Research Institute for Tropical Medicine

Large language models (LLMs) are increasingly embedded directly into social media platforms, enabling users to request real-time fact-checks of online content. Using an exhaustive dataset of 1,671,841 English-language fact-checking requests made to Grok and Perplexity on X between February and September 2025, we provide the first large-scale empirical analysis of how LLM-based fact-checking operates in the wild. Fact-checking requests comprise 7.6% of all interactions with the LLM bots, and focus primarily on politics, economics, and current events. We document clear partisan asymmetries in usage. Users requesting fact-checks from Grok are much more likely to be Republican than Democratic, while the opposite is true for fact-check requests from Perplexity -- indicating emerging polarization in attitudes toward specific AI models. At the same time, both Democrats and Republicans are more likely to request fact-checks on posts authored by Republicans, and - consistent with prior work using professional fact-checkers and crowd judgments - posts from Republican-leaning accounts are more likely to be rated as inaccurate by both LLMs. Across posts rated by both LLM bots, evaluations from Grok and Perplexity agree 52.6% of the time and strongly disagree (one party rates a claim as true and the other as false) 13.6% of the time. For a sample of 100 fact-checked posts, 54.5% of Grok bot ratings and 57.7% of Perplexity bot ratings agreed with ratings of human fact-checkers, which is significantly lower than the inter-fact-checker agreement rate of 64.0%; but API-access versions of Grok had higher agreement with fact-checkers than did not significantly differ from inter-fact-checker agreement. Finally, in a preregistered survey experiment with 1,592 U.S. participants, exposure to LLM fact-checks meaningfully shifts belief accuracy, with effect sizes comparable to those observed in studies of professional fact-checking. However, responses to Grok fact-checks are polarized by partisanship when model identity is disclosed, whereas responses to Perplexity are not. Together, these findings show that LLM-based fact-checking is rapidly scaling, is generally informative although far from perfect, while also becoming entangled with polarization and partisanship. Our work highlights both the promise and the risks of integrating AI fact-checking into online public discourse.

Introduction While pathogen genomics using next-generation sequencing (NGS) has been recommended by the WHO as an essential tool for national communicable disease surveillance programmes, procurement and supply chain management (PSM) systems for this new technology are still evolving. To assess the status of PSM systems for pathogen genomics, we examined perspectives from end-users and manufacturers across South and Southeast Asia.Methods Between 2022 and 2023, a cross-sectional survey was conducted among institutional partners supporting pathogen genomics among primarily low- and middle-income countries in South and Southeast Asia. This was complemented by qualitative interviews with the major regional NGS manufacturers. A PSM framework was employed to assess sales, procurement, production, distribution and post-sales support. Analyses are expressed as proportions and means or medians for continuous variables.Results A total of 42 partners across 13 countries, 3 genomics manufacturers and 22 laboratory personnel contributed data to this assessment. PSM challenges were reported by all countries and for all sequencing platforms. High costs of equipment and consumables were identified by 85% of respondents. Long equipment purchasing lead times and reagent re-supply times were reported by 69% and 77% of countries, respectively, with reagent resupply times averaging 8 weeks (IQR 6.2-9.0). Additional barriers included customs clearance, variability of import procedures, taxes and duties. Manufacturers reported a range of strategies to respond to PSM bottlenecks, including establishing regional hubs, distributor networks and financing schemes.Conclusion Coordinated national and regional efforts are required to improve PSM systems for pathogen genomic sequencing to enhance timely early disease detection and response capacity in South and Southeast Asia.

Ciprofloxacin resistant Klebsiella pneumoniae is common or emerging in many geographies, and knowledge of local resistance rates is important for empirical therapy. While K. pneumoniae ciprofloxacin resistance determinants are known, there is a lack of systematic data on the effect of these determinants, alone and in combination, and there are no publicly accessible tools for predicting ciprofloxacin resistance from whole genome sequence data. The KlebNET-GSP AMR Genotype-Phenotype Group aggregated a matched genotype-phenotype dataset of n = 12,167 K. pneumoniae species complex (KpSC) isolates from 27 countries between 2001 and 2021. We developed a rule-based classifier to predict ciprofloxacin resistance by categorizing the number of mutations in the quinolone resistance determining regions of gyrA and parC and the number of plasmid-mediated quinolone resistance genes, including the presence/absence of aac(6')-Ib-cr (which can acetylate ciprofloxacin). Predictive performance was assessed using the discovery dataset, with the reassessment of discrepant and concordant isolates; and validated using externally contributed datasets (n = 7,030 KpSC isolates). The rule-based classifier predicted R vs S/I with categorical agreement, sensitivity, and specificity > 96

Pathogen genomic data provide important insights for public health microbiology, yet genome analysis options often remain highly technical and beyond the reach of many microbiologists and public health practitioners. Pathogenwatch () is a platform that translates pathogen genome data into outputs directly usable for surveillance and public health action. The platform contextualises bacterial, viral, and fungal genomes within a unified framework integrating organism identity, variant or lineage assignment, antimicrobial resistance and virulence gene detection, and geographic and temporal context. Pathogenwatch provides multilocus sequence typing (MLST) for more than 37 bacterial species and core genome MLST (cgMLST) schemes for over 20 priority organisms, with user-uploaded genomes automatically compared against over 875,000 curated public bacterial genomes. The platform has been adopted by 14,389 registered users across 165 countries. In 2025, users uploaded 328,676 genome assemblies and 20,830 read datasets. Pathogenwatch replicates analysis results of complex bioinformatics pipelines. Benchmarking of SARS-CoV-2 lineage assignment against an established reference dataset demonstrated complete concordance for all Variants of Concern and Interest, and full concordance with contemporary Pangolin calls across non-VOC/VOI lineages. Pathogenwatch operates as a continuously deployed, containerised system designed for scalability, reproducibility, and rapid incorporation of new pathogens, positioning it as durable infrastructure for both endemic surveillance and genomic response to emerging threats. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This research was funded by the NIHR (NIHR133307) using UK international development funding from the UK Government to support global health research. The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR or the UK government. Additional funding was provided by the Gates Foundation (grant ref INV-025280 to DMA). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All public data represented in Pathogenwatch are available for download within the application ().

Background Zoonotic disease outbreaks are a growing global health concern, highlighting the need for tools to strengthen outbreak responses. Process mapping is a systems thinking approach that allows the complex multilevel, multisectoral network of actors and processes involved in an outbreak response to be clearly visualised and analysed using a One Health lens. We applied process mapping to a rabies outbreak in the previously rabies-free Philippine province of Romblon to demonstrate its utility.Methods Key stakeholders were interviewed to inform the development of a provisional map of the actors, processes and challenges involved in the outbreak response. During a facilitated workshop, stakeholders worked in multisectoral, multilevel groups to refine the map and co-develop solutions to the challenges identified.Results Process mapping revealed multiple challenges within rabies surveillance and outbreak response, particularly relating to case testing, recording and reporting, as well as intersectoral communication, community sensitisation and dog vaccination. Many of these issues were exacerbated by structural factors such as the Philippines’ archipelagic geography and decentralised governance of animal and human healthcare delivery. Broader issues linked with rabies’ endemic and neglected status were also identified: insufficiently One Health-orientated national guidance on outbreak declaration and response; resource constraints exacerbated by political devolution; and persistent weaknesses in intersectoral coordination and data sharing, compounded by gaps in international guidance on rabies and One Health practice. Suggested solutions that were possible to implement quickly with minimal resources, such as staff training, were undertaken. Recommendations requiring structural change, like enhancing reporting systems, were communicated to the provincial, regional and national government as policy briefs.Conclusion Our research demonstrates the potential value of process mapping for enhancing One Health disease surveillance and outbreak response, strengthening health systems, and improving preparedness. To address the more systemic challenges revealed, outbreak response guidance must be improved, for rabies specifically and within One Health frameworks more generally, and the implications of political devolution on the control of neglected tropical diseases should be explored. These steps will be essential to achieving the global target of zero human deaths from dog-mediated rabies by 2030.