Against the background of pressure in the book-retail market, intensified competition for platform traffic, and the growth of content commerce, influencer marketing helps publishers generate incremental demand, reach segmented readers, and circulate books across online communities. Yet changing traffic mechanisms, dispersed influencer resources, and rising matching and cooperation costs have exposed the limits of traditional practices based on experience, broad outreach, and opportunistic collaboration. Based on the resource action perspective, this study examines how artificial intelligence is being incorporated into influencer marketing in publishing and how it may change the identification, combination, and use of book content, influencer resources, platform traffic, and reader demand. The study used semi-structured interviews with 14 practitioners from 14 publishing organizations. Two researchers coded the interview materials to examine current application patterns, the forces driving change, and possible directions of development. Three application areas were identified: influencer screening and matching; content-resource provision and asset accumulation; and business operations and process coordination. Specific practices include integrating influencer data to support recommendations, decomposing books into customized promotional materials and reusable content assets, and assisting cooperation follow-up and information sharing. These applications reduce information-processing and communication costs, but they remain constrained by unstable outputs, dependence on prompting, and hallucinations. At present, artificial intelligence mainly improves local workflows rather than fully replacing extensive outreach or temporary resource bricolage. Further analysis attributes the emerging shift to changes in the external resource environment, growing complexity in internal marketing tasks, and the enabling conditions created by artificial intelligence. Publishers therefore need to reorganize the resource-action chain in four respects. Resource identification should move beyond follower counts and historical sales by combining book-content features, influencer performance, user feedback, and network relationships. Decision-making should incorporate auxiliary attribution, counterfactuals and process monitoring, with human correction used to improve reliability and interpretability. Influencer resources should be called upon in tiers, with portfolios and campaign pacing adjusted to pre-launch promotion, content interpretation, concentrated conversion, and long-tail diffusion. Marketing relationships should also become layered, combining sustained cooperation with key influencers and flexible participation by peripheral creators. The study conceptualizes influencers as relational resources that connect reader demand, translate the value of books, and help publishers absorb market and user knowledge. Artificial intelligence can strengthen publishers' capacity to identify, evaluate, use, and coordinate such resources, creating conditions for a shift from temporary resource assembly to more systematic resource action. This potential, however, depends on reliable outputs, human oversight, accumulated organizational data, and continuing coordination with influencers. The analysis provides a resource-action explanation of artificial intelligence's role in publishing marketing and offers practical guidance for influencer-resource management, human–AI collaboration, and more efficient resource use. Publishers must convert isolated tools into sustained, interpretable, and reliable decision support and relationship management.