Files, analysis, and deliverables

Home Tools Platforms ChatGPT Files, analysis, and deliverables

Turn supplied documents or data into a reviewable analysis and an editable file.

Conversation & research · Documented · Documentation checked 2026-10-04

ChatGPT Chat and Work; local Codex and the OpenAI API are separate surfaces.

Tools, usage, and controls vary by plan, client, region, rollout, and workspace policy.

The worked example is illustrative and has not been tested in a signed-in product account.

What it does and how it connects

Turn supplied documents or data into a reviewable analysis and an editable file.

Files supply evidence; code or analysis tools transform it; a generated document is the output. Work can create reviewable documents, presentations, spreadsheets, and PDFs. Desktop previews and annotations differ from web review and CLI output paths. A plausible chart is not evidence that the underlying calculation is right.

Walkthrough

Use the product named in the access panel. The steps below combine documented controls with the editorial example.

  1. Attach a small permitted document or data export and explain its columns or sections.
  2. Request an analysis plan that names missing values, units, and assumptions.
  3. Ask for the desired file type plus an explanation of the transformations.
  4. Open the actual file, inspect formulas or figures, and request a focused revision.

Worked example

Use an invented CSV with month, attendance, and event type to prepare a library-program report. Keep the original CSV alongside the output. A useful result includes totals, a readable chart, and an explanation of excluded rows; the totals must be checked independently.

Analyze the attached attendance CSV. First identify column meanings, missing values, and duplicates. Calculate monthly totals and show the calculation method. Create an editable spreadsheet and a one-page summary. Do not invent missing records. List every excluded row and tell me how to verify the chart against the source.

Inspect and verify the result

Recount a sample of rows yourself. Compare chart axes and units with the table, inspect formulas in the downloaded spreadsheet, and confirm that the PDF or slides actually open. Preview annotations are revision requests, not proof that all pages were reviewed.

Limits and troubleshooting

Unsupported formats, scanned pages, and ambiguous headers can reduce extraction quality. If the output omits a sheet or section, supply a smaller export and name the exact content. Never describe generated results as observed measurements without checking the source.

Data, permissions, and actions

Uploading makes selected information available to the task. A generated file may persist in chat or the available file system. Sharing or exporting can create another copy with different access; inspect that destination before distributing it.

Projects, instructions, and memory

Keep related sources together and distinguish deliberate project instructions from recalled context.

Explore Projects, instructions, and memory

From evidence to a beginner lesson

Create a sourced workshop lesson and an editable handout from approved material.

Explore From evidence to a beginner lesson

A weekly evidence digest with review

Repeat a public-source investigation and deliver a private draft digest.

Explore A weekly evidence digest with review

Evaluations: does the system do the job?

Create representative tests for quality, tool behavior, refusal, escalation, and the effects of a change.

Explore Evaluations: does the system do the job?

Sources and verification

Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.

Work with files · OpenAI

Using ChatGPT: Chat and Work · OpenAI

Share this page