The paper describes a system designed by Advacheck team to recognise sophisticated compound machine-generated and human-written texts in two subtasks at the Voight-Kampff Generative AI Detection 2025 workshop organised as part of PAN 2025. Our developed system is a multi-task architecture with shared Transformer Encoder between several classification heads. As multiclass heads were trained to distinguish the domains presented in the data, they provide a better understanding of the samples. This approach led us to achieve 99.95% mean metric on validation set in Task 1 and the third place in the official ranking in Task 2 with 60.85% macro 𝐹1-score on the test set and bypass the baseline by 13%.