (Feb 2020 vs mid-2022 peak), and population (all tech vs software specifically). The 28% figure cannot be directly compared to the verified range because its methodology is unknown.
Structural or cyclical? The evidence is mixed.
Structural (AI-driven)
Stanford/ADP: a 13% employment decline for young workers in high-AI-exposure roles persists after controlling for firm and industry shocks. Burning Glass: the decline is concentrated in occupations with high AI-exposure scores; low-exposure entry-level roles actually grew. Revelio: a 10pp increase in AI exposure is associated with an 11% decrease in entry-level demand, independent of macroeconomic conditions.
Stanford Digital Economy Lab, Burning Glass Institute, Revelio Labs
Primarily cyclical
Indeed Hiring Lab: junior postings are declining roughly in line with overall postings, not disproportionately. Only about a third of sectors show a decrease in the share of junior roles. The absolute decline mirrors the overall posting decline. BLS: software developer employment is still projected to grow 15% through 2033.
Indeed Hiring Lab, BLS Occupational Outlook
My assessment: The decline is partially structural. AI exposure has a statistically significant independent effect (Stanford, Burning Glass, Revelio all find this), but the overall macroeconomic environment also plays a role. The most defensible position: the decline would exist without AI, but AI makes it significantly worse, especially in tech. One important caveat: the junior-senior divergence is also compatible with a conventional hiring-freeze explanation (employers hold existing staff and stop hiring entry-level). The data cannot by itself distinguish AI substitution from a standard downturn. The experience-requirement shift in tech is a structural signal consistent with AI, though other factors may contribute.
Metric 2 of 4
US job postings vs pre-pandemic baseline
The original claim: US tech job postings are down 36% from February 2020 levels. This traces directly to Indeed Hiring Lab data. It is confirmed, but the real story is the divergence between junior and senior roles.
Verdict: Confirmed with nuance
The -36% figure is directly traceable to Indeed Hiring Lab. But the key finding is the junior-senior divergence.
Junior tech titles are down 34% from Feb 2020, while senior tech titles are down only 19%. This 15-percentage-point gap appears in tech among the sectors Indeed analysed. In other occupations they tracked, experience requirements actually eased. The experience-requirement shift in tech is a structural signal consistent with AI reshaping the entry point, though other factors (budget constraints, reduced training capacity) may also contribute.
Source
Figure
Scope
Confidence
Indeed Hiring Lab
-36%
All tech postings vs Feb 2020
HIGH
Indeed Hiring Lab
-34% junior
Junior tech titles vs Feb 2020
HIGH
Indeed Hiring Lab
-19% senior
Senior tech titles vs Feb 2020
HIGH
Indeed / FRED
-33%
Software dev postings vs Feb 2020
VERY HIGH
Indeed Hiring Lab
37% to 42%
Share of tech postings requiring 5+ years experience
HIGH
Tech job posting index: junior vs senior titles (Indeed)
Indexed to Feb 2020 = 100. The gap between junior and senior has widened from 0 to approximately 15 percentage points, a divergence unique to the tech sector.
Junior titles Senior titles
The 15-percentage-point gap is a structural signal worth watching. In other occupations Indeed tracks, experience requirements eased after the pandemic. In tech, they tightened. This divergence coincides with the period of rapid AI adoption in the sector. A causal link is plausible but not established by this data alone. The -36% figure is correct, but the real story is that senior roles recovered while junior roles did not.
Metric 3 of 4
Technology sector redundancies, Jan to May 2026
The original claim: 148,092 tech workers displaced across 354 events since January 2026, a daily rate of 981, running 46% above the 2025 average. This traces to TrueUp, a layoff tracker. The figure is real but the definition of "tech" is broad.
Verdict: Traceable to primary source, but scope is broad
The 148K figure is directly from TrueUp's tracker. However, TrueUp's definition of "tech" includes companies that are not primarily technology firms but have significant tech workforces.
The 46% increase over the 2025 daily rate is calculated correctly (981 per day in 2026 YTD vs 674 per day in 2025). At this pace, 2026 is on track for approximately 355,000 tech-sector redundancies for the full year, which would exceed both 2023 and 2025 peaks. Note: tech layoffs are historically front-loaded in Q1 (January in particular carries the most layoff announcements due to fiscal year planning). Annualising from Jan-May may overstate the full-year total. A seasonally-adjusted projection would likely be lower. The direction is clear even if the exact scope and total are debatable.
2026 is annualised from the current rate of 981 per day. At this pace, it would exceed both 2023 and 2025.
Annual redundancies
The direction is unambiguous, even if the exact count is debated. Two separate trackers (TrueUp and Layoffs.fyi), both aggregating from public reports with overlapping source material but different inclusion criteria, show tech redundancies at or above 2023 and 2025 levels. The 46% daily rate increase over 2025 is significant, though it does not by itself distinguish between structural restructuring and a cyclical downturn. The caveat: TrueUp's broad definition means some of these redundancies are at companies where technology is not the core business.
Metric 4 of 4
Rise in applications per vacancy
The original claim: applications rose 7% over the same period. This figure traces to SoftwareSeni without a specific primary source. However, the direction is confirmed, and the magnitude is likely higher.
Verdict: Cannot fully verify — direction confirmed, magnitude likely higher
The 7% figure cannot be traced to a primary dataset. But Handshake reports a 21% increase in applications per posting, and NACE hiring projections collapsed from +7.3% to +0.6%.
The original number is likely an understatement. The combination of falling postings and rising applications means the effective competition for each role has intensified significantly more than 7%. The NACE hiring projection collapse is a particularly stark demand-side indicator: employers planned to hire 7.3% more graduates in Fall 2024, but by Spring 2025, that projection had fallen to just 0.6%.
Source
Figure
Scope
Confidence
Handshake
+21%
Applications per posting on Handshake platform, year-on-year
HIGH
Burning Glass / BLS
5.2% to 6.2%
Unemployment rate for young graduates (20-24, BA+)
HIGH
NACE
+7.3% to +0.6%
Employer hiring projections for Class of 2025 vs Class of 2024
HIGH
Indeed Hiring Lab
Implied increase
Junior postings down 34%, applicant pool unchanged or growing
MODERATE
Hiring demand proxy: NACE employer projections
Only the two verified NACE data points are shown. NACE publishes one employer hiring projection per year, not a continuous time series, so no historical line is included. Handshake's +21% applications figure (shown in the text below) is a more robust measure of competition intensity.
NACE employer hiring projection (year-on-year %)
The demand-side squeeze is real, but the NACE data is thin. The +7.3% to +0.6% comparison comes from a single NACE report. NACE publishes one employer hiring projection per year; the change between the initial Fall 2024 projection and the revised Spring 2025 projection is striking, but it is one data point, not a trend. Handshake's +21% applications figure (year-on-year) is a more robust measure of competition intensity, and it confirms the direction.
Overall Assessment
How grounded is Signal 01?
Two of the four metrics are confirmed by primary sources. One is directionally accurate but imprecisely sourced. One cannot be fully verified but the direction is confirmed and the actual magnitude appears to be larger than claimed.
Confirmed
-36% job postings
Directly from Indeed Hiring Lab. The junior-senior divergence makes this even more concerning than the headline figure suggests.
Confirmed (broad)
148K redundancies
From TrueUp. The number is real, but the definition of "tech" is broad. The 46% daily rate increase is calculated correctly.
Cannot compare
-28% entry-level roles
The exact 28% traces to a commercial blog with unknown methodology. The verified range from primary sources is 20% to 45% depending on metric and period. The signal is real; the precise number is not comparable to the verified figures.
Cannot fully verify
+7% applications per vacancy
The 7% figure traces to SoftwareSeni without a primary source. Handshake reports +21%. The real squeeze is likely much larger than 7%.
Bottom line
Signal 01 remains Critical. The data is grounded, not speculative.
Every primary source confirms the direction. The Stanford/ADP study (50M workers, regression-controlled) finds a statistically significant independent effect of AI exposure on employment decline for young workers. Burning Glass finds the decline concentrated in AI-exposed occupations. Indeed finds a unique junior-senior divergence in tech that does not exist in other sectors. The only counter-argument (Indeed's "primarily cyclical" position) acknowledges the decline but disputes its cause. Even if the cause is debated, the effect is not.
Appendix
Validated sources
Every source cited in this report, grouped by metric, with confidence rating and direct link. Only sources with a confidence rating of HIGH or above are included. Secondary or low-confidence sources used for context only are marked with an asterisk.
Metric 1: Entry-level graduate roles vs 2022 peak
Source
Key finding
Methodology
Confidence
Link
Indeed Hiring Lab / FRED
-33%
Job Posting Index, non-seasonally adjusted, 7-day trailing average. Software development postings vs Feb 2020 baseline.
ADP payroll data covering 50M workers. Regression with firm fixed effects. Employment of software developers aged 22 to 25 from late 2022 peak. Working paper, not peer-reviewed.
Confidence key: VERY HIGH = published government dataset or peer-reviewed study with large sample. HIGH = primary dataset with transparent methodology. MEDIUM-HIGH = primary aggregator with broad scope. MEDIUM = secondary source or different methodology. MODERATE = contextual only, not directly comparable. LOW = qualitative or untraceable. Sources marked with * are secondary or contextual.
Lewis Barclay · Personal research, not professional advice · Updated June 2026
