Selective antioxidant recognition plays a critical role in various scientific and clinical fields, however, development of straightforward detection mechanisms and reduction of technique operation are still barriers nowadays. For example, convenient detection of epigallocatechin gallate (EGCG), the most effective antioxidant among all tea catechins, is highly sought but is severely limited due to the lack of appropriate indicators. To overcome the aforementioned obstacles and to provide more adequate processes that can be employed in antioxidant detection, specific analyte targeting and efficient signal reporting are necessary for the platform design. In this work, a fluorescence "turn on" strategy based on gold nanocluster (AuNC) interface-mediated radical scavenging for selective antioxidant detection is reported. This approach utilizes Lcysteine capped AuNCs with great fluorescence stability as the reporter, where the presence of reactive oxygen species (ROS) generated from the Fenton reaction results in obvious fluorescence quenching. The introduction of antioxidants, that is, EGCG molecules, interestingly, neutralizes the ROS by initiating direct electron and hydrogen atom transfer to eliminate the unpaired free radicals. Besides, the phenolic OH groups and the gallate ring of EGCG also provide a selective and efficient complexation orientation toward Fenton reaction metal ions, where ROS generation is consequently prohibited. This dual-effect phenomenon recovers the originally ROS quenched AuNC fluorescence, which is dependent upon the introduced EGCG quantity. Relying on this fluorescence turn on strategy, the current platform delivers a low detection limit of 1.20 μM with excellent selectivity toward EGCG under the presence of other interfering species. Meanwhile, this approach also provides great accuracy on detecting total antioxidant capacity in commercial green tea samples. This antioxidant fluorometric turn on design therefore gives a more straightforward route on EGCG detection, and its specific recognition capability is very suitable for practical real sample analysis.