Technology Push or Demand Pull? Supply-Chain AI Spillovers and the Quality of Green Innovation Toward Sustainable Production in Chinese Listed Firms

Artificial intelligence (AI) diffuses beyond adopting firms into the wider production network, yet the direction and quality of these spillovers remain underexplored, an omission that matters because only high-quality green innovation advances genuine sustainable transformation. We ask how supply-chain AI spillovers shape the citation-weighted quality of firms’ green patents and whether technology-push and demand-pull channels carry equal weight. Building inter-industry AI exposure from a pre-sample (2012) input–output structure for 4963 Chinese A-share listed firms over 2011 to 2023, and identifying effects with a shift-share instrument, we find that AI exposure significantly raises green innovation quality. Upstream technology-push spillovers dominate downstream demand-pull spillovers on average, but the asymmetry is conditional on supply-chain structure: the upstream channel prevails when suppliers are concentrated, and the downstream channel re-emerges when they are dispersed. The effect concentrates in green invention patents rather than utility models and operates through embedded technology transfer rather than in-house R&D. Sequencing AI toward upstream, input–output-dense sectors offers a network-level lever for steering digital technology toward high-quality green innovation and a more sustainable production system.