Where did the junior jobs go?

How much experience employers ask for

Each bar splits the roles posted in the last month by the years of experience the ad actually demands. Junior-friendly openings (two years or less) are a real minority, roughly one in five in software and far fewer in product, but nowhere near as scarce as job titles alone suggest. Most of the squeeze is the wall of roles asking for three to five years, the rung that used to be the second job, not the first.

Live postings from our job board, US-located, first seen in the last 30 days (a flow of new openings, not the standing stock, which would over-weight slow-to-fill senior roles). Experience is the minimum years parsed from the posting body. Bars cover only the ~50% of postings that state a requirement; the rest are unstated and shown as coverage. New-grad hiring is also seasonal (autumn-heavy) and often flows through university channels we do not scrape, so this undercounts the true junior pipeline.

Junior share over time

The same junior share, tracked over time per discipline. This is the line to watch: if AI is hollowing out the bottom rung, the entry-level share keeps drifting down. Our posting archive is shallow, so this series builds forward from the day we began logging it; it cannot reach back before the AI boom.

Roles asking for two years or less as a percent of US postings that state a requirement, in each function, one point per snapshot.

Software demand never recovered

Zoom out and the level of software hiring overall (every seniority, not just juniors) has fallen well below its early-2020 mark and stayed there. This is the pool everyone competes in, and the context for the experience story above: a shrunken market squeezes the least-experienced applicants first, AI or no AI. We cannot yet show whether the junior slice fell faster than the whole, that is what the forward trend is for.

Indeed Hiring Lab software-development postings index for the US, all seniorities, where 100 = the level on February 1, 2020. Note that 2020 was a hiring peak after a long expansion, so it is a high baseline, not "normal." Daily series downsampled to monthly. Indeed Hiring Lab, CC-BY.

US software roles by experience

The same software market as the line above, split by the experience employers ask for. This is the picture the backdrop can't show: whether the junior slice is shrinking faster than the mid and senior ones. It builds forward from the day we began logging it, so the line is short at first.

US software postings first seen in each 30-day window, split by required experience (junior 0 to 2y, mid 3 to 5y, senior 6y+), one point per snapshot. "Share of roles" is each bucket as a percent of postings that state a requirement (the view that shows whether junior is shrinking faster); "Number of roles" is raw volume. Note this is our board's title-classified software function, a different universe than the Indeed software-sector index above, so read them side by side, not as the same line.

Employers are asking for AI skills

The clearest AI signal in the data is demand, not displacement: the share of US postings that ask for AI or generative-AI skills has climbed, though it is still a low single-digit slice. That is employers wanting people who can use AI, which if anything adds roles, not a measure of jobs replaced by it. Rising AI demand and a thin junior pipeline are happening at the same time; this page does not claim one caused the other, and nothing here can.

Share of US job postings containing AI / GenAI-related terms. Indeed Hiring Lab AI tracker, CC-BY, monthly. This measures skill demand (adoption), not automation of existing jobs.

Sources, definitions, and sharp edges

• SIGNAL Seniority here is the minimum years of experience parsed from the posting body (junior 0 to 2, mid 3 to 5, senior 6+), not the job title. Titles badly undercount juniors because most early-career roles are titled just "Software Engineer" with the real bar ("2+ years") in the text. Parsing is heuristic, so we only count postings that state a number and report that coverage; the rest are "unstated" (shown as the faded segment on each bar). Postings that state a requirement also skew toward larger, more formal employers, who write structured reqs, so the stated slice is not a random sample of the market.
• FLOW Counts cover postings first seen in the last 30 days, the flow of new openings, not the open stock. Senior vacancies stay open longer and would inflate the senior share in a stock snapshot, biasing the junior number down. Each over-time point is one of these trailing-30-day windows, and consecutive snapshots overlap heavily, so the lines are smoothed and lag real turning points by roughly two weeks; read the trend, not the day-to-day wiggle.
• COVERAGE Our board scrapes ATS feeds (Greenhouse, Lever, Ashby, big tech). New-grad and entry hiring leans on university recruiting, Handshake, and seasonal early- career portals we do not scrape, and peaks in autumn, so the junior pipeline here is understated. US-located only; remote roles with no parsed country are excluded.
• DEPTH Postings are deleted about a week after they go stale, and our snapshots only began recording this breakdown recently, so the internal trends build forward and cannot reconstruct the past. This is why the "US software roles by experience" chart and the junior-share trend both start from the day we began logging and are short at first. The Indeed backdrop is a single all-seniority index (Indeed publishes no seniority split), so it cannot be broken down by experience the way our own forward-accruing series can.
• BACKDROP The multi-year software demand line (all seniorities) and the AI-terms share are Indeed Hiring Lab open data (CC-BY): the software index is rebased to February 2020 = 100 (a peak), both daily and downsampled to monthly. The AI share is demand for AI skills, an adoption signal, not automation.
• CAUSATION Nothing here proves AI caused anything. The 2022 to 2023 tech layoffs, higher interest rates, and the unwinding of pandemic over-hiring all hit junior roles first and overlap exactly with the rise of generative AI. We show the timelines side by side and let you judge.