Analyze PDF, Word, and Excel Files

ChatGPT reads and analyzes files directly within the conversation. You upload a PDF, Word document, or Excel spreadsheet and ask for a summary, analysis, data extraction, or rewrite. What used to take an evening of reading now takes a few exchanges—provided you know how to frame your request. Simply uploading a file and asking, “What do you think of this document?” will only yield a generic response.

Upload a file

Upload files using the + button in the input bar: select your file and pair it with a clear instruction. File uploads are available on all plans, including the free tier; audio features are reserved for paid plans and workspaces. Supported formats cover most business use cases.

OpenAI’s Help Center specifies several limits: 512 MB per document or presentation, 2 million text tokens per document, approximately 50 MB for a spreadsheet, 20 MB per image, and 80 files every 3 hours (reduced to 3 files per day on the free plan). For large files, upload only the relevant sections. Google Docs files cannot be imported directly; export them as PDF or DOCX first. Also, consider the file type: a scanned PDF is an image, not text. ChatGPT reads it using vision capabilities, which is useful but less precise than native text. Whenever possible, choose the original digital document over a scan.

Get the most out of a PDF

PDFs dominate professional communication, and this is where you’ll see the biggest time savings. For a 50-page report, start with a synthesis: “Summarize this report in 10 key points. Identify the main conclusions and recommendations.” For a contract, aim for surgical extraction instead: “In this contract, identify all clauses related to late penalties and termination conditions. Present them in a table.”

Uploading two files in the same conversation enables comparison—a tedious task manually, but instant here: “Compare these two versions and list the differences clause by clause.” Finally, ChatGPT can play the skeptical reader, often the most valuable role: “Read this financial report and identify inconsistencies, missing data, or claims unsupported by figures.” Note the difference between these four requests—summarize, extract, compare, critique. The verb you choose drives the quality of the response.

Work with Excel files

With spreadsheets, ChatGPT doesn’t just read; it uses its code execution environment. This means real calculations, not estimates—a critical distinction when accuracy matters. Start by mapping the content: “Describe the contents of this Excel file: how many sheets, how many rows per sheet, and what are the main columns?” Then move to analysis: “Calculate the mean, median, and standard deviation of the ‘Revenue’ column. Identify the 5 highest and 5 lowest values.” You can immediately request visuals: “Using this monthly sales data, create a chart showing the trend over the year with a trendline.” If the file looks suspicious, diagnose it before drawing conclusions: “This file contains duplicates and empty cells. Identify the issues and propose a cleaning plan.”

Revise a Word document

For text documents, three approaches cover most needs. First, rewriting: “Rewrite this document in more accessible language, keeping all information but simplifying technical jargon.” Second, structural review, useful before distributing an important memo: “Analyze the structure of this document. Is it logical? Propose a reorganization if needed.” Third, action extraction, which turns meeting minutes into a work plan: “Read these meeting minutes and extract all action items with responsible persons and deadlines.”

Handle long documents

For documents spanning hundreds of pages, a single broad request yields a diluted response. Proceed in two steps. First, map the structure: “Give me a summary of this document’s structure: section titles, page count per section, and main themes.” Then drill down where it matters: “Now, analyze Section 4 on financial projections in detail.” This way, you direct the analysis rather than letting it happen to you.

Your turn

Read a document with ChatGPT, then without. Test this method on a professional document you’ve read recently—a report, contract, or presentation—using this prompt: “Analyze this document and provide: (1) a 5-point summary, (2) the 3 most important pieces of information, (3) questions this document leaves unanswered, (4) one priority action item based on its content.” Compare the result with what you retained from your own reading. The gap will teach you as much about your reading habits as about the tool. One final reflex before every upload: files pass through OpenAI’s servers, so evaluate the sensitivity of what you’re sharing. Also note that a file saved in your library isn’t deleted when you delete the conversation; you must delete it separately.

Official sources: Uploading files and audio to ChatGPT · Data analysis with ChatGPT (OpenAI Help Center)

Key takeaways

• ChatGPT analyzes PDF, Word, Excel, PowerPoint, and text files directly in the conversation
• For long files, proceed in two steps: overview first, then targeted deep dive
• Excel files benefit from code execution for real calculations
• Be specific in your analysis requests to get actionable results
• Evaluate data sensitivity before uploading to ChatGPT