Research · August 2026
African Humanitarian AI Citation Index 2026
Hezron Ochiel | Hezron Insights, Nairobi, Kenya
Only 5.5% of citations came from organisations based in the African country being discussed.
10
African countries
4
AI platforms
90
production questions
1,083
verifiable citations
Introduction
Whose knowledge becomes visible?
Humanitarian decisions depend on information: where people are displaced, which services have failed, who faces the greatest protection risk, whether food security is deteriorating and which actors can reach communities. Increasingly, some people seeking that information will not begin with a humanitarian dashboard, a country cluster page, or a search engine results page. They will ask a generative artificial intelligence system and receive a synthesised answer, often accompanied by citations.
That shift changes the politics of visibility. Generative systems select a smaller set of sources and present a synthesised narrative. Citation therefore becomes more than a technical reference mechanism: it determines which institutions are visible as credible knowledge producers.
The African Humanitarian AI Citation Index (AHACI) is a structured study of source selection across ten African countries, examining whose knowledge becomes visible in AI-generated answers about humanitarian crises.
Research at a glance
The study in numbers
Countries
10
AI platforms
ChatGPT, Gemini, DeepSeek and Perplexity
Production questions
90
Platform-question observations
360
Verifiable citations
1,083
Definition
What is citation localisation?
The study defines citation localisation as the degree to which an AI-generated answer about a humanitarian situation visibly attributes evidence to knowledge producers based in the affected country.
The measure concerns visible attribution, not the full provenance of the underlying evidence.
Key findings
01
Most visible evidence came from international sources.
Of 1,083 verifiable citations, 60 were country-local, versus 745 from UN/multilateral organisations.
02
Citation attention was concentrated.
The top five producers accounted for 40.3% of citations and the top ten for 58.8%. UNICEF, UNHCR, OCHA/Humanitarian Country Team products, OCHA and IOM were prominent.
03
Kenya stood out.
31 of the 60 country-local citations came from Kenyan sources, and Nigeria contributed nine. DRC and Sudan had none in the recoverable cross-platform dataset.
04
Local evidence mostly appeared alongside international sources.
84.8% of the 33 cited responses containing country-local sources also cited a UN or multilateral organisation. Only one response relied exclusively on country-local sources.
Why this matters
Citation visibility is part of localisation
AI-generated explanations can determine which humanitarian organisations and knowledge producers people encounter. Citation visibility is therefore an additional dimension of the localisation debate alongside funding, implementation and decision-making.
Methodology
How the study was conducted
AHACI began with a 1,000-question bank spanning ten African countries, ten humanitarian themes and ten question types. A randomised batch of 100 questions included ten pilot questions used to freeze the protocol; the remaining 90 were tested on the four platforms on 23 August 2026.
The resulting recoverable dataset yielded 1,083 answer-linked citations, coded by source producer, geography, publisher category, country-local status, multilateral status and hosting provenance.
Important limitations
Reading the results carefully
- The study measures visible citations, not the systems’ internal retrieval operations.
- Some platform citation records could not be recovered.
- All prompts were in English.
- There was no denominator of all available relevant local evidence, so the 5.5% share alone does not establish algorithmic bias.
- The complete ChatGPT subset showed 7.2% country-local and 70.9% UN/multilateral citations.
Main takeaway
The study argues for making credible local evidence discoverable, attributable and citable alongside established international sources; it does not advocate arbitrary citation quotas.
Download the research
Citation Localisation in Generative AI Answers
Whose Evidence Is Visible in African Humanitarian Crises? — full research article (PDF, 14 pages).
Suggested citation
Ochiel, H. (2026). Citation Localisation in Generative AI Answers: Whose Evidence Is Visible in African Humanitarian Crises? Hezron Insights, Nairobi, Kenya.
About the researcher
Hezron Ochiel
Hezron Ochiel is an award-winning strategic communications and public relations professional, researcher and founder of Hezron Insights. His work focuses on artificial intelligence, search, digital publishing, institutional visibility and knowledge discoverability, with particular attention to African organisations.
Research collaborations and media enquiries
For interviews, briefings on the findings, data-sharing collaborations or invitations to present the study, get in touch.