The emergence of highly pathogenic avian influenza (HPAI) H5N1 in dairy cattle has presented a major challenge for dairy producers in the USA, who have had to make biosecurity decisions with limited information while balancing evolving guidance, existing beliefs, and farm management constraints. This study aimed to 1) describe dairy producers´ perceptions and understanding of HPAI; 2) characterize experiences and perceived impacts among producers affected by HPAI; 3) identify farm characteristics associated with implementation of biosecurity changes in response to HPAI risk; and 4) assess whether perceived biosecurity effectiveness was associated with implementation of biosecurity changes and whether concern about HPAI and perceived farm-level risk helped explain this relationship. A questionnaire mailed to Michigan dairy facilities yielded a 15.6% response rate (n = 120/766). Descriptive statistics, modified Poisson regression, logistic regression, and parallel mediation analysis using parametric g-computation were used to evaluate survey responses and associations with implementation of biosecurity changes. Overall, 18 respondents (15%) reported confirmed or suspected HPAI on their farms. Most respondents reported being moderately to very familiar with HPAI (76%) and receiving sufficient guidance about the disease (83%). The most frequently selected information sources were veterinary professionals (72%), government health agencies (52%), and industry associations (48%). Producers most frequently identified wildlife (74%) as a likely route of HPAI introduction, followed by airborne transmission (34%), cattle movement (23%), and sharing workers with other operations (12%). Among affected farms, the reported clinical phase lasted with a median of 21 d, whereas reduced productivity lasted with a median of 40 d. Median self-estimated economic loss was $152 per lactating cow. Among farms without reported confirmed or suspected HPAI, farms with ≥ 15 employees were more likely to report implementing biosecurity changes than farms with < 5 employees (RR = 2.26; 95% CI: 1.24–4.10), and farms using robotic milking systems were more likely to report changes than farms using parlor or rotary systems (RR = 2.07; 95% CI: 1.21–3.55). From a mediation analysis (n = 93), the estimated probability of implementing biosecurity changes was 45.7% among producers who perceived biosecurity as effective, compared with 21.4% among those who did not, corresponding to a 24.3 percentage-point difference (95% CI: 5.4–42.4). On the odds-ratio scale, perceived biosecurity effectiveness was associated with higher odds of implementing biosecurity changes (OR = 3.17; 95% CI: 1.21–8.26), higher concern about HPAI (OR = 4.26; 95% CI: 1.58–11.51), and greater odds of perceiving the farm as lower risk than other farms in the state (OR = 2.56; 95% CI: 1.05–6.25). The indirect effect through concern about HPAI was 9.4 percentage points, accounting for approximately 39% of the total association, whereas the indirect effect through perceived farm-level risk was small and not clearly different from zero. Overall, these findings suggest that implementation of biosecurity changes may be supported by trusted communication strategies that emphasize the effectiveness of specific biosecurity practices and present the disease as a relevant but still manageable farm-level threat, and better align producers´ perceptions with evidence-supported transmission pathways, particularly modifiable routes such as animal movement, personnel movement, and shared equipment.