For shaking table tests, it is essential to increase the density of the structure by adding artificial masses, which serves as an equivalent density. This modification ensures adherence to the dynamic similarity criterion. However, for underwater shaking table and wave tests, the density of water cannot change, and an increase in the equivalent density of the structure results in a distortion of the test simulation of the fluid–structure interaction. To solve this problem, an innovative similarity law is proposed for the design of fluid–structure models by changing the cross‐sectional dimensions of the structure. A series of bidirectional underwater shaking table tests and unidirectional wave tests were conducted to validate the proposed similarity law, and the traditional similarity law was analyzed for comparison. The test results indicated that the traditional scale model predicted the hydrodynamic pressure acting on the prototype with an error of approximately 50%. In contrast, the proposed scale model reduced this error to about 10%. Furthermore, the relative errors in predicting the displacement, acceleration, and strain of the prototype using the proposed scale model were all below 15%. This demonstrated the reliability of the proposed similarity law applied in the test design for fluid–structure models in underwater shaking table and wave tests.