Skip to content

Making presentations, documents and spreadsheets with ChatGPT

Getting usable files out of ChatGPT — how preview and review differ by surface, using annotations for targeted revisions, and why you should always ask where it saved the output and how it checked it.

CurrentLast verified

Platforms

  • ChatGPT desktop app (preview and annotations)
  • ChatGPT on the web (Work)
  • Codex CLI (no visual preview)
  • Codex IDE extension

What the official documentation says

  • When a task produces a file, give ChatGPT the source data, expected file type, structure and review criteria that matter for the task; the preview and review tools depend on the surface you use.

    Work with files

  • The ChatGPT desktop app previews generated documents, presentations, spreadsheets and PDF files alongside the chat, and when automatic previews are enabled it can open a generated file after a task finishes.

    Work with files

  • When HTML previews are available, generated .html and .htm files can open as interactive previews, and you can switch between the rendered preview and source view.

    Work with files

  • In ChatGPT Work on the web you attach source files or ask ChatGPT to create a document, presentation, spreadsheet or PDF, review it in the chat, and download it when needed.

    Work with files

  • Codex CLI can create and edit files in the working directory but does not include a visual file preview or annotation interface.

    Work with files

  • The IDE extension can create and edit files in the workspace; text and code files are reviewed in the editor, and documents, presentations, spreadsheets or PDFs open in a compatible viewer.

    Work with files

  • For spreadsheets and presentations, the documented advice is to describe the sheets, columns, charts, slide sections and checks you expect, and to ask ChatGPT to explain where it saved the output and how it checked the result.

    Work with files

  • Annotations let you point to a specific part of a file and say what to change, and the same annotation workflow available for code, Markdown files and websites also works with documents, spreadsheets and presentations.

    Work with files

  • Annotations are described as particularly useful after the first draft, when the work needs review and iteration.

    Work with files

Good files start with stated acceptance criteria

The opening of the official page is the whole lesson: when a task produces a file, give ChatGPT the source data, the expected file type, the structure, and the review criteria that matter for the task.

A more specific instruction follows: for spreadsheets and presentations, describe the sheets, columns, charts, slide sections and checks you expect, and ask ChatGPT to explain where it saved the output and how it checked the result.

Unpacked, that is a prompt checklist:

  1. what the source data is
  2. what file type you want
  3. what the structure looks like (sheets / columns / charts / sections)
  4. what counts as passing (the checks)
  5. ask it to report where it saved the output
  6. ask it to report what it checked

Items 5 and 6 are the ones most people skip, and they are what turns "it looks finished" into "it is finished".

Your surface decides how you can review

The documentation splits this out per surface, and the differences are practical:

ChatGPT desktop app — previews generated documents, presentations, spreadsheets and PDF files alongside the chat. With automatic previews enabled, it can open a generated file after a task finishes. When HTML previews are available, generated .html and .htm files open as interactive previews, and you can switch between the rendered preview and the source view. This is also where annotations work — point at a specific part of a supported preview and request a focused revision.

ChatGPT Work on the web — attach source files or ask ChatGPT to create a document, presentation, spreadsheet or PDF. Review the generated file in the chat, download it when needed, and give targeted feedback for the next version.

Codex CLI — can create and edit files in the working directory, but has no visual file preview or annotation interface. The documented substitute: ask Codex to report each output path and the checks it ran.

IDE extension — can create and edit files in the workspace. Review text and code files in the editor; open documents, presentations, spreadsheets or PDFs in a compatible viewer.

Annotations: the tool for the second draft

Annotations let you point to a specific part of a file and tell ChatGPT what to change. What is worth noticing is their reach — the same annotation workflow that works for code, Markdown files and websites also works with documents, spreadsheets and presentations.

The three documented examples show the range:

  • Select a navigation bar on a website and ask ChatGPT to change its font.
  • Highlight a claim in an investment thesis and ask for its source.
  • Mark a chart on a slide and request a clearer label.

ChatGPT uses the selected area as context for your request, so you can refine the file without starting over or changing the parts you already like.

The documentation names the right moment explicitly: annotations are particularly useful after the first draft, when the work needs review and iteration. Write the first version with a careful prompt; converge the rest with annotations. That division of labour is the fastest path.

Phrasing a revision on the web

The web has no annotation path, so the wording has to carry the precision. The documented approach:

  • Name the page, slide, sheet, table or passage that needs attention.
  • Describe what should stay unchanged.
  • Ask ChatGPT to report the new file name and the checks it performed before you download the next version.

The second bullet is the habit worth forming. "Fix slide three" invites collateral edits to the parts you were happy with.

Steer while it runs

One easily-missed capability in the desktop app: use the chat sidebar while a task runs. It can surface the agent's plan, sources, generated files and chat summary, so you can steer the work, inspect generated files and request another pass mid-flight — rather than discovering at the end that it went the wrong way.

How to do it

  1. Assemble the source data and decide which file type you actually want.
  2. Describe the structure in the prompt — sheets, columns and charts for a spreadsheet; sections for a deck; and the checks you expect for either.
  3. Explicitly ask ChatGPT to explain where it saved the output and how it checked the result.
  4. Review the preview alongside the chat in the desktop app, or in the chat on the web, downloading when needed.
  5. For a targeted revision, use annotations to point at the specific part and say what should change and what should stay.
  6. Iterate with focused feedback about the structure, data, layout or validation that needs another pass.

On Windows

  1. The Windows desktop app has the same file preview capability as macOS — documents, presentations, spreadsheets and PDFs preview alongside the chat.
  2. The annotation workflow applies equally, including annotations on code, Markdown files and websites.

On mobile

N/AThe official page describes preview and review behaviour per surface for the desktop app, the web, Codex CLI and the IDE extension. It does not separately document file preview or annotation behaviour on mobile, so this tutorial makes no claim about it.

Use cases

  • Turning interview notes and research into a deck with clear sections and supporting evidence.
  • Turning notes and material into a comparison spreadsheet with scores, flagged risks and a summary tab.
  • Generating an HTML page and switching between the rendered preview and the source view to inspect it.

Common mistakes

  • Asking for "a deck" without naming sections or checks. The documented advice for spreadsheets and presentations is to describe exactly those.
  • Not asking where the output was saved. The documentation recommends asking ChatGPT to explain the location and the checks it ran.
  • Expecting a visual preview in Codex CLI. The documentation states it has no visual file preview or annotation interface.
  • Requesting a revision without saying what should stay unchanged, and losing the parts you were happy with.

FAQ

Which surfaces can preview generated files?
The desktop app previews generated documents, presentations, spreadsheets and PDFs alongside the chat, and can open a generated file after a task finishes when automatic previews are enabled. ChatGPT Work on the web reviews the file in the chat with a download when needed. Codex CLI can create and edit files but has no visual preview or annotation interface. The IDE extension reviews text and code files in the editor and opens documents, presentations, spreadsheets or PDFs in a compatible viewer.
What can annotations be used on?
The documentation says annotations let you point to a specific part of a file and tell ChatGPT what to change, and that the same annotation workflow available for code, Markdown files and websites also works with documents, spreadsheets and presentations. Its examples: selecting a navigation bar on a website and asking for a different font, highlighting a claim in an investment thesis and asking for its source, marking a chart on a slide and requesting a clearer label.
Can I see a generated HTML file rendered?
Yes. When HTML previews are available, generated .html and .htm files can open as interactive previews, and you can switch between the rendered preview and the source view to inspect the output or its underlying HTML.
How should I phrase a revision request on the web?
The documented approach is to name the page, slide, sheet, table or passage that needs attention, describe what should stay unchanged, and ask ChatGPT to report the new file name and the checks it performed before you download the next version.

Official sources

These are the pages this tutorial is checked against. Follow them if you need the vendor's exact wording.

Source status