5.4 Working with Documents and Images

Multimodal AI can help read PDFs, spreadsheets, images, and scans. Instead of “summarize the file,” define scope and output: which fields to extract, whether to cite pages, and what wording must remain exact.

Extracting facts before summarizing usually makes checking easier. Page or paragraph references help you return to the source.

Check file quality first

Missing pages, blurry scans, complex tables, handwriting, and incorrect orientation can affect reading. AI may not know what a file omitted, so the user must check page count and completeness.

For a long file, work chapter by chapter and maintain a list of confirmed facts. Before using contracts, personal data, or internal reports, determine whether uploading is allowed and remove sensitive details when necessary.