For decades, Higher Education Academy (HEA) Fellowship has been presented as evidence of effective university teaching. Whether achieved through an institutional programme or direct application to Advance HE, the process has traditionally centred on one activity: writing a reflective account demonstrating alignment with the Professional Standards Framework (PSF).

That model made sense in a world where reflective writing was assumed to be an authentic expression of professional judgement.

The arrival of generative AI calls that assumption into question.

This is not primarily a debate about academic misconduct. Nor is it simply another chapter in the long history of plagiarism detection. The more interesting question is whether a carefully constructed written portfolio can still serve as reliable evidence of teaching capability when large language models can generate fluent, coherent reflections that satisfy many of the formal expectations of assessment.

Increasingly, the higher education sector appears to be concluding that it cannot.

Reflection was always a proxy

The purpose of an HEA Fellowship application has never been to reward elegant writing. Written reflection has instead acted as a proxy—a practical way of inferring deeper qualities such as pedagogical judgement, professional values and scholarly engagement.

The problem is that proxies only remain useful while they continue to measure what they were designed to measure.

Generative AI exposes an uncomfortable weakness. If a language model can produce a reflective narrative that aligns convincingly with the Professional Standards Framework, then assessors face a difficult question. Are they evaluating genuine professional understanding, or simply a well-crafted piece of text?

Stephen Wheeler has argued more broadly that AI has not broken assessment so much as revealed weaknesses that were already there. Technology, he suggests, simply exposes poor assessment design rather than creating it.

That observation extends well beyond Fellowship applications.

The emergence of "Synthetic Credentialism"

One of the more interesting ideas beginning to appear in discussions of AI and assessment is what some researchers describe as Synthetic Credentialism.

Although the terminology is still emerging and has yet to become established within mainstream educational literature, the underlying concern is becoming increasingly familiar. The argument is that AI-assisted production may allow learners to obtain qualifications or professional recognition that no longer provide dependable evidence of the competencies those credentials are intended to represent.

Viewed in this way, the problem is structural rather than individual.

The issue is not whether one candidate used ChatGPT.

It is whether institutions can continue treating static written artefacts as convincing evidence of professional capability when sophisticated AI systems can generate those artefacts with increasing ease.

If that assumption no longer holds, then the validity of the assessment itself—not simply student behaviour—deserves scrutiny.

Universities are beginning to redesign assessment

Interestingly, parts of the sector have already started responding.

One notable example comes from the Journal of Perspectives in Applied Academic Practice, where Daniel Cole describes the replacement of traditional written portfolios within a Postgraduate Certificate in Academic Practice with assessed Professional Conversations. Rather than relying solely on written reflection, candidates discuss and defend their practice in structured dialogue with experienced assessors. The paper argues that this produces richer evidence of professional understanding while offering greater resilience in an AI-enabled environment.

This is not a minor procedural adjustment.

It represents a shift in what counts as evidence.

Conversation allows assessors to explore uncertainty, challenge assumptions, ask follow-up questions and probe the reasoning behind decisions. Those qualities are considerably harder to simulate than a polished written narrative.

The same direction of travel is increasingly visible elsewhere across higher education.

Recent commentary from the Higher Education Policy Institute (HEPI) argues that AI has exposed the long-standing gap between assessment outputs and the capabilities institutions actually wish to certify. Rather than investing ever more heavily in AI detection, the emphasis is shifting towards assessment redesign, with greater use of oral examination, iterative project work and authentic demonstrations of judgement.

That represents a philosophical change as much as a technical one.

Constructive alignment may no longer be sufficient

Much of contemporary university teaching rests upon ideas developed during the late twentieth century.

Constructive alignment, learning outcomes and reflective practice remain enormously influential, and for good reason. They brought coherence to programme design and assessment.

The difficulty is that many of these frameworks evolved before anyone imagined systems capable of producing graduate-level reflective prose in seconds.

The assumptions underpinning those models deserve renewed examination.

Several educational developers have argued that assessment should increasingly value process, transparency and the demonstration of judgement rather than polished final products alone. The emphasis moves from What was submitted? towards How did the candidate think? and Can they explain and defend their reasoning?

That distinction matters.

Teaching has always involved more than producing convincing documentation.

Fellowship still matters—but perhaps not in its current form

None of this suggests that HEA Fellowship has become worthless.

Professional recognition remains important. Universities need ways of recognising educational expertise, encouraging scholarly teaching and supporting career development.

The question is whether the mechanism for demonstrating that expertise remains fit for purpose.

A written reflective portfolio was always an indirect measure of teaching quality. AI has simply made that indirectness impossible to ignore.

If the sector increasingly adopts professional conversations, teaching observations, recorded practice, viva-style discussion and longitudinal evidence of development, Fellowship may ultimately become more—not less—credible.

Ironically, generative AI may strengthen professional recognition by forcing universities to gather better evidence.

A broader lesson

The debate surrounding Fellowship applications reflects a much larger issue facing higher education.

For decades universities have relied on written artefacts as stand-ins for complex human capabilities. Essays stood in for understanding. Reflective portfolios stood in for professional judgement. Reports stood in for critical thinking.

Generative AI has exposed the fragility of those assumptions.

The real challenge is therefore not learning how to detect AI.

It is deciding what kinds of evidence genuinely demonstrate that someone can think, teach, question and exercise sound professional judgement.

That is a far more difficult problem.

It is also a far more interesting one.

References

Cole, D. (2023). Authentic Assessment Through Professional Conversations: An AI Friendly Assessment Method? Journal of Perspectives in Applied Academic Practice, 11(3). https://doi.org/10.56433/jpaap.v11i3.586.

Higher Education Policy Institute (2025–2026). Assessment reform and generative AI commentary, including What Generative AI Reveals About Assessment Reform in Higher Education and Verifiable Judgment: What AI Actually Demands of Universities.

Wheeler, S. (2025). Reimagining Assessment in Practice. Personal blog on digital pedagogy and assessment redesign.