The Widest Door.
The Class of 2026. A quarterly snapshot of where work is going. From Navii Research.
AI hiring is 1.42 times more open to someone entering the workforce than the rest of the job market. For the Class of 2026, it is the best entry-level bet available anywhere in hiring.
We expected the opposite. So did almost everyone.
The advice to aim at AI was right. Nobody mentioned the clock.
The dominant story about this graduating class is that AI closed the door on them. We went looking for that door, expecting to find AI hiring the tightest entry point in the market. We found the opposite, and it is the most useful thing in this issue.
AI hiring is the widest entry-level door we measured. 3.14% of AI postings are open to someone entering the workforce, against 2.22% across the job market as a whole — a 1.42× advantage. For a graduate deciding where to aim, that is not a small edge. It is the best odds on the board.
We analyzed 10,453,124 job postings opened in June 2026 against the identical calendar window in Q1, alongside 21,864 first-person messages from 3,543 graduating students on Navii's own platform, and the best external microdata available.
What we found:
- 1AI is the widest entry-level door in hiring — 1.42× the market average. We expected the opposite, and had we not measured the market baseline we would have implied the opposite.
- 2This cohort is more prepared than the one before it. 64.0% of graduating students on Navii arrive with at least one internship, up from 56.3% for the Class of 2025. Nearly a third have two or more.
- 3They are not afraid of AI. They want AI jobs — by a ratio of 221 to 2. Across 40,589 messages, displacement language appears in twelve.
- 4The internship converts at a five-year high: 63.1%. Graduates with internship experience are hired at 81.6% versus 40.7% without. There is a path, and it is working better than it has in years.
- 5The clock is real. AI's entry-level advantage fell from 1.57× to 1.42× in a single quarter. Entry-level roles grew in both markets, but far slower than hiring overall. AI went from 3,846 to 3,859 entry-level postings — a net gain of 13 — while adding 20,507 postings overall.
The Class of 2026 was pointed at AI, and the advice was right — AI is the widest door in the market and this cohort is better prepared to walk through it than any before them. What nobody told them is that the door is narrowing while they stand in it. The opportunity is real. So is the clock.
For hiring leaders, the junior pipeline you assume exists is roughly two percent of the market, and converting your own interns is the highest-leverage move available. For graduates, aim at AI, get inside any way you can, and ship something real while you are there.
AI is the widest entry-level door in the job market.
We built this issue expecting to report that AI hiring had closed to new graduates. Every prior signal pointed there: PwC's finding that AI-exposed entry-level roles are 7× more likely to demand senior skills, Stanford's 16% relative employment decline for 22-to-25-year-olds in AI-exposed occupations, Harvard's 9% junior decline at GenAI-adopting firms.
Then we measured the baseline, and the comparison inverted.
3.14% of AI postings are open at entry level. Across the whole job market, it is 2.22%. On the specific question a graduate is actually asking — where are my odds best? — the answer is AI, and it is not close.
Two questions, two different answers
This issue reports two measures that sound contradictory and are not. Keeping them apart is essential to reading the data correctly.
Question 1 — where are the odds best today?
This is a level. Answer: AI, by a wide margin.
Question 2 — which way is it moving?
This is a change. Answer: against you, and faster in AI.
| Q1 → Q2 2026 | All jobs | AI |
|---|---|---|
| Entry-level postings, Q1 | 219,945 | 3,846 |
| Entry-level postings, Q2 | 231,999 | 3,859 |
| Net change | +12,054 (+5.5%) | +13 (+0.3%) |
| Total postings added | +1,253,148 (+13.6%) | +20,507 (+20.0%) |
| Entry-level share | 2.39% → 2.22% | 3.76% → 3.14% |
Two things to read carefully. First, entry-level postings grew in both markets; the share still fell because total postings grew faster. That is arithmetic, not contradiction. Second, the small AI number is a change, not a total. There are 3,859 entry-level AI postings — that is the 3.14% share above. Thirteen is how many more there were than the previous quarter.
Put together: AI is the best place to be standing, and the worst place to stand still. It starts from a better position and is deteriorating roughly twice as fast.
We are stating this prominently because we nearly published the opposite. Had we measured only the AI market, as our Q1 issue did, we would have reported a narrowing AI entry level without the context that makes it interpretable — and implied that AI is uniquely closed to new graduates. It is not, and anyone reporting that is wrong.
Nobody is firing this cohort. They are adding senior roles far faster than junior ones.
What people think is happening: AI is destroying entry-level jobs, and this year is the crisis point.
What is actually happening: the entry level has been structurally inverted for five years, and recent-graduate unemployment is roughly flat year over year (5.67% to 5.63%). The level is the story. The trend is not.
The outlier worth noting: computer science graduates have the highest unemployment rate of all 74 majors tracked (7.0%, and 7.8% for computer engineering) — alongside the lowest underemployment and the highest wages. The degree is not broken. The entry gate is.
This cohort is more prepared than the one before it.
The dominant story about graduates is that they are underprepared for an AI-shaped labor market. In our first-party data, the opposite is true. That is a 7.7-point increase in a single year, and nearly a third have two or more internships.
AI-named fields of study went from 0.33% of the Class of 2022 to 6.21% of the Class of 2026, with 7.84% already in the 2027 pipeline.
External data agrees on why this matters: employers now rate “internship with our organization” at 4.5 out of 5 as the deciding factor between equally qualified candidates — the only attributes in the “very much influence” band — while GPA screening collapsed from 73.3% of employers in 2019 to 42.1% in 2026. This cohort built exactly the credential that now decides hiring.
They adapted faster than the discourse about them did.
They want AI jobs. They do not fear them.
Across 21,864 first-person messages from 2,525 graduating students, we searched for the fear that dominates coverage of this generation. It is essentially absent.
Of 438 students who mention AI, 221 want an AI job title. A sweep of all 40,589 messages found displacement language in twelve — 0.03% — and on inspection only two students genuinely feared being replaced. Among 60 messages expressing job-search anxiety, exactly one mentions AI and zero mention replacement.
More striking: distress does not spike at graduation. Measured on identical instruments, the Class of 2026 and the Class of 2025 are statistically indistinguishable — anxiety 2.1% vs 2.4%, rejection 1.5% vs 1.8%, ghosting 0.5% vs 0.7%. Where the doom narrative predicts a spike, we found a null result.
The fear in the coverage is not the fear in the conversations. This cohort is pointed at the opportunity, not away from it.
The internship is converting at a five-year high.
For anyone still in school, this is the most actionable number in the issue. Conversion is up roughly 13 points, with an 88.3% acceptance rate. Employer intent also flipped positive: after projecting a 3.1% cut the prior year, employers now expect to bring in 3.9% more interns.
Worth noting because it is widely misreported: NACE's “conversion at a five-year low” line is from April 2025 and has since reversed sign. It is still being quoted in 2026 coverage.
The honest caveats: only about half of four-year seniors get an internship at all, 31% of four-year internships are unpaid with offer rates at or below having none, and NACE's denominator counts only graduating, job-seeking interns.
The path in is working better than it has in years. The problem is how many people get onto it.
The step from intern to junior is a deployment step.
We measured skill requirements across all 2,304 intern and 2,288 cleaned junior AI postings — population-level, not sampled.
Read the change column carefully. Almost every skill requirement falls from intern postings to junior postings — Python, PyTorch, TensorFlow, even agentic experience. Only one family rises, and it rises steeply: containers, orchestration, and cloud. Kubernetes appears six times more often. Docker nearly four.
This is good news, because it is a short and specific list. What separates a junior posting from an internship is not more theory or a better transcript. It is deployment — evidence you can put something in front of a user and keep it running.
The gap between intern and hired is not more theory. It is proof you can put something in production.
Population-level across all June 2026 postings. Not a sample.
The clock — the market grew, the door didn't.
Everything above is the opportunity. This is the urgency.
AI's entry-level advantage over the market fell from 1.57× to 1.42× in a single quarter.
The mechanism is the same in both markets, and much more extreme in AI. Entry-level roles did grow — the whole job market added 1.25 million postings and 12,054 entry-level roles, a 5.5% increase. AI hiring added 20,507 postings and went from 3,846 to 3,859 entry-level postings — a net gain of 13. The small number is a change, not a total.
Neither market is shedding entry-level jobs. Both are adding senior ones far faster, and AI is doing it at roughly twice the rate.
Inside the specific categories most people aim at, the floor is lower still. Strip internships out — they are 69% of everything tagged junior in those categories — and the genuinely entry-level share lands between 0.1% and 2.2%. AI Product Manager is 0.41%. Forward Deployed Engineer, 0.42%.
And the first rung is moving. US AI internship postings fell 45.5% while India's rose 9.0%; India now posts 716 AI internships to the United States' 518. This is not primarily an offshoring story — Indeed calls offshoring “subtle, but not primary” — but the American intern pipeline contracted sharply, and that is where most US graduates enter.
June 2026. Narrower curated basket than the market-wide figures. Categories with fewer than 500 postings omitted.
What that costs, in one student's words. Three times in an afternoon, politely:
“i want to join in a startup that considers entry level AI Engineers”
Forty-two minutes later, the same person:
“IM NOT GETTING ANY CALLS BACK , I NEED AI ENGINEER JOB”
Nothing about AI taking their job. Everything about a door that is open, and narrowing, and hard to find.
Anonymized. No name, school, employer, or location is retained.
The opportunity is real. So is the clock.
What the data means if you lead a hiring team.
The junior pipeline you assume exists is two percent of the market.
Entry-level roles are 2.22% of all postings and 3.14% of AI postings. We also checked whether the companies running the largest AI internship programs convert that pipeline into junior roles externally. At IBM, NVIDIA, Microsoft, and TikTok, intern postings held flat or rose while junior postings held flat or fell. Microsoft posted zero junior AI roles in both quarters. Amazon and Google posted zero AI internships. Conversion may well be happening internally — that is the honest caveat — but externally, these employers have no junior door at all.
You are not reading the applications you already receive.
Applications per job are up 111%, and per recruiter up 411%. Interviews per hire rose 4.5%. A candidate today is roughly half as likely to reach an interview as five years ago. The bottleneck is not applicant supply or applicant quality — it is that the volume broke the process, and the response was to stop reading rather than to read differently. The leverage in your funnel is at the top of it.
Stop screening for AI skills you rank last.
Employers rank AI skills eighth of eight competencies (2.8/5) and sixteenth of twenty-one résumé attributes; 42.5% are not seeking them at all. It is the only competency where employers rate graduates more proficient than the skill is important. Meanwhile students are optimizing hard for exactly this signal. If you want a differentiated pipeline, screen for the deployment skills in Finding 5 — they are scarce, teachable, and actually predict the work.
If you are allocating budget this quarter, the leverage is in converting your own interns and in reading your existing funnel — not in competing for senior AI talent everyone else is bidding on.
What the data means if you're navigating a career.
The internship is no longer an advantage. It is the door.
Intern-to-full-time conversion reached 63.1% in the most recent cycle, a five-year high, with an 88.3% acceptance rate. Graduates with internship experience are hired at 81.6% versus 40.7% without. Employers rate an internship at their own organization 4.5/5 as the tiebreaker. With the external entry-level door at 2.22% of the market, the internal path is not one route among several — it is the main one. If you are still in school, this is the single highest-return use of your time, and it outranks GPA by a wide margin.
AI is still the wider door. Walk through it with your eyes open.
AI hiring is 1.42× more open at entry level than the market average. That advantage is real and worth pursuing — but it shrank from 1.57× in a single quarter, and AI entry-level postings went from 3,846 to only 3,859 while the AI market added 20,507 postings overall. Aim at AI because the odds there are better than average, not because they are good.
Your competitive edge is deployment, not another model course.
The measured gap between intern and junior postings is Docker (4.7% to 17.5%), Kubernetes (2.1% to 13.0%), Azure (9.3% to 17.8%), and RAG (10.5% to 15.0%). Ship one thing to production that a real person uses. That single artifact addresses the “1–2 years experience” wall more directly than any additional coursework, because it is evidence of the exact capability the junior tier is screening for.
Do not optimize only for the AI label.
AI-named degrees went from 0.33% to 6.21% of graduating cohorts in four years — the signal is crowding fast, and employers rank AI skills last among competencies they actually screen. The scarce thing is not AI familiarity. It is judgment, communication, and the ability to finish something. Employers rank critical thinking and communication first. Pair the domain you know with demonstrated AI fluency; do not abandon the domain to chase the label.
If you are planning your next six months: get inside a company any way you can, and ship something real while you are there.
What we're watching for Q3 2026.
This quarter's data points to three things we will check — and publish — in the next issue.
Does the entry-level share recover with the autumn cycle?
US summer-internship recruiting peaks the preceding autumn, so a June window sits at a seasonal trough by construction. If the Q3 all-market entry share does not recover above 2.39%, and the AI share above 3.76%, the contraction is real rather than seasonal.
Does AI's entry-level advantage keep eroding?
It fell from 1.57× to 1.42× in one quarter. If it drops below the market average, AI stops being the better bet for new graduates, and the advice in this issue changes.
Does India's internship lead survive a US-favorable window?
India leads 716 to 518 on a window that disadvantages the US. Q3 captures the American autumn cycle. If India still leads, the crossover is structural.
We will revisit all three in the Q3 issue, including any that prove us wrong.
Statistics the field is getting wrong.
In preparing this issue we found several widely circulated statistics to be wrong or misattributed.
- ▸There is no “WEF Future of Jobs Report 2026.” The most recent edition is January 2025. The 170M/92M/78M figures belong to that edition.
- ▸A frequently cited “80% entry-level decline at AI-adopting firms” misstates the underlying Harvard finding of 7.7%.
- ▸A “35% entry-level decline” attributed to the World Economic Forum comes from an opinion column carrying WEF's own disclaimer, and refers to postings, not employment.
- ▸NACE's dashboard cites bachelor's unemployment of 8.9% for June 2026. BLS data shows 8.9% is June 2025 for all 20-to-24-year-olds, and June 2026 is 7.7%. The sign is reversed.
We flag these because we nearly published two of them ourselves.
Data sources
A proprietary global job-posting index (June 1–30, 2026 window, matched against the same window in Q1); Navii first-party platform data; and published external research cited inline — NY Fed Labor Market for Recent College Graduates, BLS, NACE, PwC's 2026 Global AI Jobs Barometer, Stanford Digital Economy Lab / ADP, Harvard (Hosseini & Lichtinger), Handshake, Indeed Hiring Lab, Greenhouse, Ashby.
Time window
Q2 close, June 1–30, 2026, compared against the identical calendar window in Q1. We used explicit date bounds rather than a trailing-30-day window, which on our publication date would have measured Q3.
Sample
10,453,124 postings market-wide and 122,879 AI-titled postings in the Q2 window. Entry-level counts are seniority-tagged postings minus internships, applied identically to both the market and AI samples. Category-level analysis covers the tracked AI categories; 2,304 intern and 2,288 cleaned junior postings analyzed at population level for skills. Navii first-party: 3,543 Class of 2026 students identified by structured graduation-year extraction, compared against 4,739 from the Class of 2025; 21,864 first-person messages from 2,525 students.
A note on two AI baskets
This issue reports AI hiring at two levels of resolution. The market-wide comparison uses a broad AI title basket (122,879 postings) so it can be set against the whole market on identical terms. The category table in Finding 1 uses the narrower curated categories we track quarter to quarter. The two are not interchangeable and we label which is which at every use.
Limitations
- ▸Postings are hiring intent, not hires. Title-based search undercounts; all figures are lower bounds.
- ▸Two data points cannot separate trend from season. The June window sits at a seasonal trough for US internship recruiting. See Outlook.
- ▸Seniority tags are assigned by the source platform and are imperfect in both directions. We use them because they are applied uniformly across our market and AI samples, which is what the central comparison requires.
- ▸Only 50.9% of Navii profiles parse into structured education data. Our users self-selected into an AI career product and are not representative of all graduates.
- ▸Category-level junior measurements required cleaning: the underlying search matches title words in any order, admitting senior titles into junior queries. Uncleaned, category junior growth reads +21.6%; cleaned, +5.3%.
- ▸A single employer, Bending Spoons, represents 7.9% of global junior AI postings. Excluding it, junior postings grew 14.5% rather than 5.3%. We disclose rather than exclude.
- ▸Salary data contains hourly rates mislabeled as annual and understates Indian compensation. Indian comp figures are order-of-magnitude only.
On continuity with Issue 1
Issue 1 reported a 31-category index totaling 14,934 postings. Those category definitions were not preserved in a directly reproducible form and we could not restate the total here. We have since recovered the full 31-category list — 18 technical, 13 non-technical, with postings, geography, seniority and salary — from the published Issue 1 record, and republish it with this issue alongside our exact query parameters. The series is comparable from Issue 1 forward. Issue 1's promised daily archival did not begin in Q2; it begins now.
Cross-verification
Q1 figures were re-queried on their original windows and reproduced within ±1–4%, confirming negligible backfill. 430 queries were logged with exact payloads for audit.
Disclosure
Navii sells career products to job seekers and hiring products to employers, and therefore has a commercial interest in this subject. Where our first-party data supports a favorable reading, we have sought external corroboration and reported it alongside. Where the data contradicted our expectations — AI is more open to entry-level candidates than we assumed, not less — we report that too. And where our own product falls short: our matching engine shows graduating students senior-level roles 34% of the time and internships only 6.1% of the time.
About & citation
The Navii Index is a quarterly research publication from Navii, the AI-powered career development platform. Each issue analyzes the labor market through a dual lens — what the data means for hiring leaders, and what it means for individual career builders. Series editor: Aakanksha Upadhyay, CEO and Co-founder of Navii.
When citing, please use: “The Navii Index — Q2 2026: The Class of 2026, Navii Research, heynavii.ai/the-navii-index/q2-2026-the-class-of-2026”
Contact research@heynavii.ai for dataset access, custom slices, or media inquiries.