Author affiliations
1 BMJ Global Health, London, UK
2 Global Surgery Advisors, New York City, NY, USA
Research in global health continues to evolve, and BMJ Global Health is committed to keeping pace. Year after year, the research submitted to our journal has increased in reach and rigour, all while embracing novel methodologies such as stochastic frontiers, interpretive phenomenological analysis, spatial modelling and quasi-experimental designs.
Simultaneously, the on-the-ground practice of global health continues the push for ever more equitable access to care, for testable hypotheses and for implementable solutions to shared challenges. Crucially, global health allows for deep interplay between research and practice. This cocreation of knowledge, in the service of prioritarian goals, is a hallmark of the field.
Meanwhile, the widespread, and ever expanding, use of large language models (LLMs) in daily life has spilled over into the world of scientific research. A marked increase in LLM-generated—and sometimes LLM-fabricated—manuscripts, reference lists and reviews has hit scientific journals.
BMJ Global Health is responding to these changes in three ways.
Introducing two new article types
First, we are excited to launch two new manuscript formats: Global Health in Practice and—limited only to papers that form part of a journal supplement—Essay.
The Global Health in Practice article type will allow for the concise description of on-the-ground implementation of global health programmes. The purpose of this article type is to give authors the opportunity to describe their experience in the implementation of global health interventions, including those specific to a community, subnational or national setting. Authors are encouraged to articulate a clear problem and describe their response to it with enough contextual detail and objectivity to allow the reader to assess whether the implementation might apply to similar problems elsewhere. The article should elaborate worthwhile conclusions that go beyond a statement that more research is needed.
These articles are aimed at a broad readership, one with an interest in and knowledge of global health, but without expertise in the specific implementation or setting in question. Jargon and specialised terminology should therefore be avoided, and references to organisations or practices that are specific to one country must be explained.
Because these articles are limited to 1500 words and 15 references, clear writing and an engaging presentation are essential. This means that conference proceedings and outcomes, or papers detailing framework development, are not currently of interest.
We are also introducing the Essay article type, exclusively for manuscripts that form part of a journal supplement. Essays will allow authors to synthesise the data and findings presented elsewhere in the supplement and to discuss their implications. This may include setting the supplement’s findings in the context of other research, presenting the framework underpinning the supplement or undertaking a broad evaluation of the evidence on the supplement’s topic.
As with Global Health in Practice, we look for an engaging style, an even-handed approach in evaluating evidence, a lucid line of argument and a worthwhile conclusion that goes beyond a call for more research. Essays are limited to 2000 words and 20 references, without a formal abstract.
Note that manuscripts that describe systematic synthesis of evidence—such as scoping reviews, systematic reviews and meta-analyses—continue to be classified as Original Research and should be submitted using that manuscript type. By contrast, the Essay offers supplement authors the opportunity to present an overview to a broad readership who may then want to explore the science within the rest of the supplement.
Three current article formats—Original Research, Commentary and invited Editorials—will remain unchanged…
The next section states that the journal is retiring two article types: the Practical article type and the Analysis article type…[not covered here].
Artificial intelligence in global health research
Finally, our response to the use of artificial intelligence (AI), and particularly LLMs, is evolving. Two recent events have highlighted the importance of such an evolution.
In April, the New England Journal of Medicine retracted an article containing a clinical image purporting to show bronchial casts retrieved from a patient in the setting of a forest fire. The image displayed clear signs of AI generation or manipulation.
A few weeks later, the Lancet published a report documenting a 10-fold increase in the prevalence of fabricated citations in biomedical manuscripts. In 2023, approximately 1 paper out of 2800 had a fabricated citation; by early 2026, this number had risen to 1 in 277. The authors only reviewed published manuscripts, meaning that the fabricated citations had made it all the way through the screening, editorial, review and publication process.
We at BMJ Global Health have also noted an increase in the use of LLM in submitted manuscripts. More worryingly, we have found increasing incidents of authors and reviewers denying LLM usage, even when explicit LLM artefacts have been left in situ.
LLMs must be used with extreme caution in the generation or review of science. They are sycophantic by design, which has the effect of reinforcing a user’s beliefs and incorrect assumptions. They also tend to generate fictional or so-called ‘hallucinated’ references, as well as medical errors and scientific fallacies, all of which puts the reliability of published work at risk.
At BMJ Global Health, we have a zero-tolerance policy for the undeclared use of LLMs in the creation of scientific research and commentary, or in the undertaking of peer and editorial review. In order to properly assess the validity of a piece of work, editors, peer reviewers and readers need to know that LLMs have been used. Failing to declare the use of LLMs distorts and abases the scientific process and borders on academic misconduct.
For this reason, BMJ Global Health requires authors and reviewers to declare and explicitly justify why an LLM was used instead of traditional methods. We will reject any paper or peer review in which undeclared use is detected and will report more egregious cases to the authors’ universities and home institutions.
Technology to automate the detection of LLM usage is improving, but it has not caught up with the usage itself. This means that any journal, ours included, is at risk of inadvertently publishing AI-fabricated papers or references. As we strive to improve our AI guardrails, we ask you, our readers, to alert us to AI-fabricated manuscripts and hallucinated references.
We look forward to continuing to receive your high-quality work in the pursuit of excellence, innovation and equity in global health research and practice.
[Editor’s Note: I approached the previous section on AI in global health research with trepidation. My lack of knowledge of the extent of influence of AI and of whether or not it is being allowed in any/many medical/research journals is that I hope it is being avoided like the plague. I was very glad indeed to learn that the BMJ Global Health is excluding it from what they are willing to publish. I hope everyone else follows their lead.]