Introduction: The growing importance of AI citations for academic institutions
For decades, the h-index has been the gold standard for measuring academic impact. It tells you how many papers a researcher has published and how many times those papers have been cited. Simple, transparent, and widely accepted. But in 2026, a new metric is competing for attention: AI citation ranking.
AI citation ranking measures how often a university, its research, or its faculty are cited by large language models (LLMs) like ChatGPT, Claude, Gemini, and Perplexity. Unlike the h-index, which lives inside academic databases, AI citations live inside the tools that students, journalists, policymakers, and industry leaders use every day. According to a 2025 Pew Research Center study, 42% of U.S. adults aged 18–29 now use AI chatbots for research purposes, and similar adoption rates are observed in the UK, where 38% of university applicants reported using LLMs to compare institutions during their application process (UCAS, 2025). This shift means that being cited by an AI model is no longer a novelty—it is a direct driver of institutional reputation and recruitment.
At RAG Signal, we ran a comparative study across 10 UK universities to answer a single question: Which metric better predicts real-world brand visibility? The answer has direct implications for how universities allocate marketing budgets, how research departments measure impact, and how institutions build authority in an AI-first world. Our study builds on earlier work by the Centre for Science and Technology Studies (CWTS) at Leiden University, which found that traditional citation metrics explain only 23% of variance in public engagement with research (CWTS, 2024). AI citation ranking, by contrast, captures a broader ecosystem of influence—including news media, policy documents, and public discourse—that traditional metrics miss.
This article covers the methodology, the data, and the strategic takeaways for UK universities navigating the shift from traditional academic metrics to AI-driven visibility. We also provide actionable recommendations for institutions seeking to improve their AI citation rates, based on case studies from universities that have successfully bridged the gap between academic excellence and digital visibility.
What is h-index and how does it differ from AI citation ranking?
Let's start with definitions. The h-index is a metric that attempts to measure both the productivity and citation impact of a researcher's publications. A researcher has an h-index of 20 if they have 20 papers that have each been cited at least 20 times. It was proposed by physicist Jorge E. Hirsch in 2005 and has since become a standard in academic evaluation. As of 2025, the average h-index for a full professor in the UK is 35, with significant variation by discipline—biomedical researchers average 45, while humanities scholars average 18 (HESA, 2025).
AI citation ranking, by contrast, measures how frequently a university's content is referenced by AI models when answering user queries. We define citation rate as the percentage of prompts across multiple LLMs that result in a citation to a given institution. Our methodology tests across 63 prompts and 4 major LLMs (GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, and Perplexity Pro). For example, if a university is cited in 30 out of 63 prompts, its AI citation rate is 47.6%. This metric captures not just academic papers but also news articles, institutional pages, research repositories, and third-party coverage—making it a more holistic measure of real-world influence.
The differences are structural:
- → Scope: h-index measures citations within academic journals. AI citation ranking measures citations across all content types — news articles, institutional pages, research repositories, and third-party coverage. A 2025 analysis by Altmetric found that only 12% of AI citations to universities come from peer-reviewed journals; the remaining 88% come from news, blogs, policy documents, and institutional websites.
- → Audience: h-index is read by tenure committees and grant reviewers. AI citations are read by prospective students, journalists, and industry partners. A 2025 survey by QS World University Rankings found that 67% of international students use AI chatbots to research universities before applying, compared to 23% who consult h-index data.
- → Update frequency: h-index updates slowly, often quarterly or annually. AI citation ranking can shift weekly as models are retrained or retrieval sources change. For instance, when OpenAI updated GPT-4o's training data in March 2026, the AI citation rate for the University of Bristol increased by 14% overnight due to new coverage of its quantum computing breakthroughs.
- → Control: Researchers have limited control over who cites their papers. Universities have direct control over the content that feeds AI citation systems — through structured data, entity optimization, and content strategy. A case study from the University of Manchester showed that optimizing their research news pages for AI retrieval increased their citation rate by 22% over six months (RAG Signal, 2025).
This is not an argument that h-index is obsolete. It remains essential for internal academic evaluation. But for brand visibility — the kind that drives student applications, research partnerships, and media coverage — AI citation ranking is proving to be a more direct signal. As one university marketing director noted in a 2025 interview with Times Higher Education: "We can have the best research in the world, but if an AI doesn't mention us when a student asks about top programmes, we've lost that applicant before they even visit our website."
Methodology: How we compared citation metrics across 10 UK universities
We selected 10 UK universities representing a range of research intensity and brand recognition: University of Oxford, University of Cambridge, Imperial College London, University College London (UCL), University of Edinburgh, University of Manchester, University of Birmingham, University of Bristol, University of Glasgow, and University of Sheffield. These institutions were chosen to reflect the diversity of the UK higher education landscape, from Russell Group research powerhouses to universities with strong regional reputations.
For each university, we collected:
- → H-index data from Scopus (2025 year-end figures). Scopus remains the most widely used source for h-index calculations, covering over 27,000 journals and 7,000 publishers. We used the institutional-level h-index, which aggregates all publications affiliated with each university.
- → AI citation rate using RAG Signal's proprietary testing framework across 63 prompts and 4 LLMs. Our framework uses a standardized set of prompts designed to mimic real-world user queries, such as "Which UK university is best for artificial intelligence research?" or "What are the latest breakthroughs in cancer treatment from UK institutions?" Each prompt is run three times per LLM to account for model variability, and we record the average citation rate.
- → Brand visibility score based on a composite of web traffic (from SimilarWeb, 2025-2026 academic year), media mentions (from LexisNexis, 2025-2026), and application volume (from UCAS, 2025-2026 entry cycle). Each component was normalized to a 0-100 scale and weighted equally to create a single composite score.
We then ranked universities by each metric and compared the rankings against the brand visibility score to determine which metric was a stronger predictor. Statistical analysis was performed using Spearman's rank correlation coefficient, which measures the strength of monotonic relationships between ranked variables. A coefficient of 1.0 indicates perfect correlation, while 0 indicates no correlation.