To improve the consensus forecast of tropical cyclone (TC) tracks over the Northwestern Pacific, artificial intelligence (AI) weather prediction models are integrated with numerical weather prediction (NWP) models, and several consensus forecast schemes are constructed based on simple multi-model average (AVG), selective average (SEAV), and selective change-weighted average (SECW) methods. These schemes are evaluated for their TC track performance during 2023–2024. It is found that compared to NWP models, AI models exhibit weaker performance in short-term forecast, but much better performance in long-term forecast. Three integrated schemes combining AI and NWP leverage the strengths of both approaches, and they are proven to overall outperform pure NWP consensus schemes, while surpassing the pure AI consensus scheme within 72 h. Comprehensive evaluations indicate that the AI+NWP_SECW scheme achieves the best track performance within 72 h (10–30% error reduction from the SSTC), whereas the AI_AVG scheme is identified as optimal at 96–120 h (23–31% error reduction from the SSTC). However, the sample stability of the AI_AVG scheme is shown to be inferior to that of the integrated consensus schemes.