Definition
Plain language
Searching over page images — scans, slides, screenshots — instead of extracted text.
As stated in the literature
Retrieval where documents are indexed as rendered page images via a vision encoder rather than parsed text, preserving layout and figures but exposing the pipeline to purely visual manipulation.
Also called: visual document retrieval
Why it matters: It handles documents that text extraction mangles — slides, forms, charts — at the cost of making the picture itself something an attacker can tamper with.
For example, searching a folder of scanned invoices by looking at the page images themselves, so the total in a table or a stamped signature still counts as findable content.
Heard on the show
“There's one recent image-only line in visual-document retrieval, but its main attack leans on gradient-based optimization, and that degrades badly under black-box transfer.”Episode 247 — One Edited Photo, an Honest Caption, and a RAG System That Believes It