PDF to Excel automatically: how to get PDF tables into a spreadsheet

You need the table in a PDF as a spreadsheet because you want to sum it, filter it, reconcile it against another file or load it into a system.

If you already tried copy and paste, or an export, and ended up with numbers stored as text, merged cells, or one table split across ten sheets, we get it. It is not you: a PDF stores characters at positions on a page and has no concept of a table or a row, so every tool has to guess where the table is. Last evaluated: October 2026.

The short answer to how to convert a PDF to Excel automatically: check whether the file has a text layer or is a scan, extract the table with a tool that reads layout and not only text, join the pieces that belong together (above all tables that run over several pages), give each column its type, export to Excel or CSV, and check that the totals still add up. The rest of this guide shows where each shortcut breaks and what holds up.

If you copy, paste or export, we get it: the spreadsheet was not usable

These are the four things that go wrong, and none of them has to do with how careful you were:

• The table is not a table in the file. Text copied from a PDF often arrives as one cell per line, or one long cell, because the file only knows where each character sits.
• Multi-page tables become several tables. A statement or a price list that runs over eight pages is eight fragments, each with its own header repeated as if it were data.
• A scan has no text to copy. If you cannot select the text in the PDF, there is nothing to export until OCR reads the image first.
• Numbers and dates arrive as text. 1.234,56 in one locale and 1,234.56 in another, a date written as 03/04/2026, a minus sign in a different place: Excel will sum none of them until you fix the type.

The story does not end there for most people, because the next idea is to use the tools you already have.

So you reach for the built-in tools, and they get you partway

Excel can read tables from a PDF itself. Its Power Query PDF connector, available through Get Data, returns the tables it finds in the file, can combine similar tables on consecutive pages (the option is on by default), and lets you decide whether border lines count as cell edges. Microsoft's own page lists what to expect: for large files, pick pages one at a time or in small ranges, and when rows that span several lines are not recognised, clean them up in the editor (Power Query PDF connector, Microsoft Learn, read 15 October 2026). It does not describe OCR for scanned pages.

Adobe offers an online converter that says it transfers PDF tables into Excel rows and columns, in XLS and XLSX. The page mentions OCR only in connection with Acrobat Pro, so it does not confirm that the free tool reads scans (Adobe PDF to Excel, read 15 October 2026).

Python libraries such as pdfplumber, Camelot and PyMuPDF give you control and cost you code: each returns one table per page, so you join the pages and drop the repeated headers yourself. We ran all three on the same files in Extract tables from a PDF. And pasting the PDF into an LLM breaks on length, for reasons we cover in how to extract data from long PDFs with an LLM.

The approaches, side by side

The last row is the pattern that holds up, and it is what an agentic IDP does. For the general case of extracting fields rather than tables, see how to extract structured data from a PDF with an LLM.

The solution: an agentic IDP with OCR

anyformat is enterprise-ready agentic document processing: agentic OCR that turns any document into data you can trust, where every value is validated and cites its source. It is built for companies with serious demands on security, scale and audit, so process owners and their IT teams use it alongside the developers who evaluate and integrate it.

You describe the columns once, as a schema, and anyformat does the rest. Parse reads every page on its own, with OCR first for scans, and fixes the reading order. Extract looks at the parsed blocks of every page, recognises which table fragments belong to the same table and joins them in order. Validate runs your rules, and values with low confidence go to a reviewer in the same workflow. You use it through the API, the Python and TypeScript SDKs, or the no-code Studio. The table is an object field with nested columns, and each column has a type:

{
  "name": "line_items",
  "description": "One row per line of the table, across all pages",
  "data_type": "object",
  "nested_fields": [
    { "name": "description", "data_type": "string", "description": "Item description" },
    { "name": "quantity", "data_type": "integer", "description": "Quantity" },
    { "name": "unit_price", "data_type": "float", "description": "Price per unit" },
    { "name": "amount", "data_type": "float", "description": "Row total amount" }
  ]
}

The result is a list with one entry per row, and every value carries a confidence score and the text and page it was read from. From the platform you download results as CSV or Excel, one document at a time or in bulk, and the Excel file keeps the table structure, with sheets for subtables when there are any (anyformat docs, outputs, read 15 October 2026). The same results are available through the API.

Every value carries its evidence, it holds up on hard documents, it is a European company with ISO 27001 and GDPR, and the price covers the whole workflow, not only the extraction. The measure that matters is how many documents go through without anyone touching them, and teams go live in days, not quarters, with SDKs, templates and evals to measure accuracy as they go.

Try it on your own PDF: the Free tier includes 50,000 credits with no credit card, enough to run a few hundred pages and open the Excel file yourself.

If your PDFs are bank statements or invoices, the same flow is worked through in how to extract data from bank statements, which has a section on exporting to Excel or CSV, and in the invoice data extraction API guide.

How to check a PDF to Excel conversion is right

A table that looks right can still be wrong, so build the check into the process:

• Count the rows. If the PDF says "42 lines", the spreadsheet should have 42 rows. A different count flags the file.
• Make the totals close. The sum of the line amounts should match the printed subtotal, and subtotal plus tax should match the total. A validation rule can run this on every file.
• Check the types. Sum a column: if Excel returns zero or an error, the numbers arrived as text. Sort the date column: dates stored as text sort alphabetically.
• Look at the seams. Open the page breaks of one test file by hand: the last row of a page and the first row of the next are where a table is most likely to be cut or duplicated.
• Use the confidence. Set a threshold and route what falls below it to a person who sees the source page next to the value.

What it costs with anyformat

anyformat's pricing is credit-based, per page and per operator. At list rates, Standard Parse is 25 credits and Standard Extract 35 credits per page, so 60 credits per page; Agentic is 100 and 150 credits, so 250 credits per page. A credit costs €0.001 on Business and €0.0015 on Free and pay-as-you-go. For a 10-page PDF that is:

Validation rules are billed separately, per rule and once per extraction rather than per page; check the credits page for the current rate. Start with Standard and move to Agentic only for the files where Standard gets a table wrong.

Limits, stated plainly

• Scans and photos lower accuracy for every method. Use a confidence threshold and a review step for them, and see OCR for poor-quality scanned PDFs.
• Very irregular tables (merged cells, tables that change columns halfway) need a test on your own files before you trust them.
• Agentic mode does not support multi-file documents yet (several files treated as one document). Combine files before upload or process them separately.
• Handwriting stays the weak spot of every approach. Plan a review step.
• Accuracy depends on your documents. Before you rely on any vendor's headline figure, ours included, test fifty of your own, count rows against the source and measure field-level accuracy.

How to choose

For one small, clean table once in a while, Excel's own PDF import or a converter is enough. If you convert PDFs every week, if the tables run over several pages, if some files are scans, or if a wrong number costs you something, ask each tool three questions: does a table that spans pages come back as one table, do the numbers arrive as numbers, and what tells me a row is wrong? A tool that cannot answer the third one will fail you quietly. For how this differs from plain OCR, see IDP vs OCR.

Frequently asked questions

Start with your messiest PDF. Upload the file that breaks your current export, define the columns you need, and open the Excel file with the confidence of every value and the page it came from. The Free tier includes 50,000 credits with no credit card.

Disclosure: anyformat sells document extraction; the comparison criteria and dates are stated.