Aim To perform a systematic review, meta-analysis and Delphi exercise to evaluate diagnostic yield of combined 2-[18F]-fluoro-2-deoxy-D-glucose (FDG) positron-emission tomography and computed tomography (FDG-PET/CT) in fever of unknown origin (FUO). Materials and methods Four databases were searched for studies of FDG-PET/CT in FUO 1/1/2000–1/12/2015. Exclusions were non-English language, case reports, non-standard FDG radiotracer, and significant missing data. Quality was assessed by two authors independently using a standardised tool. Pooled diagnostic yield was calculated using a random-effects model. An iterative electronic and face-to-face Delphi exercise generated interspeciality consensus. Results Pooled diagnostic yield was 56% (95% confidence interval [CI]: 50–61%, I2=61%) from 18 studies and 905 patients. Only five studies reported results of previous imaging, and subgroup analysis estimated diagnostic yield beyond conventional CT at 32% (95% CI: 22–44%; I2=66%). Consensus was established that FDG-PET/CT is increasingly available with an emerging role, but there is prevailing variability in practice. Conclusion There is insufficient evidence to support the value of FDG-PET/CT in investigative algorithms of FUO. A paradigm shift in research is needed, involving prospective studies recruiting at diagnosis of FUO, with updated case definitions and hard outcome measures. Although these studies will be a significant undertaking with multicentre collaboration, their completion is vital for balancing both radiation exposure and costs against the possible benefits of utilising FDG-PET/CT.