Hundreds of thousands of seabirds are killed each year as bycatch in longline fisheries. Seabirds are predominantly caught during line setting but bycatch is generally recorded during line hauling, many hours after birds are caught. Bird loss during this interval may lead to inaccurate bycatch information. In this 15 year study, seabird bycatch was recorded during both line setting and line hauling from four fishing regions: Indian Ocean, Southern Ocean, Coral Sea and central Pacific Ocean. Over 43,000 albatrosses, petrels and skuas representing over 25 species were counted during line setting of which almost 6,000 seabirds attempted to take the bait. Bait-taking interactions were placed into one of four categories. (i) The majority (57%) of bait-taking attempts were "unsuccessful" involving seabirds that did not take the bait nor get caught or hooked. (ii) One-third of attempts were "successful" with seabirds removing the bait while not getting caught. (iii) One-hundred and seventy-six seabirds (3% of attempts) were observed being "caught" during line setting, with three albatross species - Laysan (Phoebastria immutabilis), black-footed (P. nigripes) and black-browed (Thalassarche melanophrys)- dominating this category. However, of these, only 85 (48%) seabird carcasses were retrieved during line hauling. Most caught seabirds were hooked through the bill. (iv) The remainder of seabird-bait interactions (7%) was not clearly observed, but likely involved more "caught" seabirds. Bait taking attempts and percentage outcome (e.g. successful, caught) varied between seabird species and was not always related to species abundance around fishing vessels. Using only haul data to calculate seabird bycatch grossly underestimates actual bycatch levels, with the level of seabird bycatch from pelagic longline fishing possibly double what was previously thought.