Purpose: The aim of this study was to compare clinical and radiographic bone level assessments to intra-surgical bone level registration around implants with peri-implantitis and to identify the clinical variables rendering peri-implant bone level assessment accuracy. Materials and methods: The study sample included 50 implants with peri-implantitis in 23 patients. Registration methods included probing of the vertical distance between the implant/abutment interface and the bottom of the pocket (= VPD), intraoral radiography, bone sounding without flap elevation and intra-surgically assessed interproximal bone level. The latter was considered the true bone level (gold standard). Twenty clinicians evaluated all radiographs. Results: VPD and intraoral radiography resulted in a significant underestimation of the true bone level by 1.0 mm (95% CI: 0.495-1.585; P < 0.001) and 2.3 mm (95% CI: 1.650-2.980; P < 0.013) respectively. Bone sounding without flap elevation did not differ significantly from the true bone level (mean difference 0.2 mm; 95% CI: -0.775 - 0.335; P = 0.429). Duplicate magnification registration of 50 implants resulted in excellent intra- and inter-rater reliability (ICC intra ≤ 0.99; ICC inter = 0.964; P < 0.001). Radiographic underestimation was significantly affected by defect depth (P < 0.001). Variation among clinicians was substantial (mean underestimation range 1.1 mm to 3.8 mm); however, clinical experience had no impact on radiographic underestimation (P = 0.796). Conclusions: Bone sounding without flap elevation was the best predictor of peri-implant bone level, whereas intraoral radiography was the most inferior. Consequently, peri-implantitis may be under-diagnosed if examination is only based on radiographs.