The Persistence of AI-Generated Errors Across Manuscript Revisions: Human Discernment, Editorial Intervention, and the Limitations of Artificial Intelligence Self-Correction

This research manuscript examines the persistence and recurrence of artificial intelligence-generated errors across successive revisions of scholarly documents, emphasizing the role of human discernment, editorial intervention, and the limitations of AI self-correction. Authored by Yeshayahu Ben-Yehudah, Founder and Director of The Ben-Yehudah Institute, the study documents a continuing human–AI investigation involving interpretive distortion, incomplete scholarly citations, grammatical errors, incorrect copyright attribution, ORCID iD formatting inconsistencies, and the repeated failure of artificial intelligence to consistently implement previously established editorial instructions. The investigation draws upon three successive versions of the manuscript Does the Transition from Waking to Dreaming Change Consciousness Itself, or Only the Conditions Through Which Consciousness Experiences? (DOI: 10.5281/zenodo.23891057). The author identifies how corrections to one category of error did not prevent other defects from persisting or recurring. Particular attention is given to the distinction between autonomous AI self-correction, human-directed AI revision, and direct human editorial correction. The study demonstrates how the author's DISCERNMENT identified shortcomings that were not independently resolved during the AI-assisted manuscript preparation process. Grounded in the Ben-Yehudah Six-Pillar Process Architecture—STILLNESS, AWARENESS, OBSERVATION, DISCERNMENT, INTEGRATION, and TRANSMISSION—the manuscript examines the relationship between observational fidelity, authorial intent, cognitive distortion, and scholarly reliability. The findings are presented as a continuing qualitative case study rather than a universal conclusion about artificial intelligence. The author has intentionally preserved certain manuscript errors for subsequent examination as part of the broader investigation. Artificial Intelligence Disclosure: Artificial intelligence assisted with the organization, drafting, literature identification, and editorial development of this manuscript. The underlying research observations, identification of errors, investigative direction, and editorial decisions originate with Yeshayahu Ben-Yehudah. The manuscript itself remains subject to the limitations of AI-assisted scholarly production examined in the study. DOI: 10.5281/zenodo.23891573 © 2026 Yeshayahu Ben-Yehudah | The Ben-Yehudah Institute. All rights reserved.