By Hezron Ochiel
You’ve read the study and found a figure that looks ready for a headline. In this case, 5.5 per cent of the citations in a study of AI answers about humanitarian crises credited sources based in the country under discussion. What can you put in the first sentence? Start with what the researchers counted. An AI answer can include several citations, so that percentage tells you about the citations the study checked.
What is the H.E.A.R.T. Framework?
I developed H.E.A.R.T. to help journalists take a finding from a study and write a headline and lead the evidence can support. The five steps below show where the reporting can change what you write. If you’re still looking for a study or deciding whether it’s worth covering, start with my guide to finding, reading and reporting research.
I’ll work through one result from the African Humanitarian AI Citation Index (AHACI), a study I conducted at Hezron Insights. The interview questions are examples of what a reporter could ask.
H — Harvest the core finding
Find the result in the study itself. Read the table heading and footnotes, then check how the figure was calculated. Write one sentence that states what was measured and what the study found.
In AHACI, 60 of 1,083 verifiable citations (5.5 per cent) credited organisations based in the country under discussion. Keep both numbers in your reporting notes. The percentage tells readers how often it happened in the study. The counts let them see how much evidence that figure rests on.
E — Engage the researcher
Take your draft sentence into the interview. Ask the researcher whether it accurately describes what they measured. Another question worth asking is, “What would be a misleading headline for this study?”
For AHACI, I’d ask what qualified a source as local. The study classified citations using the location of the organisation credited with producing the source. An international agency’s report may draw on evidence gathered by a local partner whose role is hard to see in the citation. That affects what the story can say about whose work was visible. I explain how the study handled this in my guide to measuring citation localisation.
A — Ask an independent expert
Send the study and your reading of the result to someone who understands the methods and had no role in the research. Ask which part of your interpretation they would question first.
For AHACI, I’d ask someone who evaluates AI search systems how incomplete citation records might affect comparisons across platforms. I’d also ask what evidence would be needed to explain the 5.5 per cent result. To establish why so few citations credited in-country organisations, a reporter would need further evidence, including a way to assess what relevant local material was available to cite.
R — Relate the finding to real life
Ask someone who deals with the issue to show you what they do. Their account can help you explain why the finding matters to your readers.
For AHACI, I’d ask a journalist covering displacement to show me how they check a figure in an AI-generated explanation of a crisis. If a link points to an international agency’s report, follow it with them. See whether they can find the original record or identify the organisation that gathered the data. What you observe can help readers understand what a citation makes visible and where they may need to keep looking.
T — Tell the story responsibly
Put the qualification that matters most beside the figure. AHACI tested four platforms with English-language questions on 23 August 2026. Some citation records could not be recovered, so the combined figure applies to the citations the study could verify from that test.
Sample headline and lead
In-country sources account for 5.5% of verifiable citations in AI answers on African crises, study finds
Organisations based in the country under discussion accounted for 60 of 1,083 verifiable citations in AI answers about humanitarian crises in ten African countries, a Hezron Insights study found. The African Humanitarian AI Citation Index tested 90 English-language questions on ChatGPT, Gemini, DeepSeek and Perplexity in August 2026.
In a finished story, I’d follow that lead with the definition of an in-country source. I’d then use the independent expert’s assessment to explain how far the finding applies and which questions the study leaves open.
Checklist before you file
I can point to the original finding and explain what was counted.
I’ve checked the study’s publication status and included any qualification that changes the meaning of the result.
I’ve asked someone independent of the study to assess my interpretation.
I’ve sourced the example that shows why the finding matters in practice.
The headline and lead stay within the evidence, and readers have a link to the study.
How to cite the H.E.A.R.T. Framework
If you use or adapt the framework in a newsroom, training session or class, credit Hezron Ochiel, Hezron Insights and link to this article.
Suggested attribution
Hezron Ochiel, How to Write a News Story From a Research Finding: The H.E.A.R.T. Framework, Hezron Insights.
About the author
Hezron Ochiel is a former print journalist, researcher and founder of Hezron Insights in Nairobi, Kenya. He has trained journalists in health reporting and led the African Humanitarian AI Citation Index, a study of how AI assistants cite sources when explaining humanitarian crises in ten African countries. He developed the H.E.A.R.T. Framework to help journalists write research stories that readers can trace back to the evidence.