Keyword-based auditing (KA) technology is a crucial measure to achieve cost-effectiveness in cloud auditing applications. Different from probabilistic auditing, which verifies outsourced data by random sampling and verification, KA considers the auditing requirements of multi-user and multi-attribute data by performing keyword searches and targeted audits. KA can significantly reduce auditing costs. However, existing KA schemes usually focus only on auditing the efficiency of target data while paying little attention to remedial measures such as fault localization and data recovery after audit failures. This lack of attention to remediation measures does not guarantee data availability. Therefore, this study proposes a keyword-based multi-cloud auditing scheme (referred to as KMCA) that leverages smart contracts to enable targeted auditing, batch fault localization, and data recovery. Specifically, the targeted auditing module defines the keyword-file mapping based on the searchable encryption index structure and employs Bloom filters’ false-positive rate characteristic to hide keyword frequency and protect privacy. The fault localization module uses a binary search approach to locate error-prone cloud servers in batches and fine-grained localization of corrupted data. The data recovery module formulates multi-cloud redundant storage and data recovery strategies to avoid single-point failure and improve storage fault tolerance. Under the random oracle model, KMCA is provably secure. Performance analysis shows that KMCA is feasible.