An image labelled “reconstruction of an ancient Chinese divine beast” may come from a museum, a recent illustrator, a game, or a generative model. Visual polish is not provenance. The first verification step is not deciding whether it looks ancient. It is locating the earliest traceable record: holding institution, accession number, date, material, dimensions, and catalogue title.
A museum page is a strong starting point, not an instruction to stop thinking. First record visible facts such as a horn, wing, inscription, or intertwined tortoise and snake. Keep the institution’s dating and interpretation in a second layer. Place a researcher’s or writer’s inference in a third, and contemporary additions in a fourth. This prevents “the museum catalogues this object as bixie” from becoming “every Pixiu has always looked exactly like this.”
For an image circulating without attribution, use reverse image search and compare the earliest posting dates and highest-resolution files. Cropped marks, recoloring, mirroring, and AI outpainting complicate the trail, so compare local texture, damage, and background as well. If results lead only to shops and stock sites, label the source unconfirmed instead of choosing the most authoritative-looking repost.
For text claims, prioritize primary-text databases, formal collection records, and peer-reviewed research. Keep the original name beside any translation, especially for Bai Ze, Xiezhi, or 乘黄, which can be overwritten by a familiar English animal. A secondary article is useful when it leads readers back to inspectable evidence rather than joining a citation loop whose members merely copy one another.
When publishing AI-assisted beast art, a clear disclosure names the use of AI, the reference object, the fact that the image is not an archaeological reconstruction, and the details supplied by imagination. This preserves room for creativity without inviting search systems to treat a new image as ancient proof. Trustworthy cultural content does not pretend certainty; it makes the path from evidence to conclusion visible.