Artificial Intelligence (AI) is increasingly embedded in scientific research, yet its role in systematic literature reviews (SLRs) remains methodologically underdeveloped. This study maps how generative AI has been utilized across SLR stages, identifying its benefits, risks, and ethical implications through a systematic review of 124 papers retrieved from Scopus, Web of Science, and PubMed, analyzed via qualitative content analysis. The mapping reveals a fragmented and unstandardized landscape along three dimensions: (i) usage, with AI most often applied to screening and summarization, rarely involved in methodological decision-making, and unreliable in database searching; (ii) benefits of AI adoption; and (iii) risks and ethical concerns, for which we identified mitigation practices. From these findings, we propose the HAICO-SLR Guide (Human-AI Collaboration in Systematic Literature Reviews Guide). This framework represents a methodological advancement by formally classifying each review stage as exclusively human, automatable, or hybrid, shifting the focus from addressing AI in isolation to defining explicit operational boundaries within human-AI collaboration. The guide also outlines how to report AI use in SLRs and the role of the researcher who guided it. Designed as a flexible, structured tool, HAICO-SLR can be seamlessly coupled with existing review protocols (e.g., PRISMA). This adaptability allows researchers across diverse fields, including the social sciences, health sciences, and engineering, to standardize procedural reproducibility and mitigate risk when conducting SLRs. The framework positions AI as a co-intelligence agent within a human-in-the-loop paradigm, enforcing a dual-reporting mechanism that elevates transparency to shared methodological co-responsibility.