To extract gravitational acceleration during drilling, the Magnetic Inertia Golden Jackal Optimization (MIGJO) algorithm is employed. Initially, the vibration characteristics during drilling are analyzed, and a gravity extraction model is established by categorizing non-gravity accelerations into solution vectors. Then, based on the output characteristics of the magnetic inertial sensor during drilling, an objective function for the ideal gravitational acceleration is defined, along with constraint conditions such as the gravity angle and tangent Pearson coefficient of the drilling tool diameter. Utilizing the Golden Jackal Optimization (GJO) algorithm, the solution vector from the previous step initializes a dynamically scaled random walk population, reflecting the random variations of non-gravity accelerations during drilling. A gravity factor balance algorithm is developed to perform global search and local refinement using the relative error of gravity modulus and trigonometric functions. Additionally, an attack-defense coefficient is introduced to manage the magnetic inertia golden jackal′s behavior, optimizing both attack and defense strategies to improve gravity extraction accuracy and speed. The attack strategy, based on the positions of the optimal and suboptimal solutions, enhances accuracy, while the defense strategy, utilizing upper and lower bounds and mutation points, helps the algorithm avoid local optima. The similarity between the current gravity solution and local gravity design is used to dynamically adjust the solution vector′s position, further refining the accuracy of gravity extraction. Simulated and real-world drilling experiments demonstrate that MIGJO significantly improves the accuracy of gravitational acceleration extraction, with average absolute errors in inclination and tool face angle controlled within 0.63° and 0.8°, respectively. This method effectively enhances the precision of gravity acceleration extraction during drilling.