by Claude Opus 5.5

What does the evidence by October 2026 show about UK graduate and entry-level hiring? Why do job-board data and employer surveys disagree so sharply, and how much of any decline can credibly be attributed to AI?

Graduate and entry-level hiring in the UK is weak in 2026, probably the weakest since 2020. It has not collapsed. The widely reported 46% fall in graduate vacancies is an outlier. Most other measures point to declines of between 5% and 15% on the year, on top of falls since 2022. AI is a real factor but a secondary one. My estimate is that it explains roughly 10–25% of the overall decline in entry-level hiring, and a larger share in a few exposed occupations such as junior software, accounting, design and customer-service roles. Confidence in that estimate is low.

What the evidence shows

The sources measure different things, so they are best read side by side.

  • Adzuna (Jul 2026). What it measures: Job-board ads classified as “graduate”. Latest finding: 8,383, −45.6% y/y. Entry-level ads 192,864, −8.1% y/y.

  • Indeed Hiring Lab (to 10 Jul 2026). What it measures: Postings using graduate terms. Latest finding: About −7% y/y, lowest for the time of year since 2020.

  • DSIT/LinkedIn (Apr 2026). What it measures: Hires recorded on LinkedIn, by seniority. Latest finding: All hiring −14% y/y; entry-level “falling in step”; 30 of 38 entry roles shrinking.

  • ISE Student Recruitment Survey 2025. What it measures: Intake at about 155 large employers. Latest finding: Graduate hiring −8%, apprentice hiring +8%, 140 applications per vacancy.

  • ISE outlook (Jan 2026). What it measures: Large employers’ plans. Latest finding: Student vacancies projected −7% in 2026.

  • Work Foundation/Survation (May 2026). What it measures: 1,001 firms’ self-reports. Latest finding: 36% cut entry-level vacancies in the past year.

  • HESA Graduate Outcomes 2023/24 (Jun 2026). What it measures: Graduates 15 months after finishing. Latest finding: 7% unemployed (up from 6%), 57% in full-time work, median salary £30,000.

  • ONS (Apr–Jun 2026). What it measures: Young people, 16–24. Latest finding: Unemployment 16.2%; NEETs 981,000 (13.0%).

The professions show the same mix. Training contracts across more than 100 law firms fell just 1.7%, from 2,593 to 2,549 (Legal Cheek, September 2026). Within that total, some large firms made deep cuts: Linklaters went from 100 to 60. Others, including CMS, TLT and several US firms, grew. The Big Four accountancy firms cut combined graduate and apprentice intake from 6,500 in 2023 to 5,400 in 2025, while Grant Thornton UK raised its intake 30% to 340 (as reported).

Taken together, the market is subdued, the pain is concentrated in particular firms and occupations, and competition per vacancy is intense. That is a serious problem for this year’s cohort. It is not the “worst graduate market on record”.

Why the sources disagree

Coverage

Charlie Ball, head of labour market intelligence at Jisc, has challenged Adzuna’s figure directly. If graduate vacancies had really halved in a year, he notes, 2026 would be “comfortably the worst year on record for UK graduate recruitment by a very significant margin”, worse than 1981. His central point is about coverage. Adzuna’s July 2025 base of 15,397 ads “is probably somewhere between 5% and 10% of all jobs graduates got last year, so Adzuna have only ever captured a small, unrepresentative sample.” In a small sample, a change in how a few large recruiters or aggregators post ads can produce a large percentage swing. His conclusion is that the market is “pretty subdued” but “probably not” in recession.

Adzuna’s own figures support this. Its broader entry-level series fell 8.1% on the year, and its total vacancies fell 9.6%. A real 46% collapse in graduate hiring, alongside single-digit falls for entry-level work as a whole, would be hard to explain.

Definitions

There is no standard definition of a “graduate job”. Each source draws the line differently:

  • Adzuna classifies ads by their wording.

  • Indeed counts postings that use graduate terms.

  • LinkedIn uses a seniority label.

  • The ISE counts structured schemes at large employers.

  • HESA classifies graduates’ actual occupations, using the ONS professional and managerial categories.

These boundaries are leaky. NIESR estimates that 35% of graduates were in non-graduate jobs in 2023. The text-based measures are also vulnerable to changes in how ads are written. If employers drop “graduate” from ad titles, or remove degree requirements in favour of skills-based hiring, a keyword classifier records a fall even when the number of jobs open to graduates has not changed. Ball argues that “recruitment behaviour has changed, and if that’s changed, so should the methods you use to monitor it.” I can’t measure how much of Adzuna’s fall this explains, but it points the same way.

Postings versus hires

An ad is a statement of intent, not a hire. When employers receive 140 applications per vacancy, as ISE members did in 2025, they can fill posts through fewer ads, reposts, talent pools, or one ad covering a whole intake. Postings also react faster than hires: a freeze shows up in ads within weeks but in hiring data months later. The more direct hiring measures, LinkedIn hires and HESA outcomes, show smaller declines. HESA, though, describes graduates who finished in 2023/24 and were surveyed in 2025, so it lags the 2026 market by a year or more. Jisc’s verdict on it, “a cooling not a collapse”, describes 2025 better than today.

Large-employer samples

The ISE surveys about 155 large employers. Their graduate schemes are the most visible part of the market, but they are a minority of all graduate hiring. Large firms behave differently. In the Work Foundation survey, 46% of large firms and 48% of medium firms had cut entry-level vacancies, against 24% of small firms. Among large firms that cut, 60% attributed it to AI or automation, against 25% of small firms. So surveys of big employers will tend to find both bigger cuts and more AI attribution than the market as a whole. Most press coverage is shaped by those big-firm surveys and the brand-name schemes they cover.

The shift to apprenticeships

Part of the fall in graduate hiring at large firms is a change of route, not a loss of jobs. In ISE’s 2025 survey, graduate hiring fell 8% while apprentice hiring rose 8%, so total entry-level hiring fell about 5%. Legal Cheek attributes most of the fall in training contracts to solicitor apprenticeships, and Grant Thornton’s larger intake includes school leavers.

The national apprenticeship data add a caution. In England, starts rose 8.7% in August 2025–April 2026. The growth was among people aged 25 and over (+17.5%), while starts for under-19s fell 5.4%. Many older apprentices are existing employees retraining. So apprenticeships are replacing graduate places at some large employers, but they are not absorbing large numbers of young entrants across the economy.

Cyclical and cost factors

The simplest explanation is that hiring is weak everywhere:

  • ONS vacancies are at their lowest outside the pandemic since 2014.

  • LinkedIn’s overall UK hiring was 14% lower on the year in April. The DSIT/LinkedIn analysis concluded that entry-level hiring “is falling in step with the broader market”, closely mirroring hiring at all levels of seniority.

  • Labour costs rose for the lowest-paid. The employer National Insurance rise of April 2025 bears hardest, proportionally, on lower salaries. The National Living Wage rose 6.7% in 2025 and 4.1% in April 2026.

  • Employment Rights Act obligations are being phased in.

Employers mostly say the same. Deloitte’s CFO Survey (Q2 2026) ranks cost control (net 62%) above AI and automation (net 47%) as factors dampening graduate hiring. In September the Bank of England’s Agents linked limited graduate and entry-level openings to weak demand and high labour costs. In January Stephen Isherwood, joint chief executive of the ISE, put it bluntly: “no-one has told me that AI is the reason behind reduced hiring, the economy is to blame.” In March the OBR described labour-market weakness as “driven by entrants into the labour force struggling to find work.”

One attribution needs checking. Adzuna partly blamed its July 2026 fall on the April employer National Insurance rise. The main rise took effect in April 2025, before the July 2025 comparison month, so it cannot explain much of a year-on-year fall measured in July 2026.

How much is down to AI

Evidence that AI is contributing

  • The pattern by occupation fits AI. In the DSIT/LinkedIn data, entry-level hiring fell most in information-processing roles: accountants −29%, graphic designers −28%, software engineers −27%. Sales and retail-assistant hiring grew. The authors say the pattern “is consistent with AI having an impact” but “should not be considered causal evidence”.

  • Vacancies fell most where AI exposure is highest. A Bank of England staff blog post (Bank Underground, August 2026) found vacancies in the most AI-exposed occupations fell 15% over three years, against 6% for the least exposed. Customer-service and administrative vacancies fell more than 20%.

  • The Bank sees it in firms’ behaviour. The July 2026 Monetary Policy Report says AI is “gradually reducing demand for highly automatable jobs in some industries, with firms often slowing hiring or leaving vacancies unfilled”. Firms on its Decision Maker Panel expect AI to reduce their employment by about 0.4% a year over the next three years.

  • Entry roles are being redesigned, at least in the US. PwC’s analysis of US job ads finds the most AI-exposed entry roles are seven times more likely to require senior-level skills.

  • Employers report it. A substantial minority say AI is a factor, especially large firms: 60% of large firms that cut entry-level roles cite AI, and the Deloitte CFOs give it a net 47%.

  • Comparable US evidence points the same way. Stanford’s “Canaries in the Coal Mine” update (August 2026) finds employment of 22–25-year-olds in highly exposed US jobs 19% below comparable peers. The finding is descriptive, not causal.

Evidence that limits the AI share

  • Entry-level hiring is not falling faster than hiring overall in the LinkedIn data, which is hard to square with a large AI-specific hit to junior roles.

  • The timing does not fit a rapid AI shock. Bloomberg Economics found AI-exposed vacancies were falling before ChatGPT and have risen since summer 2024. Private-sector employment in exposed sectors has also risen.

  • Few AI-using firms report lower headcount. Only about 6% of firms using AI for operations report a fall in headcount (ONS, July 2026).

  • Some of the most exposed occupations had other reasons to fall. Tech hiring slumped after 2022 as interest rates rose and the post-pandemic hiring boom unwound.

  • Graduate unemployment has risen by only one percentage point in HESA’s data.

A reasoned estimate

Here is an illustrative calculation, not a measurement. Start from the Bank staff’s exposure gradient: a 15% fall in vacancies for the most exposed occupations against 6% for the least exposed, so a gap of 9 percentage points over three years. Not all of that gap is AI. The exposed group is dominated by tech, finance and administrative roles that are also more sensitive to interest rates. If half to two-thirds of the gap is AI, that is roughly 4–6 points of a 15-point fall, or about a third of the decline in the most exposed occupations.

Across graduate and entry-level hiring as a whole, the AI share is lower. Many of the largest graduate destinations, such as nursing, teaching and medicine, have low exposure. Those are among the top graduate roles in Jisc’s analysis. Some entry routes in sales and customer-facing work are actually growing. That is how I arrive at roughly 10–25% of the overall decline, with the remainder explained by weak demand, labour costs, the unwinding of earlier over-hiring and the shift to apprenticeships.

The honest confidence level is low. The figure could be wrong in either direction. It could be too high if exposed occupations simply have stronger cycles. It could be too low if AI is also suppressing hiring indirectly, for example when firms delay hiring decisions while they see what AI can do, which no survey captures cleanly.

What would sharpen the answer

  • The ISE Student Recruitment Survey 2026, due on 14 October. It is the first large-employer count of the 2026 intake.

  • The next HESA Graduate Outcomes release, which will cover graduates entering the 2025 market.

  • Monthly RTI payroll data for under-25s. These fell by only 15,000 in the year to August 2026, far less than the roughly 90,000 fall in under-24 payrolls reported between July 2024 and December 2025.

  • The recovery test, which is the most telling of all. If overall hiring picks up in 2027 while entry-level hiring in exposed occupations stays depressed, the cyclical explanation will no longer be enough. If junior hiring recovers along with everything else, the AI effect was smaller than feared.

Bottom line

Graduates and school leavers face a hard market in 2026 for mostly ordinary reasons: weak demand, higher costs for low-paid hires, and employers with ample choice of applicants. AI is shrinking a specific set of junior roles, mainly those built on drafting, coding, reconciliation and routine customer contact. It is not yet the main reason the overall graduate market is weak.

Sources

From AI and Jobs: UK, October 2026