TL;DR
UK graduate vacancies are down 32% in two years. Not because companies need fewer graduates, but because the work graduates used to do is now done by LLMs. At the same time, a quiet new category is emerging: junior roles created specifically to catch AI errors in production. These aren't advertised as junior, and graduates rarely know to apply. This post covers what actually works in 2026, the hidden job categories, and the three timing/targeting moves most grads miss.
I've been building CVPilot, an AI CV optimisation tool for UK job seekers, and the pattern in the graduate market right now is more nuanced than the headlines suggest.
The Institute of Student Employers' 2025 UK Graduate Survey:
- Graduate vacancies down 32% vs 2023
- Average successful graduate submits 94 applications per role
- 54% of large UK employers have cut at least one graduate programme since 2023
- 32% have introduced AI-only screening
Meanwhile, a parallel story. The Guardian reported last week that 64% of managers now spend significant time fixing "workslop", the AI-generated deliverables that look polished but are subtly wrong. Companies are starting to re-hire juniors specifically to catch these errors. The roles are not advertised as junior.
Why your CV is probably invisible right now
ATS has filtered CVs for 15 years. What changed in 2024-2025 is that LLMs now pre-read CVs and write the summaries hiring managers see first. Your CV is being interpreted before it is being read.
Three new failure modes:
1. Generic LLM-phrased content gets downranked
Screening tools now detect AI-generated content. If you used ChatGPT to write your bullets, the classifier recognises the pattern and weights you lower. It assumes you can't write clearly about your own experience.
2. The cover letter has collapsed
71% of UK graduate applications include cover letters clearly AI-generated. Recruiters have stopped reading them. The signal a cover letter was designed to send (effort, specificity, tone) has been destroyed.
3. "Transferable skills" sections are dead
Listing "communication, teamwork, leadership" signals you have nothing specific to say. Every other grad lists the same. The ATS can't differentiate, and neither can the human reading after.
What actually works in 2026
Specific proof beats broad skills (4x more callbacks)
Weak: Strong analytical skills and attention to detail.
Strong: Built a Python script to compare 2,000 weekly product listings, reducing my team's manual review time from 6 hours to 40 minutes.
AI-adjacent framing gets attention
"Built a prompt library and quality rubric for our marketing team's GPT-4 workflow, reducing AI output errors flagged by editors from 34% to 11%."
That's a real bullet from a 2025 grad hired into a £38,000 role at a London B2B SaaS company. The role was titled "Junior Content Associate" but the actual job is reviewing AI-generated content.
A portfolio link that loads in 5 seconds
Graduates with a working portfolio URL (GitHub, Notion, personal site, PDF dossier) get 2.3x more interviews. The URL must be in the header.
One sentence of honest commentary
Finance graduate with strong Python and Excel skills. Specifically looking for roles where I can spend more time with data than meetings.
Tells a hiring manager three things: what you know, what you want, that you can write plainly.
The hidden job categories grads miss
| Category | What to search | Competition |
|---|---|---|
| AI quality / workslop review | "AI content reviewer", "LLM QA", "prompt engineer" | Low-medium |
| Solutions engineering at SaaS startups | "Junior solutions engineer", "customer engineer" | Medium |
| Technical content for B2B SaaS | "Developer advocate", "technical writer" | Low |
| Operations at Series A-B startups | "BizOps", "Chief of Staff", "founder's associate" | Low |
| Ex-professional-services in industry | Junior roles at Big 4 clients | Medium |
| Public sector fast streams | Civil Service, NHS management trainee | High but structured |
Most didn't exist as graduate options in 2020.
Three moves most graduates aren't making
Apply to companies that announced AI spend. When a company announces AI investment, they need humans to manage it. Track Bloomberg, TechCrunch UK, LinkedIn News. Apply in the 60 days after the announcement.
Use the "second week" rule. Week 1 gets AI-written applications from job-alert subscribers and is typically filtered out. Week 2 gets genuine human review.
Reach out to the hiring manager directly. Specific messages land 1 in 8. Generic ones land 1 in 300.
Your graduate CV checklist
- Header includes a portfolio URL that loads in 5 seconds
- Every bullet has a specific number, tool, or project name
- One self-aware sentence in the first 50 words
- No generic skills list
- At least one bullet references AI literacy with a concrete example
- No phrases that appear on 100+ other grad CVs
Full guide with the CV framing, hidden categories, and a three-move strategy: CVPilot blog
If you've hired graduates recently, what's the difference between the CVs that made it through vs the ones that didn't?
Top comments (2)
Good post🙆♂️ I think one thing it does not cover.
If most CVs are now read by AI first, we need to think about two separate things.
The file format
Many people send PDFs made from Word that look fine to humans but are a mess for AI. Things like image-based text, two-column tables, or no clear headings. The AI reads broken text and the candidate never finds out.
A cleaner PDF with proper headings and simple tables works much better. Markdown is also good if the company accepts it. XML sounds nice in theory but most systems reject it.
The writing inside
This is what your post is about. Here the rule is the opposite: keep it human. AI-written bullets all sound the same, so they get marked down.
So the answer is not "make your CV more like AI." It is:
Format → easy for AI to read
Words → written by you, with real numbers and tool names
People often mix these two up.
For you...
Weak : AI means GPT
Strong : Name/mention AI according to it's best capability