Mapping the Evidence

1.1 The Funding Split

Only 48.6% of listings provide direct project funding. A significant 21.5% require students to enter university competitions, where funding is allocated at the faculty level rather than tied to a specific project. Although such competitions often provide full fee coverage and stipends, they introduce an additional layer of uncertainty for prospective candidates.

Furthermore, 23.4% of listings do not provide explicit information on funding arrangements, leaving applicants unable to assess the financial feasibility of pursuing the position. In a highly competitive field such as AI, this lack of transparency may deter candidates without independent financial support. It is also worth noting that, because of the LLM extraction method, not all funding details were captured — some listings may include funding information on their institutional pages.

1.2 Stipend Benchmarks

Adverts cluster heavily around the UKRI minimum of £20,780. Although the financial security of prospective postgraduate researchers is often downplayed by unclear funding statements, those that include financial details align with current UKRI guidelines. It is important to note that UKRI announced an increase of almost 5% in PhD stipends for students starting their studies in the academic year of 2026/27 — an announcement released just after our data collection, and therefore not captured in this dataset.

1.3 The Top-Up Gap

Beyond core funding, additional financial resources dedicated to training, conferences, and research dissemination serve as vital components of funding security. Only 30.3% of listings mention such top-ups, suggesting that opportunities for professional development support are often implicit. Only 715 listings included information on the financial top-up.

Industrial partner projects may offer up to £10,000 in additional stipend, while others provide only £1,000 per annum — inadequate given that leading AI conferences such as NeurIPS and ICLR charge over US$400 (~£300) for full-time student registration alone.

1.4 Institutional Disparities in Top-Up Availability

Top-up availability varies substantially across institutions. The University of Bristol sits near the middle of the distribution at approximately 22%, highlighting significant variation in additional research support across the doctoral landscape.

1.5 Ambiguity as a Filter

Among funded listings, 84.2% (962 listings) include explicit statements of secured funding. However, 172 listings remain vague and 9 are conditional — deterring applicants without institutional insider knowledge of university funding mechanisms. Some positions that appear funded may in fact entail an additional selection process.

1.6 Who Pays for AI PhDs?

The most common contributors are host Universities (630 listings) and UKRI Research Councils (604 listings). The Engineering and Physical Sciences Research Council (EPSRC) accounts for approximately 9.6% of the analysed PhD listings. Direct industry and charity contributions remain comparatively low, as illustrated in the chart above.

2.1 The Advertised Openness

A high degree of advertised openness exists: 2,262 listings do not explicitly restrict applications to UK nationals. However, this surface openness conceals a deep structural barrier driven by national UK Research and Innovation (UKRI) policies.

While UKRI rules permit up to 30% of a cohort to be international students, the financial award is capped at the domestic fee rate. Unless the host university agrees to absorb the difference, international applicants are legally permitted to apply but are systematically priced out. Ultimately, this relies on omission in academic advertising, leaving applicants to navigate complex funding caveats entirely on their own.

2.2 The International Paywall

The stacked bar chart above illustrates the "false hope" funding trap — comparing fully funded positions against the volume of projects that invite international applicants but leave them to cover significant fee shortfalls. This pattern is not unique to any one institution; it is a systemic feature of the UK AI doctoral market, reinforcing geographical barriers of the broader AI ecosystem.

3.1 Softening Entry Standards

The overwhelming dominance of the unspecified degree requirement (grey bar) and the broad acceptance of 2:1 degrees illustrates a broad institutional shift. Only a small minority of listings explicitly demand First-Class honours, suggesting that formal grade-based gatekeeping is declining across the sector.

We must be cautious, however: the absence of strict formal requirements does not necessarily reflect informal shortlisting practices at the institutional level, including supervisory preferences or competitive dynamics among applicants.

3.2 Limited Alternative Pathways

Beyond degree classifications, alternative routes into doctoral study remain limited. Only around 35% of listings explicitly recognise equivalent practical or professional experience as a valid basis for assessing research suitability.

Despite some movement towards flexibility in formal requirements, academic credentials continue to dominate as the primary mechanism for evaluating candidates. The majority of listings require a degree only, with no recognition of professional equivalence.

4.1 Prestige Signalling

Elitist keywords — such as references to "world-class" institutions or "leading" research environments — are relatively uncommon overall. This suggests that symbolic signalling of prestige is not a dominant feature of AI PhD listings, though a small subset of universities stand out with notably higher average buzzwords per listing.

4.2 The Networking Hurdle

A subset of universities requires students to contact supervisors before applying. While often framed as early engagement, such requirements can create a barrier for candidates with limited prior academic networks or confidence in approaching senior academics — functioning as an additional gatekeeping mechanism for those without social capital.

4.3 Passive Inclusion Dominates

We examined the presence and framing of diversity initiatives within PhD listings, distinguishing between active diversity language (concrete measures such as guaranteed interviews or targeted bursaries) and passive diversity language (generic encouragements with no actionable support).

The vast majority of listings contain no diversity statement. Of those that do, passive Tier B language far outweighs active Tier A provisions. This absence is particularly striking given the growing institutional emphasis on equality, diversity and inclusion within UK higher education.

5.1 The Visible "Gender Void"

The market is dominated by presumed male-only supervisory teams, accounting for the largest single category. Research confirms that the lack of visible female role models significantly lowers the AI self-efficacy of female applicants, potentially causing them to self-select out of the field before they even apply.

This finding aligns with broader gender imbalances in STEM subjects (e.g. Frachtenberg & Kaner, 2022; Kalim et al., 2025; MIOIR, 2025; Stathoulopoulos et al., 2019) and reinforces the geographical barriers of the broader AI ecosystem.

6.1 Disciplinary Openness

Contrary to common assumptions, AI PhD programmes are relatively open in disciplinary framing. Only 96 openings were classified as exhibiting explicit "disciplinary tunnel vision," where eligibility was restricted to a narrow set of technical degrees. Most projects appear open on paper, yet they filter candidates through implied technical competencies embedded within project descriptions.

6.2 The Python Barrier

Programming languages, most notably Python, are the most commonly required technical prerequisites alongside references to machine learning frameworks and computational methodologies. This suggests a shift from formal credential-based gatekeeping to skill-based filtering.

Rather than excluding applicants through narrow degree requirements, listings signal technical competencies. Such signals may function as implicit barriers, particularly for applicants from non-technical or interdisciplinary backgrounds who may meet formal eligibility criteria but lack confidence in applying without demonstrable coding experience.