HARF: A Human–AI Collaborative Framework for Cultural Heritage Reconstruction with Expert-Guided Multi-Platform Generative AI and Systematic Prompt Engineering

The destruction of cultural heritage through conflict and environmental degradation has created urgent needs for reliable digital reconstruction, yet current generative AI approaches often prioritise aesthetic plausibility over cultural authenticity, a risk exemplified by the loss of the Western Buddha of Bamiyan in 2001. This study introduces and validates the Heritage-Aligned Reconstruction Framework (HARF), a reproducible method that integrates archaeological evidence, dimensional analysis, and domain expertise to constrain AI-mediated reconstruction. HARF is structured around three components; a blueprint that encodes verified archaeological and iconographic information into prompts, a quantitative index assessing prompt completeness, and a geometric pre-evaluation stage that screens outputs against site-specific tolerances. Validation proceeded in two phases involving multidisciplinary feedback from 33 international specialists and targeted scoring by Bamiyan experts. HARF consistently improved evidentiary alignment compared with unstructured prompting, with the strongest candidates reaching the minor-revision band of historical plausibility, though none attained exemplar-level fidelity. Expert assessments identified dimensional accuracy and niche geometry as the most reliable predictors of trust, while persistent challenges remained in head morphology and drapery articulation. By operationalising the principles of the London Charter and Seville Principles (Denard, 2009; López-Menchero Bendicho & Grande, 2017) within a transparent, auditable workflow, HARF provides a culturally grounded and methodologically accountable approach to reconstructing destroyed monuments. It offers a transferable model for guiding AI-generated heritage imagery toward evidential rigour, interpretive clarity, and cultural sensitivity, and can be applied to other destroyed or endangered monuments where archival imagery, dimensional documentation, and basic iconographic evidence are available. Preprint notice: This is the author manuscript/preprint version of a manuscript currently under review at Digital Applications in Archaeology and Cultural Heritage. It has not yet been formally accepted or published as the Version of Record. If accepted, this Zenodo record will be updated to link to the final published article.