AI has made top college grades easier to earn and harder to interpret. For startups hiring young talent, the transcript is becoming a weaker shortcut.
The problem with an A grade is no longer that it looks too good. The problem is that it may not say enough. Since ChatGPT arrived, the classroom work most exposed to generative AI, essays, reports, coding assignments and take-home projects, has become easier to polish and submit at a higher level than before.
That matters beyond campus. Startups hire heavily from early-career talent pools because they need energy, speed and ambition before candidates have long resumes. A strong GPA used to be one of the faster ways to screen for discipline and raw ability. It was never perfect, but it helped narrow a crowded field. Now that signal is getting noisier.
According to a Wall Street Journal report published May 14, the shift is being pushed by a new University of California, Berkeley working paper from Igor Chirikov at the Center for Studies in Higher Education. The paper analyzed more than 500,000 grades at a large public research university in Texas from 2018 to 2025 and found that courses with more AI-exposed work saw the share of A grades rise by 13 percentage points after ChatGPT's release, about 30 percent relative to the 2022 baseline.
The important detail is where the rise showed up. The increases were larger in classes where homework carried more weight, which points to AI helping with graded work outside the classroom rather than every student suddenly learning more. Some students are probably using AI as a tutor, editor or coding assistant in ways that resemble modern work. But the same grade can now represent very different levels of independent skill.
For a large company, a weaker GPA signal is annoying. For a startup, it can be expensive. A bad early hire does not disappear into a department. They slow product cycles, drain founder time and change the quality bar for everyone around them.
This is why grade inflation is also a labor-market story. Handshake data cited in the coverage shows employers are already reacting: among job postings that ask for grades, the share requiring at least a 3.5 GPA rose from 9 percent in 2020 to 25 percent this year. That looks like a higher bar, but it may also be the wrong bar. If A grades are becoming more common, raising the cutoff can push more candidates into the same crowded bucket.
The better response is to ask what the job actually requires. A software startup does not need a transcript to know whether a junior engineer can reason through a bug, read unfamiliar code and explain a tradeoff under time pressure. A sales-led company does not need a GPA to test whether a candidate can research an account, write a concise outreach note and handle a live objection. These are not abstract abilities. They can be observed.
Grades are not useless. They still say something about persistence, attendance, course selection and the ability to finish structured work. But founders should treat them as supporting evidence, not the center of the case. The candidate with a 3.9 and no visible work may deserve a screen. The candidate with a 3.3, a useful open-source contribution or a serious internship may deserve a longer look.
The New Talent Filter
The obvious next move is more work trials, but founders need to be careful. Long unpaid assignments are a bad filter because they punish candidates with less spare time and make the company look careless. The better version is short, paid and close to the real job. Give a candidate a messy customer note, a small dataset, a broken feature or a product brief. Ask them to show their thinking, not just submit a polished answer.
Interviews will also need to become more AI-resistant without pretending AI does not exist. A good interview can allow candidates to use AI, then ask them to explain what they accepted, what they rejected and why. That is closer to the way many jobs now work. The skill is whether they can judge the output, spot weak logic and connect it to the task in front of them.
This opens a real opportunity for edtech and HR-tech companies. The market does not need another badge that says a person completed a course. It needs better proof-of-skill infrastructure: verified portfolios, live assessment records, apprenticeship credentials, project histories and references tied to actual work. If colleges cannot make transcripts more meaningful, someone else will build a layer that sits between education and hiring.
Startups should also rethink where they look for junior talent. Apprenticeships, bootcamps, campus labs, student startups and creator portfolios may become more valuable because they show work in public or under closer review. A transcript shows an outcome. A portfolio can show judgment, taste and consistency over time.
The market implication is simple. As AI makes academic performance easier to dress up, hiring will move toward evidence that is harder to fake. The strongest founders will not ignore college credentials, but they will stop treating them as a shortcut. The next hiring advantage may belong to companies that can identify real ability before a transcript has the chance to blur it.
Also read: Poetiq shows model orchestration can beat bigger coding models • The metals selloff shows gold and silver are trading like liquidity assets • Tech layoffs are changing the startup hiring market in 2026
