Abstract
<jats:p>Effective biodiversity policy and spatial planning depend on species occurrence data. Yet data that informs conservation decisions typically originate from heterogeneous sources that potentially introduce systematic biases that shape downstream ecological inference and policy outcomes. To explore these potential biases, we analysed reptile location data from Bangladesh and India – a mega-diverse region encompassing multiple biodiversity hotspots as a case study. We quantified sampling bias in museum collections, social media, and citizen science platforms across species traits, habitat zonation, conservation status, and environmental gradients. Museum records were associated with higher values for crepuscular activity, aquatic habits, threatened status and small geographic ranges, but lower values for terrestrial substrates, diurnal activity, and an affinity to colder temperatures. Social media records were more associated with built-up environments, Not Evaluated taxa, terrestrial and saxicolous species, active foragers and smaller-ranged species. Citizen science records showed stronger associations with fossorial and semi-aquatic substrates, mixed foraging, larger ranges, and higher human-footprint gradients. Our findings show that biodiversity datasets are not neutral windows to nature, but rather filtered representations shaped by how records are sought, generated, shared and preserved. Strategic integration of heterogeneous data sources can therefore improve the ecological representativeness of biodiversity evidence and reduce blind spots in conservation planning.</jats:p>