Gemini now creates and edits documents, PDFs and spreadsheets in chat, including from photos of handwritten notes

Google's latest Gemini upgrade lets users create and edit documents, PDFs, spreadsheets and more directly inside Gemini, according to a hands-on test published by TechRadar. Users can generate a file from a prompt and refine it without repeatedly downloading new versions. TechRadar's writer Eric Hal Schwartz reports that any Gemini user can try it.

How it works

The workflow starts with an upload. A user adds a file to Gemini and asks it to transform that file into a document. The result can be previewed directly in the conversation. From there, users have two editing options:

• Expand the document to make manual edits.
• Highlight a particular section and ask the AI to change it.

According to TechRadar, this means a minor correction no longer has to turn into a long conversation with multiple regenerations of an entire document.

The test: handwritten notes to PDF

Schwartz took handwritten notes during a short presentation on the history of sandwiches, photographed the page and uploaded the image. He asked Gemini to turn the notes into a neat, readable PDF. The notes covered the ancient origins of putting food on bread, the story of the fourth Earl of Sandwich, industrialization, and the introduction of sliced bread in 1928.

The content largely survived. Gemini kept the story of John Montagu supposedly ordering meat between bread during a card game in 1762, and the explanation that sandwiches became popular because they were inexpensive and convenient. It also preserved the caveat that the gambling story is probably exaggerated.

The test: editing

For the edit, Schwartz asked Gemini to move the line about medieval stale bread into its own section, add more detail, and otherwise keep the document the same. He reports that it did "pretty well" but with flaws:

• Gemini added visible citation markers that did not cite anything, even though the document was based on his own notes. He says this could be fixed in a follow-up.
• The second version simplified some original wording. The explanation of industrialization became shorter, and the sliced bread detail was reduced to a brief mention of 1928.

Schwartz says none of these changes ruined the document, but they show how asking AI to improve something can cause users to lose details they wanted to keep.

Caveats from the reviewer

Schwartz says users still need to check that Gemini interpreted handwriting correctly. In his assessment, the feature can misunderstand a phrase or introduce unsupported details, and a polished layout can make those mistakes easier to overlook. He lists students turning lecture notes into study guides and meeting attendees turning hurried notes into shareable documents as likely uses.

Model name, context window, pricing and rollout details for the feature were not part of the TechRadar report.

What this means

The feature moves document creation and revision into the Gemini conversation itself, so a user no longer needs to download and re-upload files to refine them. Schwartz argues that this in-conversation refinement is arguably more valuable for ordinary users than an impressive first draft.

The test also shows the trade-off in that approach. A request to expand one section produced unrequested citation markers and shortened passages elsewhere in the document. Because the output looks polished, careful proofreading remains necessary, particularly when the input is handwriting or when the model adds information beyond the original material.