The Credential Question: AI Is Repricing College, Not Killing It

Cover Memo

This report draws on CHIP50 Wave 38.6 (total unweighted N = 23,462; fielded June 21–July 13, 2026). Unlike the other reports in this series, "the credential question" is not solely a current-student question, so it uses four distinct populations, each labeled at point of use:

• General adult sample (~23,400) — views on whether AI makes degrees (bachelor's, master's, PhD) and specific fields more or less valuable.
• High-school students (n ≈ 1,180) — whether AI has changed their college plans (ai_college_forego).
• Parents (n ≈ 7,000) — whether AI has changed their thinking about sending children to college (parent_view) and how much they value AI in the curriculum (parent_curric).
• Currently-enrolled students (n ≈ 2,330) — how well their program has prepared them to use AI (ai_edu_prep).

All percentages are survey-weighted to national demographic targets; unweighted N is a reliability cue only. Figures are self-reported.

Key Takeaways

Introduction

The previous report in this series asked whether AI levels the playing field inside school, and found a leveler with an asterisk: real help for the students who have the fewest alternatives, reaching them least. This report asks the question one step further out — whether the thing they are enrolled in is still worth being enrolled in.

Public faith in college was already sliding before AI arrived, and it is still sliding: Gallup finds confidence in U.S. higher education at 38% in 2026, down from 57% a decade earlier, with cost, politics, and "poor workforce preparation" the leading complaints.¹ Into that low-confidence moment, generative AI drops a pointed new question — if a machine can write the essay, pass the exam, and do the entry-level job, what is the degree still for? Gallup finds Americans uneasy: 46% think AI will make a college degree less important over the next five years, against just 20% who think more.¹

CHIP50's Wave 38.6 supplement lets us take that anxiety apart — by degree level, by field of study, and, crucially, by asking the people whose decisions actually matter: high schoolers weighing whether to go, parents weighing whether to send them, and students already enrolled. The result is more specific, and less apocalyptic, than the headline dread. Americans are not writing off college so much as repricing it — discounting some degrees and fields while holding or bidding up others — even as the students inside the system worry it isn't teaching them the one skill the moment demands.

A Discount, Not a Write-Off

Asked whether AI makes a college degree more or less valuable for future success, Americans lean negative — but only slightly, and with a large undecided middle. Among all adults, 32% say AI makes a bachelor's degree less valuable (20% somewhat, 12% much), versus 23% who say more (10% much, 13% somewhat); 30% say it makes no difference and 15% aren't sure (Table 1). That is a real net-negative tilt, and it is directionally consistent with Gallup's more starkly worded finding — but it is a discount, not a write-off: a majority either see no change or can't say.

The discount also shrinks as the credential gets higher. Repeating the question for a master's degree, the net tilt narrows (24% more valuable vs. 29% less); for a PhD, it flips positive (27% more valuable vs. 23% less). In the public mind, in other words, AI erodes the value of the entry-level credential most and the advanced credential least — the bachelor's is where the anxiety concentrates, while the doctorate is seen as comparatively AI-proof.

The discount is also not evenly felt across groups. Net of party, income, and education, Black and Hispanic adults are more likely than White adults to see a degree as retaining its value in an AI economy — a notable vote of confidence in the credential from groups for whom it has often been the surest ladder up. (Asian adults, by contrast, look like White adults once the large "not sure" share is handled properly.) Higher-income adults are likewise more bullish on the degree, while urbanicity and region make no difference. The devaluation anxiety, in other words, is concentrated among White and lower-income Americans more than among the Black and Hispanic Americans the "is college still worth it" debate often invokes. (This item was asked of the general adult sample, not of students.)

Comparative context. Gallup (2026) finds 46% of Americans expect AI to make degrees less important within five years vs. 20% more¹ — a sharper version of the same tilt CHIP50 sees, with CHIP50's larger "no change / not sure" share reflecting its "value for success" framing.

STEM Up, Humanities Down

The degree anxiety is not uniform — it is a bet on fields. Asked which undergraduate majors AI makes more or less valuable, Americans draw a sharp line between the technical and the humanistic (Table 2). Engineering (42% more valuable vs. 26% less, net +17) and Computer Science (44% vs. 29%, net +15) top the list, followed by Biology (net +7). At the bottom sit the humanities and social sciences: Philosophy (net −8), Political Science (−7), and Economics (−6), with English and Communication roughly even.

The pattern is intuitive and a little ironic — the public expects AI to raise the value of the very fields (computer science, engineering) that build it, and to lower the value of the fields (philosophy, political science) most concerned with how societies should govern it. It also lends popular backing to the "AI is coming for the liberal arts" narrative: the degrees Americans think AI devalues are precisely the writing-and-reasoning majors whose signature outputs a chatbot can now imitate.

The People Deciding Aren't Fleeing

For all the ambient anxiety, the people actually making college decisions have barely moved. Among high-school students, 60% say AI has not affected their plans to pursue a college degree at all; the rest split almost evenly between those it has made less likely to go (12%) and more likely (13%), with 15% undecided (Table 3). Whatever the think-pieces say, AI is not — yet — triggering a measurable exodus from college among the teenagers closest to the decision.

Parents lean the other way, if anything. Asked whether AI has changed their thinking about sending their children to college, 27% of parents say it makes them more likely to send them, versus 12% less (61% report no change). And parents want colleges to adapt rather than be abandoned: 58% say it is at least somewhat important that their child's college actively integrates AI into its curriculum, including 23% who say it matters "a lot" (Table 3). The demand signal from families is not "skip college" but "make college teach this."

The Real Gap Is Preparation

If there is a credential crisis in this data, it is not abandonment — it is preparation. Asked how well their academic program has prepared them to use AI for work, 42% of currently-enrolled students say "not very well" or that their program has not addressed AI at all (21% each), against just 14% who feel "very well" prepared and 33% "somewhat" (Table 4). Roughly a fifth of students, in other words, are in programs that — by their own account — are ignoring the technology reshaping their field, even as two-thirds of them have used it for coursework and their parents are asking schools to teach it. (The intensity is lower than the alarm suggests: 18% of currently-enrolled students do "most" or "all" of their schoolwork with AI. The complaint is not that programs are failing to keep up with heavy users; it is that they are silent on a tool most students have already picked up.) Report 3 documents that silence from inside the classroom: AI is largely absent from course requirements, and where students meet a stance at all it is an individual instructor's, not a policy.

This is the credential question's real bite, and it connects directly to why confidence in college is falling: Gallup finds "poor workforce preparation" among the top reasons Americans lose faith in higher education.¹ The threat to the degree in an AI economy may be less that AI makes college obsolete than that colleges are too slow to teach AI — leaving graduates credentialed in the old skills and self-taught in the new ones. That is the thread this report hands to the companion "AI @ Work" series, where the same question reappears from the employer's side of the desk: what is a degree worth when the skills that matter are the ones the degree didn't cover?

Conclusion

The fear that AI will kill the college degree is, on this evidence, overstated — but the complacent opposite is wrong too. Americans are repricing higher education in real time: marking down the bachelor's and the humanities, holding or raising STEM and advanced degrees, and doing it against a backdrop of decade-low confidence in college generally. Yet the households making the actual choice are staying the course — high schoolers' plans are steady, parents lean in and want AI taught — which suggests the degree's near-term problem is not demand but relevance. Students are telling us where the gap is: two in five say their programs aren't preparing them for the tool their working lives will run on. Close that gap and the credential question largely answers itself; leave it open and the slow devaluation the public already senses becomes self-fulfilling. Either way, the pressure point is not whether to go to college but what college teaches once you're there.

Appendix A — Methods

Data. CHIP50 Wave 38.6, fielded June 21–July 13, 2026; total unweighted N = 23,462.

Populations. Degree-value and field items (ai_deg, ai_deg_masters, ai_deg_phd, ai_degree_fields_1–9): general adult sample (~23,400). College-plan item (ai_college_forego): high-school students (n ≈ 1,180). Parent items (parent_view, parent_curric): parents (n ≈ 7,000). Preparation item (ai_edu_prep): currently-enrolled students (n ≈ 2,330).

Weighting. Survey-weighted to national demographic targets; unweighted N is a reliability cue only. Self-reported throughout.

Measures (verified coding). ai_deg / ai_deg_masters / ai_deg_phd — "Does AI make a [bachelor's/master's/doctoral] degree more or less valuable for future success?" (1 = much more … 5 = much less; 6 = not sure). ai_degree_fields_1–9 — "Will the following undergraduate degrees be more or less valuable due to AI?" (1 = more valuable, 2 = no change, 3 = less valuable); fields: 1 Economics, 2 Math, 3 Computer Science, 4 Political Science, 5 Communication, 6 English, 7 Philosophy, 8 Engineering, 9 Biology. ai_college_forego — "Has AI affected whether you plan to pursue a college degree?" (1 = less likely, 2 = more likely, 3 = no change, 4 = undecided). parent_view — "Has AI made you rethink whether to send your children to college?" (1 = much more likely … 5 = much less likely). parent_curric — "How important is it that your child's college integrates AI?" (1 = not at all … 4 = a lot). ai_edu_prep — "How well has your academic program prepared you to use AI for work?" (1 = very well, 2 = somewhat well, 3 = not very well, 4 = has not addressed AI at all, 5 = not sure).

Appendix B — Data tables

Table 1. Does AI make a degree more or less valuable? (general adult sample; weighted %)

Base ≈ 23,400 adults. "More" = much + somewhat more; "less" = much + somewhat less. Weighted %; rows may not sum to 100 due to rounding.

Table 2. Which undergraduate fields does AI make more or less valuable? (general adult sample; weighted %)

Base ≈ 23,400 adults. Middle category ("no change") omitted. Weighted %. Net is computed on unrounded percentages and will not always equal the difference of the two rounded columns shown (e.g. Engineering is 42.34 − 25.54 = +16.8, displayed +17).

Table 3. The decision-makers: high schoolers and parents (weighted %)

Bases: HS students n ≈ 1,180; parents n ≈ 7,000. Weighted %.

Table 4. How well did your program prepare you to use AI? (currently-enrolled students; weighted %)

Base ≈ 2,330 currently-enrolled students. Weighted %. "Poorly / not addressed" (rows 3–4) = 42%.

Demographic robustness (SES, urbanicity, region)

The race finding on degree value — Black (−0.19) and Hispanic (−0.12) adults more likely than White adults to see a degree as remaining valuable, net of controls and computed on the "not sure"-excluded outcome (n ≈ 19,900) — is integrated in the body ("A Discount, Not a Write-Off"). Asian adults do not differ from White adults on the clean outcome (β = −0.02, p =.51); the −0.12 in an earlier run was an artifact of leaving ai_deg's "6 = not sure" category in as an ordinal extreme, the very contamination this report's ai_deg note warns against. SES matters strongly: higher-income adults are more bullish on the degree (with a ai_deg "not sure" caveat on the education gradient). Urbanicity and Census region are null. On the preparation item (ai_edu_prep), Black and Asian students report feeling better prepared net of controls, but that item's "not sure" category muddies its mean, so it is not featured.

Appendix C — Question wording and response codes

Source: the Wave 38.6 Qualtrics instrument provided by the CHIP50 / COVID States team (the Wave 38.6 codebook, 2026-07-20), which is the authoritative record of item wording and stored numeric codes for this wave. Stems below are the codebook's condensed forms, not screen-verbatim text; response options and codes are verbatim. Where an item's stored code order differs from its on-screen display order, the stored codes are authoritative and are the ones used throughout this report.

• ai_deg (all adults) — "Does AI make a college degree more or less valuable for the future?" (1 = Much more; 2 = Somewhat more; 3 = Neither; 4 = Somewhat less; 5 = Much less; 6 = Not sure). "Not sure" is a substantive category and is retained in the denominator.
• ai_deg_masters / ai_deg_phd (all adults) — the same question about a master's degree and a doctoral degree (same scale).
• ai_degree_fields_1–9 (all adults) — "Will these undergraduate degrees be more or less valuable because of AI?" (1 = More valuable; 2 = No change; 3 = Less valuable). Fields: 1 Economics · 2 Math · 3 Computer Science · 4 Political Science · 5 Communication · 6 English · 7 Philosophy · 8 Engineering · 9 Biology.
• ai_college_forego (high-school students) — "Has AI affected your plan to pursue a college degree?" (1 = Less likely; 2 = More likely; 3 = No effect; 4 = Undecided).
• parent_view (parents) — "Has AI made you rethink sending your children to college?" (1 = Much more likely; 2 = Somewhat more likely; 3 = No change; 4 = Somewhat less likely; 5 = Much less likely).
• parent_curric (parents) — "How important is it that your child's college integrates AI?" (4 = A lot; 3 = Somewhat; 2 = Not so much; 1 = Not at all).
• ai_edu_prep (currently-enrolled students) — "How well has your program prepared you to use AI for work?" (1 = Very well; 2 = Somewhat well; 3 = Not very well; 4 = The program hasn't addressed AI; 5 = Not sure).