Background <p>Although investigations of the intratumoral microbiota date back thousands of years, breakthrough transformations have only recently been achieved through high-throughput sequencing and multiomic technologies. These advances have revealed diverse and tumor type-specific microbial communities that drive carcinogenesis via immunomodulation, metabolic reprogramming, and genomic instability. Current cornerstones of cancer therapies—including chemotherapy, radiotherapy, immunotherapy, and targeted therapy—are limited by systemic toxicity, localized tissue damage, drug resistance, and low patient response rates. These constraints underscore the urgent need for more effective and precise therapeutic strategies.</p> Main body <p>This review comprehensively integrates artificial intelligence (AI) technologies into the characterization of the intratumoral microbiota, facilitating the development of novel computational pipelines for mapping microbe–host crosstalk. We systematically summarize recent advances in engineered microbial therapeutics, including bacteria designed for targeted antitumor activity and engineered microorganisms that enable the localized delivery of therapeutic agents. Furthermore, this review critically evaluates the safety profiles of microbiota-based interventions and discusses key challenges in clinical translation.</p> Conclusions <p>By combining cutting-edge computational technologies, biological research, and clinical insights, this review aims to bridge the gap between microbiome science and oncological practice, pioneering innovative strategies for microbiota-guided diagnostics and personalized cancer therapy.</p>