ERIC search strategy and title-and-abstract screening dataset for a review of generative AI in EFL learning

This dataset documents a focused search of ERIC conducted on 16 October 2026 and the subsequent title-and-abstract screening of 323 unique records. Three export batches contained 333 record occurrences; ERIC OID deduplication produced 323 unique records. The screening dataset records 287 core-inclusion decisions, 4 supplementary-use records retained for contextual or measurement purposes, and 32 exclusions. The search used one database and one precise query. Screening was conducted by one author and represents the title-and-abstract stage only; it is not a claim of exhaustive coverage, independent dual screening, full-text eligibility assessment, or risk-of-bias appraisal. No article abstracts, full-text PDFs, or third-party figures are included. Bibliographic metadata were obtained from ERIC. ChatGPT assisted with title-and-abstract screening and later PDF reading, Claim extraction, cross-study synthesis, and parts of manuscript drafting and Figure 1 development. The author reviewed 64 core Claims and five disposition records, but did not independently re-screen all 323 records or verify every screening label record by record. The exact ChatGPT product/build label was reported by the author as ChatGPT-6 Luna (High) and could not be independently verified.