Is the job market actually worse for AI-exposed majors, or is it just me?
It is genuinely worse, and you should stop blaming yourself for it. A US Census Bureau working paper (Center for Economic Studies, CES-WP-26-56, "Graduating into Disruption," a 2026 working paper reported by Bloomberg on September 14, 2026) found that in the most AI-exposed decile of college majors, the likelihood of initial employment fell by five percentage points and full-quarter initial earnings fell by thirteen percent after the introduction of ChatGPT. The authors compare that decline to graduating into a large recession.
That framing matters. Economists have long documented that graduating during a downturn can suppress earnings for years, and yet nobody blames those graduates for the timing of their birth. You are in a comparable situation. The market shifted underneath a degree that, when you enrolled, looked like one of the safest choices available.
According to the study authors (reported via Bloomberg and Staffing Industry Analysts on September 14, 2026), the most AI-exposed decile is dominated by computer science, information systems, and other software-related fields. So if your major sits in that group, the headwind is not imaginary and it is not a reflection of your ability.
Why can't I find work when everyone said tech was recession-proof?
Part of the reason is that the entry-level rung has thinned, and part is that graduates are being pushed into work far outside their field. The Census study authors noted that some of the earnings decline came from graduates taking lower-paying jobs in fields such as retail and restaurants. That is the mechanism behind the numbers: it is not only that fewer roles exist, but that people are accepting jobs well below their training to keep income flowing.
The broader picture confirms this squeeze. The Federal Reserve Bank of New York's report on the labor market for recent college graduates (updated in 2026 with Q2 2026 data) put recent-grad unemployment at roughly 5.6 percent and, more strikingly, the underemployment rate at 42 percent. Underemployment measures graduates working in jobs that typically do not require a degree. When more than four in ten recent graduates are underemployed, a hard search is the norm, not the exception.
Understanding this changes your strategy. If the entry-level software title you trained for is scarce, the goal shifts from landing the perfect first role to getting a foothold that keeps you building relevant skills and avoids the underemployment trap for as long as you can.
What should I actually do first?
Lead with demonstrable work, not just the degree. Employers hiring into AI-exposed fields are increasingly skeptical that a credential alone signals capability, partly because AI tools have made a lot of routine output cheap. What has become more valuable is proof that you can define a problem, use the available tools well, and ship something that works.
Build a small portfolio of projects that show judgment, not just syntax. That might mean an application that solves a specific, real problem for a specific group of people, documented so a hiring manager can see your reasoning. Show how you used AI tools inside your workflow rather than pretending you avoided them; the graduates who thrive in an AI-exposed field are usually the ones who direct the tools rather than compete with them. Position yourself as someone who makes AI productive, because those are the roles least threatened by it.
Write about what you built. A short technical post explaining a decision you made and the tradeoffs involved does more to establish credibility than another line on a resume. It also gives interviewers something concrete to ask about, which shifts the conversation toward your strengths.
Where should I look if software roles are scarce?
Widen the target beyond the obvious titles, and go where technical fluency is rare rather than abundant. Every industry now needs people who can bridge the gap between technical systems and the humans who use them. Healthcare, logistics, manufacturing, insurance, and the public sector all employ technical talent, and they often compete less aggressively for it than the large software firms do. A computer science graduate who can talk to non-technical stakeholders and automate a tedious internal process is valuable in places that would never appear in a search for "software engineer."
Also consider adjacent roles that keep you close to the work: data analysis, technical support that leads to engineering, implementation and solutions roles, quality and testing, and operations positions that involve tooling. These are not detours if you choose them deliberately. They keep you employed above the underemployment line, they build a track record, and many convert into the role you originally wanted within a year or two.
If you do have to take a job outside your field to pay rent, protect a few hours each week for skill-building and applications. The Census research suggests the real long-term risk is not one imperfect job; it is drifting into work that erodes the momentum of your training and makes the next move harder.
How do I rebuild earning power over time?
Treat the first job as a launchpad and negotiate from evidence as soon as you have it. The thirteen percent earnings gap the Census study measured is an initial figure, tied to first jobs. Initial does not mean permanent. As you accumulate shipped work and references, your leverage grows, and internal moves or a switch after twelve to eighteen months are often where the pay recovery happens.
Keep your search relationship-driven rather than purely application-driven. In a thin market, referrals and direct conversations with people doing the work you want move you past the automated filters that reject most applicants. Reach out to alumni and to engineers whose projects you admire, ask specific questions, and let those conversations surface roles that were never posted. A challenging market rewards persistence and specificity far more than volume.