AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: AI’s Place In The Graduate Job Debate: What Data Shows And Omits on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

UK online job adverts fell more in occupations rated highly exposed to AI than in less-exposed roles, while Australian graduate postings also weakened. The figures show a pattern, not proof that AI caused the declines; other economic and workplace changes may also be responsible.

Job adverts have fallen more sharply in UK occupations rated highly exposed to AI than in less-exposed roles, while Australian graduate postings also declined in 2025, according to research and job-ad data reviewed by ThorstenMeyerAI.com. The figures point to pressure on some early-career opportunities, but do not show that AI caused the declines; researchers cite several competing explanations.

A Bank of England staff analysis by Haley Schlicht compared occupational AI-exposure scores with UK online job adverts from the Office for National Statistics. It found adverts fell 15% in the most-exposed third of occupations, compared with 10% in the middle third and 6% in the least-exposed third. Customer service adverts fell 23%, and administrative adverts fell 22%. The figures describe changes in adverts, not confirmed job losses or hiring decisions.

Schlicht wrote that evidence of AI-driven change in hiring was “proving considerably more tenuous than the headlines suggest” and said “confident attribution remains premature.” Pandemic-era over-hiring, a weaker economic cycle, remote work, energy costs and higher labour costs could also affect recruitment. The ONS estimated 691,000 UK vacancies in July to September 2026, down 21,000, or 2.9%, from the previous quarter. It said smaller firms may be holding back recruitment because of higher labour costs.

In Australia, graduate job postings fell almost 15% in 2025 and declined more in AI-exposed jobs, according to the source material’s account of Indeed data. The decline reportedly began in 2023, before wide adoption of generative AI. Apprenticeship postings, described as having little AI exposure, fell 46% from their peak. The source does not provide the underlying series or the comparison period for that peak, so these figures cannot by themselves identify a cause.

At a glance
reportWhen: Published November 17, 2026; draws on U…
The developmentA review of UK and Australian job-ad data finds steeper declines in AI-exposed and graduate hiring, while the available evidence does not establish AI as the cause.

AI and the entry level · explainer

From career pyramid to diamond

Why the first rung of the ladder is where AI could show up first, and why the cause is not settled yet.

A workforce by level (schematic)

Senior Junior AI takes tasks

Pyramid: many juniors at the basedoing scheduling, correspondence and routine information processing UK job adverts: −15% vs −6%in the most vs the least AI-exposed third of occupations Diamond: hollow at the base“centred on experienced staff”

How the squeeze could work

  1. First jobsRoutine tasks: scheduling, correspondence, routine information processingthe base of the pyramid
  2. AI steps inSenior workers can do more of the work that once formed the bedrock of junior hiringBank Underground
  3. Fewer advertsUK online job adverts fell 15% in the most AI-exposed third, 6% in the leastBank Underground, ONS data
  4. DiamondThe pyramid may turn into a diamond, hollow at the baseBank Underground
  5. Not settledPandemic over-hiring, the business cycle, remote work and labour costs hit at the same time“confident attribution remains premature”
The open question for employers: if the base hollows out, who becomes the experienced staff of the next decade?

Source: Haley Schlicht, Bank of England, Bank Underground blog, Aug 6, 2026 (staff view, not necessarily the Bank’s), based on ONS online job adverts through early 2026; the pyramid-to-diamond idea is from research the post cites. The blocks are a schematic, not data. The question in the orange box is ours.

Graphic: Powered by Thorsten Meyer AI

Pressure on the First Career Step

Entry-level jobs can give graduates the routine tasks and supervised practice needed to build professional judgement. If employers reduce those roles, the effect could reach beyond a smaller number of openings: businesses may have fewer workers developing the experience required for more senior positions later. That is a plausible concern, not an outcome established by the figures reviewed here.

The pattern matters to students, universities and employers because it raises questions about what new hires are expected to contribute. A commentary by law and technology lecturer Armin D Ashenberg argues that graduates will need to judge whether AI output is good, identify its shortcomings and improve it. The available job-ad data, however, does not show whether employers are replacing graduates with AI, changing the skills they seek, or simply hiring less across the board.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

Warnings Meet Uneven Evidence

The data review follows two warnings discussed in the source material. On November 15, Ashenberg wrote in The Conversation about how AI may affect graduate employment. Two days earlier, The Telegraph published a report attributing a warning about “very painful” job losses to the head of the UK Office for Budget Responsibility. The article was behind a paywall, and the source material says it found no OBR or parliamentary record of the remarks. The warning should therefore be treated as reported by The Telegraph, not independently confirmed.

The OBR’s October 2026 Economic and fiscal outlook offers a more qualified assessment. Its central forecast is that AI could lift productivity growth by 0.2 percentage points a year by the medium term, while stating that the scale and timing of the effect are “highly uncertain.” The report, citing the ONS, said 30% of businesses reported using at least one AI technology in September 2026, up from 15% two years earlier. It said it would monitor emerging evidence on productivity and employment.

““Confident attribution remains premature.””

— Haley Schlicht, Bank of England Data and Statistics Division

AI’s Role Is Not Isolated

The main uncertainty is causation. Job adverts are weaker in more AI-exposed occupations, but the Bank of England analysis says several forces changed at the same time. The data cited does not establish how much of the decline, if any, is attributable to employers adopting AI rather than to economic conditions, labour costs or other shifts.

Important details also remain unavailable. The Australian figures are not accompanied here by the full underlying dataset, definitions or comparison windows. The OBR chief’s reported warning could not be checked against a speech, transcript or parliamentary record in the source material. Nor do the figures establish whether fewer adverts translate into fewer hires, whether jobs are being redesigned, or how graduates’ prospects vary by field, region and employer size.

More Hiring Data Will Test the Pattern

The OBR says it will monitor emerging evidence on productivity and employment as it prepares future forecasts. Further ONS vacancy releases and updated job-ad series can show whether the differences between high- and low-exposure occupations persist, widen or narrow. Those figures will still need to be read alongside employer hiring, workforce and AI-adoption data to clarify what is driving the changes.

Universities and employers are also likely to face continued debate over graduate preparation. Ashenberg’s argument is that teaching should include evaluating and improving AI output, not only learning to use tools. Whether that approach changes hiring outcomes remains untested in the figures cited. For now, the evidence supports a narrower conclusion: early-career hiring is under pressure in some exposed roles, but AI’s share of that pressure is unknown.

Key Questions

Does the data prove AI is causing graduate job losses?

No. The figures show steeper advert declines in some AI-exposed occupations, but the Bank of England analysis says confident attribution is premature. It identifies other possible influences, including economic conditions and labour costs.

How much did UK job adverts fall in AI-exposed occupations?

In the Bank of England staff analysis, online adverts fell 15% in the most AI-exposed third of occupations, compared with 10% in the middle third and 6% in the least-exposed third. These are advert changes, not a count of jobs lost.

What happened to Australian graduate job postings?

The source material reports that postings fell almost 15% in 2025, with larger declines in AI-exposed jobs. It says the fall began in 2023, before wide AI adoption, and does not provide enough underlying detail to establish the cause.

Has the OBR confirmed the reported warning about painful job losses?

Not in the records described by the source material. The warning was attributed to the OBR chief in a Telegraph report, but no OBR or parliamentary record was found there. The OBR’s published October forecast instead describes AI’s effects as highly uncertain.

Source: ThorstenMeyerAI.com

You May Also Like

Why Multi-Step Forms Lead to Three Times More Completions

Discover how breaking your form into multiple steps can triple your completion rates. Simple tips to improve user experience and get more leads.

AI’s Next Big Moves In 2026 You Should Follow

Explore the confirmed AI developments in 2026, why they matter, and what remains uncertain as the industry advances rapidly this year.

The Compounding Error Problem — Why 99.9% Alignment Decays to 60% in 500 Generations

Analysis of how 99.9% alignment accuracy degrades to 60% after 500 generations, highlighting risks in recursive AI self-improvement.

IdeaClyst: The Validation Council

IdeaClyst introduces a new multi-model council to rigorously evaluate ideas before inclusion in roadmaps, enhancing decision quality and reducing costly errors.