To screen the presence of Anisakidae third stage larvae (L3) in fish, fast methods such as the visual inspection and candling have been widely used by the industry over the last 50 years, and they are regulated by the European Parliament and the Council. These methods are ineffective to detect L3 embedded in fish muscles, consequently alternative methods, such as the artificial digestion (AD) and the UV-Press (UVP) are increasingly applied, but their performance needs to be evaluated. The aims of the present work were: 1) to compare the performance of AD and UVP. methods by a Ring Trial (RT) involving highly experienced laboratories; and 2) to evaluate the potential transferability of the best performing method to the industry by a collaborative study involving industrial partners (beta-testing). For RT, each participating laboratory (n = 5) received 6 samples of 100 g of fish fillets spiked with 1 L3 (1 sample), 3 L3 (2 samples), 7 L3 (2 samples), and a negative control sample (without L3). In each positive sample, there were live Anisakis pegreffii L3 collected from a naturally infected fish. The result evaluation was based on the agreement between the number of reported and the number of spiked L3. No false positive sample was detected. The L3 number detected by the UVP method showed higher (90%) level of agreement with the number of spiked L3 than the number of L3 detected by the AD method (83.3%); however, no significant difference in terms of accuracy (p = 0.32) was detected when the two methods were compared. Moreover, considering only the presence/absence of L3 in the samples, the UVP reached 100% of accuracy and 100% of sensitivity; whereas, AD showed 98% of accuracy and 96% of sensitivity. The variability of the UVP method was lower than that of the AD method, indicating a better reproducibility. On the basis of the RT results, the UVP method was selected for the L3-testing. Each industrial partner (n = 3) received 15 samples of 100 g of fish fillet spiked with 1 L3 (2 samples), 2 L3 (2 samples), 3 L3 (2 samples), 4 L3 (2 samples), 5 L3 (2 samples) and 6 L3 (2 samples) and three negative control samples (without L3). The number of L3 counted in 34 out of 45 samples (75.6%) by the UVP method was in agreement with the number of spiked L3. One company reached 93.3% of agreement; whereas the other two companies reached an agreement of 66.7%. Two false negative results were found; whereas, no false positive results were obtained. Moreover, at the industrial level, considering only the presence/absence of larvae in the samples, the UVP reached 97% of accuracy, 94.4% of sensitivity, and 100% of specificity. However, the UVP method, in spite of its accuracy, needs further investigations to provide new time temperature combinations that could allow a reduction of the testing time and its integration in the fishing deck. (C) 2016 Elsevier B.V. All rights reserved.