**Background:** Japan introduced Foods with Function Claims (FFC) in April 2015, allowing manufacturers to notify the Consumer Affairs Agency (CAA) about health functions of food products without pre-market evaluation. Companies can substantiate claims using either clinical trials (RCTs) or systematic reviews (SRs). By January 2023, 94.5% of 6,247 notifications used SRs. Previous studies using the original AMSTAR tool in 2017 and 2019 found poor SR quality, but no improvement trend was observed. This study aimed to assess SR quality using the updated AMSTAR 2 tool.
**Methods:** From 519 SRs (H1-H535) published on the CAA website between April 1, 2022 and October 31, 2022, 40 SRs were randomly selected. Exclusion criteria included clinical/observational studies and duplicate SRs for the same functional ingredient. Four reviewers independently assessed quality using AMSTAR 2 (16 items, 7 critical domains). Inter-rater reliability showed substantial agreement (91.9%, k=0.849). Items were scored as yes (1), partial yes (0.5), or no (0), with percentage scores categorized as good (80-100%), poor (50-79%), or very poor (0-49%). Overall confidence was rated as high, moderate, low, or critically low.
**Key Results:** Of the 40 SRs, 25 (62.5%) were supplements, 14 (35.0%) processed foods, and 1 (2.5%) fresh food. 36 (90%) were qualitative reviews without meta-analysis, and 4 (10%) included meta-analysis. The mean AMSTAR 2 percentage score was 51.1% (SD 12.1%, range 19-73%). Overall confidence ratings were: high 0 (0%), moderate 0 (0%), low 2 (5%), and critically low 38 (95%). Among critical domains, items with good descriptions included: #11 (appropriate meta-analysis methods, 100%), #15 (publication bias investigation, 100%), #7 (list of excluded studies, 92.5%), and #9 (satisfactory RoB assessment, 80%). Critical domains with very poor descriptions were: #2 (review protocol, 5%), #4 (comprehensive literature search, 0%), and #13 (accounting for RoB in interpretation, 47.5%). Non-critical domains with very poor descriptions included: #5 (study selection in duplicate, 17.5%), #8 (adequate description of included studies, 10%), #10 (reporting funding sources, 2.5%), #14 (explanation of heterogeneity, 15%), and #16 (reporting conflicts of interest, 17.5%). The number of critical domain deficiencies per SR was: 4 deficiencies in 7.5%, 3 in 52.5%, 2 in 35%, and 1 in 5%.
**Clinical Implications:** The overwhelming majority (95%) of SRs in the FFC system are of critically low quality, meaning they should not be relied upon to provide accurate summaries of available evidence. This poses risks to consumer protection and informed decision-making. Key deficiencies—lack of protocol registration, inadequate literature searches, and insufficient RoB consideration—are correctable through mandatory pre-registration, requiring publication in peer-reviewed journals, and providing methodological guidance based on AMSTAR 2, PRISMA 2020, and GRADE frameworks. The study's limitations include potential publication bias from random sampling, lack of author clarification for unclear items, no assessment of primary study quality or functional mechanisms, and inability to link results to specific companies due to legal concerns.