Loop-mediated isothermal amplification (LAMP) is a potential screening test for avian influenza (AI), but its narrow detection spectrum hinders the field of applications. To improve this narrow detection spectrum, three types of primers were compared for detecting diverse H5-subtype hemagglutinin (HA) genes. Thus, 4 and 6 genes of 10 genetically different H5-HA genes tested were detected with S primers specific for A/duck/Tsukuba/9/2005 (H5N2) and with M primers, which contained mixed bases, respectively. In contrast, all 10 HA genes became positive with the population (P) primers, comprised of a mixture of primers designed for each subpopulation of 2202 HA genes. Our study indicated that the P primers were essential at FIP and BIP primer sites for exhaustive detection, whereas those at F3, FL, BL, and B3 sites were exchangeable with M primers. A base-mismatch experiment demonstrated that HA genes with ≦2 base-mismatches per primer site and ≦10 base-mismatches per HA gene were amplifiable. The RT-LAMP was also broadly reactive, specific for H5-subtype HA genes, and applicable to field samples with the sensitivity of a rt-PCR. The in silico analysis suggests that most H5-HA genes (2586/2588) registered in the GenBank database might be amplifiable. These results indicate that use of subpopulation primers in LAMP allows exhaustive detection of diverse HA genes, and the H5-LAMP can be used as a reliable AI screening test at general diagnostic laboratories.