ChatGPT users can now upload a photo of a paper or PDF form, describe what should go in each field through natural conversation, and receive a completed version of the form. OpenAI rolled out the feature on May 22 without a formal product announcement, surfacing it through an in-product update and a single developer post on X.
The mechanics are straightforward: the model reads the form image, identifies the field structure, asks the user for any missing information, and renders a completed output that preserves the form’s original visual layout. The output can be downloaded as an image or a PDF for printing or digital signing.
OCR and form-field detection are not new. Adobe Acrobat has handled both for years. The substantive change is the natural-language fill step. Where a traditional form-fill workflow requires the user to click into each field and type, the chat-based pattern lets the user supply all the information conversationally and have the model handle field-by-field placement. For users completing a form they have never seen before (a tax document, a medical intake, a government application), the cognitive overhead drops sharply.
The competitive implication for forms-as-a-service products is clear: chat-based completion will become a baseline customer expectation by mid-2026. Teams building DocuSign-adjacent products, intake forms, or any structured-document workflow should plan to ship a conversational completion path on the same timeline.
Reported via threadreaderapp.com on 2026-05-22.
