Based on gold nanoclusters (AuNCs) with good optical stability, biocompatibility and simple and non-toxic preparation method, this paper developed a highly selective, highly sensitive and visualized uric acid (UA) sensor. We synthesized BSA-AuNCs using bovine serum albumin (BSA) as a template. Under the catalysis of urate oxidase, UA produces stoichiometric hydrogen peroxide (H₂O₂), which causes the fluorescence of AuNCs to be quenched. In addition, we found that BSA-AuNCs mimic the peroxidase activity in this system, catalysing the substrate's oxidation 3,3',5,5'-tetramethylbenzidine (TMB) to ox-TMB by H₂O₂. At this time, the emission spectrum of BSA-AuNCs overlaps the absorption spectrum of ox-TMB, and a kind of fluorescence resonance energy transfer (FRET) occurs. BSA-AuNCs acts as a donor to transfer the excitation energy to the acceptor ox-TMB, which makes ox-TMB produce fluorescence. At the same time, the fluorescence intensity of BSA-AuNCs is significantly lower than when it exists alone, which significantly improves the sensitivity of UA detection. Under optimum conditions, it is found that the UA concentration has an excellent linear relationship between the quenching degree of 2~100 μmol·L^-1, the linear equation is (F₀-F)/F₀=0.005 85c_UA+0.103 64, the linear correlation coefficient is 0.995 4, and the detection limit is 0.26 μmol·L^-1, which is far below the minimum limit of normal human UA levels (90 μmol·L^-1). Meanwhile, the recovery of UA in blood samples was studied, and the recovery rate was 97.3% to 104.7%, indicating that this method is effective in clinical blood samples. The detection has tremendous application potential, and provides an excellent theoretical basis and methodological guidance for further clinical analysis.