UK graduate employers now receive an average of 140 applications per vacancy, yet only around 8% configure their ATS to auto-reject CVs based on content. This data analysis examines what AI screening actually does during the spring recruitment cycle and where the graduate labour market is heading.
When the Institute of Student Employers (ISE) released its 2025 Student Recruitment Survey, one figure captured the state of UK graduate hiring: 140 applications for every single vacancy. That ratio, drawn from 155 employer members who collectively received over 1.8 million applications for roughly 31,000 early careers roles, represents a two-year high. The surge is partly a consequence of AI-powered tools and one-click apply features that have made mass applications nearly frictionless for candidates.
At the same time, the supply of graduate positions contracted. The ISE survey recorded an 8% decline in graduate hiring during 2024/25, the weakest year for graduate recruitment since the pandemic-era 12% drop in 2020. Of the employers surveyed, 42% reduced their graduate intake, while 25% held steady and 33% reported an increase. The outlook remains cautious: ISE data points to a further 7% reduction in graduate hiring for the 2025/26 cycle, driven primarily by sharp declines at a small number of large employers.
The Higher Education Statistics Agency (HESA), drawing on its Graduate Outcomes survey of 917,610 graduates from the 2022/23 cohort, reports that 82% of respondents were in employment or unpaid work 15 months after graduation. However, full-time employment for first-degree graduates fell from 57% to 54%, while the unemployment rate rose by one percentage point to 6%. These are modest shifts in isolation, but they reflect a broader cooling in the graduate labour market that makes understanding screening systems all the more relevant.
The mechanics of automated CV screening are frequently misunderstood, and the confusion begins with one widely cited statistic: that 75% of CVs are rejected by applicant tracking systems before a human ever reads them. This claim, which circulates across career advice platforms and social media, has been traced by analysts to Preptel, a now-defunct recruiting services company. No methodology for the figure was ever publicly disclosed. Recruiter survey data paints a substantially different picture.
A 2025 survey of 25 recruiters published by Enhancv found that 92% manually review applications, even in high-volume scenarios. Only 8% (two out of 25 respondents) reported configuring their ATS to automatically reject CVs based on content match scores. Separately, HR.com published similar findings in late 2025, reporting that the vast majority of recruiters use filtering functions to prioritise and sort applications rather than to eliminate them outright. The real bottleneck, according to these surveys, is human bandwidth. When a recruiter faces 140 applications for a single role, even those who review every submission will inevitably spend limited time on each one.
That said, virtually all employers do use what the industry terms "knockout questions": mandatory fields such as right-to-work status, minimum qualification level, or willingness to relocate. Candidates who do not meet these hard criteria are typically filtered out before human review. This is a binary compliance check, not an assessment of CV quality.
The technical architecture of screening software has evolved considerably. Earlier generations of ATS relied on straightforward keyword matching, scanning CVs for exact terms from the job description. According to industry analysis, the current generation of tools, including platforms such as Workday, Greenhouse, Lever, and iCIMS, increasingly uses natural language processing (NLP) and, in some cases, large language models (LLMs) to assess context rather than simply counting keyword frequency.
In practical terms, this means that a skills statement such as "developed automation scripts in Python to reduce manual processing time by 40%" is likely to carry more analytical weight than simply listing "Python" in a skills section. The system can parse the verb, the tool, and the outcome, recognising a demonstrated competency rather than a standalone keyword. UK-specific platforms are also part of this landscape: Trac is widely used across NHS trusts, SuccessFactors is common in banking, and Oleeo handles much of the Civil Service's graduate recruitment pipeline.
Even so, advanced NLP systems depend on being able to read and parse document content accurately. Formatting choices that prevent clean parsing, such as embedded text boxes, multi-column layouts, or image-based headers, can still cause information loss regardless of how sophisticated the underlying AI may be.
This analysis draws primarily on three data sources. The ISE Student Recruitment Survey 2025 covers 155 employer members across a range of sectors, with data on over 31,000 early careers roles and more than 1.8 million applications. It is the most widely cited benchmark for UK graduate recruitment trends, though it skews toward larger, structured employers and may not fully represent SME hiring patterns.
The HESA Graduate Outcomes survey captures responses from 917,610 graduates of the 2022/23 academic year, surveyed between December 2023 and November 2024. As of mid-2025, this dataset has been embedded in the Department for Education's Longitudinal Educational Outcomes (LEO) data, marking its integration into official government statistics.
Labour market context comes from the Office for National Statistics (ONS), which reported 2.0 unemployed people for every job opening in the period from December 2024 to February 2025. ATS-specific data draws on recruiter surveys published by Enhancv, HR.com, and Select Software Reviews, with caveats around sample size noted in the limitations section below.
The UK graduate recruitment calendar typically follows a pattern that peaks in two main windows. According to data from TargetJobs and GRB, the autumn cycle (September to November) tends to capture the largest structured graduate schemes, particularly in sectors such as finance, consulting, law, and engineering. The spring cycle, running from approximately March through June, represents a second significant wave of hiring activity.
During the spring window, assessment centres are generally held in March and April, with job offers typically extending through April, May, and June. Direct entry graduate roles, as distinct from structured graduate schemes, tend to be advertised from around Easter onward. Some structured schemes that did not fill all positions in the autumn cycle may also re-open applications during this period.
For international graduates in the UK, the spring cycle holds particular significance. Candidates who completed their studies in the preceding academic year may be operating within time-sensitive post-study frameworks, and the spring window often represents one of the final major hiring surges before summer. Comparable spring hiring patterns exist across European markets; spring recruitment cycles in France, for instance, follow a broadly similar timeline for early-careers roles.
Competition levels during the spring cycle can vary considerably by sector. According to ISE data, technology, finance, and professional services roles tend to attract the highest application-to-vacancy ratios, while public sector, education, and charity roles may see lower but still substantial volumes.
Given that the primary function of ATS software is to parse and organise application data, document formatting is a practical concern rather than a cosmetic one. According to ATS optimisation guides from Jobscan and Resume.io, .docx (Microsoft Word) remains the most universally parseable file format across major platforms. While many modern systems handle text-based PDFs without difficulty, some older or sector-specific systems may struggle with PDF files that contain multiple columns or graphical elements.
Standard section headings, such as "Experience," "Education," "Skills," and "Contact Information," tend to parse more reliably than creative alternatives. ATS platforms are generally configured to recognise these conventional labels, and recruiters reviewing parsed data also tend to navigate structured CVs more efficiently. Single-column layouts, standard fonts (Arial, Calibri, or Times New Roman at 10 to 12 points), and the avoidance of text boxes, tables, and embedded images are commonly cited as formatting practices that support clean parsing.
The shift from pure keyword matching to semantic analysis has implications for how skills are presented on a CV. Industry guidance from multiple ATS vendors suggests that contextualised skill descriptions, those that pair a verb with a tool or competency and a measurable outcome, tend to carry more weight in modern screening than isolated keyword lists.
A common recommendation across career services and ATS optimisation platforms is to include both acronyms and their full forms (for example, "SEO" alongside "Search Engine Optimisation") to account for variation in how search queries may be configured. For candidates holding technical certifications, listing both the credential name and the issuing body can serve a similar function in improving discoverability.
It is worth noting that the distinction between ATS screening and human screening is often less clear-cut than popular career advice suggests. In many organisations, ATS software ranks and sorts applications rather than making binary decisions. The recruiter then reviews a prioritised list. Framing a CV to communicate clearly with both the parsing software and the human reader is, in practice, the same exercise: clarity, specificity, and relevant evidence of skills and accomplishments.
Understanding where demand and compensation concentrate can add strategic context to the spring cycle. According to ISE data from 2025, the average starting salary on structured graduate schemes is approximately £35,170, though this figure reflects larger employers and formal programmes. HESA's broader dataset, which encompasses graduates in a wider range of employment types, puts the average closer to £28,731.
Sector-level variation is substantial. As reported across multiple industry surveys in early 2026, approximate median starting salaries include the following:
Regional variation adds another dimension. London-based roles typically offer starting salaries in the range of £32,000 to £34,000, while positions in Scotland and the South East cluster around £28,000 to £29,000. These figures are broadly consistent across multiple data sources, though precise medians vary depending on survey methodology and sample composition.
For candidates weighing opportunities across borders, salary figures gain additional context when adjusted for cost of living. A £34,000 starting salary in London represents a different standard of living than an equivalent nominal figure in a lower-cost city. The same principle applies when comparing UK graduate salaries with opportunities in markets such as the Gulf region, where sectors tied to national development strategies may offer distinct compensation structures.
Several converging trends suggest that AI's role in graduate screening will expand substantially over the coming years. According to ISE data from 2025, 62% of surveyed employers expect to use AI in recruitment within five years, and 70% anticipate greater automation in their hiring processes overall. Among those already using AI in selection, 94% report improved speed and efficiency, and 81% cite enhanced capacity to analyse large data volumes.
At the same time, candidate use of generative AI is creating new tensions. ISE reports that 61% of employers have caught or suspected candidates using AI during interviews without permission, yet 45% have not provided applicants with any guidance on what constitutes appropriate AI use. This policy gap is likely to narrow as employers increasingly formalise their expectations around AI in selection processes.
The broader trajectory points toward a graduate recruitment landscape in which both sides of the hiring process are increasingly AI-mediated. Candidates use AI to generate and optimise applications; employers deploy AI to parse, rank, and evaluate them. How this dynamic evolves, particularly with regard to fairness, transparency, and effectiveness, remains one of the most closely watched questions in labour market analytics.
Candidates who successfully navigate initial screening will still encounter human-led assessment stages. Behavioural interview preparation, for example, remains a distinct competency that no amount of CV optimisation can replace.
Several important caveats apply to the data presented in this analysis. The ISE survey, while authoritative, covers 155 employer members and skews toward larger, more structured organisations. SMEs, which collectively employ a significant share of UK graduates, are underrepresented. The 140 applications-per-vacancy figure may be higher or lower at employers outside the ISE sample.
ATS-specific data, particularly the recruiter surveys on auto-rejection rates, relies on small sample sizes (as few as 25 respondents in some cases) and self-reported behaviour. Actual ATS configurations at individual employers may differ from what recruiters report in surveys.
HESA Graduate Outcomes data captures employment status at 15 months post-graduation. It does not track career progression, job quality, or whether graduates are in roles related to their field of study. The 6% unemployment figure, while useful as a benchmark, does not capture underemployment or precarious work arrangements.
Salary figures vary across sources due to differences in methodology, sample composition, and the inclusion or exclusion of London weighting. Readers are encouraged to consult primary sources directly and to seek guidance from qualified career professionals for analysis tailored to individual circumstances.
Written By
Labour Market Reporter
Labour market reporter covering data-driven job market analysis, employment trends, and salary benchmarking worldwide.
Content Disclosure
This article was created using state-of-the-art AI models with human editorial oversight. It is intended for informational and entertainment purposes only and does not constitute legal, immigration, or financial advice. Always consult a qualified immigration lawyer or career professional for your specific situation. Learn more about our process.
Singapore's Q2 2026 tech hiring market favours specific CV formats depending on candidate profiles and recruiter expectations. This comparison breaks down when each approach works, where it falls short, and how to navigate ATS filters in one of Asia's most competitive talent markets.
Bangalore's tech sector increasingly relies on AI screening to filter candidates. This guide explores how evidence-based resume structures, market-specific formatting, and keyword optimization align with ATS expectations in India's largest tech hub.
International job seekers targeting Belgium often face CV rejection over preventable formatting errors. This guide reports on the most common formatting pitfalls in Belgian job applications and how proactive auditing can reduce the risk of automated and human screening failures.