Screenshot or e-invoice: keep PNG. Photo of paper: JPG is fine.
Merge the images in the order the invoices are listed, one per page.
Confirm the smallest amount and the stamp are still sharp before submitting.
A screenshot of an e-invoice is sharp-edged text on flat white; PNG stores it losslessly and the glyph edges stay clean. JPG is lossy, and at the compression ratios people usually pick the characters grow a halo, which is exactly what drops an OCR match in an expense system. A paper receipt photographed with a phone is the opposite case: it is a photograph, and JPG is both smaller and honest about what the sensor captured.
A PNG screenshot can be 1-2 MB per page, so twenty invoices become 30 MB and the mail gateway refuses it. Compressing afterwards converts those PNGs to JPEG, which is where the text turns soft. The right lever is resolution before conversion: cap the screenshot at roughly 2000 pixels on the long edge rather than scaling a 4K capture down after the fact.
Most expense systems OCR page by page and assign one record per page. Two invoices on one page means one of them silently disappears from the claim. And scan in colour: the red invoice stamp is the field many systems look for, and grayscale printing can drop it entirely.
Reviewers work down your list, so merge in the same order the claim lists the invoices, then name the file something searchable like 2026-10-claim-name-8pages. A forty-invoice blob with a meaningless filename gets sent back more often than it fails OCR.
It was probably compressed at some point. Re-convert from the original PNG or high-quality JPG.
Technically yes, but most systems read one per page and will lose the second.
Preferably. The red stamp is a recognisable field, and black-and-white scanning can erase it.