Convert PDF to Excel

PDF to Excel starts with a picture, not a promise

A PDF page does not contain an Excel-like table waiting to be opened. It contains text fragments and drawing instructions placed at coordinates. The horizontal rules may be lines, rectangles, a background image or nothing at all. A merged heading may sit between two ordinary rows without carrying any machine-readable instruction that says it spans three columns. That is why a converter can look convincing on one invoice and put the next one into a grid that bears no relation to the page.

This converter makes that uncertainty visible. It looks for repeated row baselines and column anchors, then shows every proposed area as a cropped picture from the PDF. The picture is the independent reference. Beside it is the cell grid the workbook would receive. Choose the candidates that are really tables, correct any cells that landed in the wrong place, and export only when the two agree.

A PDF page and an Excel workbook, in brief

PDF

Full name	Portable Document Format
Extension	.pdf
Format type	Fixed-layout document
MIME type	application/pdf

XLSX

Full name	Office Open XML Workbook
Extension	.xlsx
Format type	Spreadsheet — a ZIP archive of XML parts
MIME type	application/vnd.openxmlformats-officedocument.spreadsheetml.sheet

Choose several tables and get one workbook

Every selected candidate becomes a separate worksheet in the same .xlsx file. The default names include the source page and table number so their order is still understandable after the PDF is closed, and each name can be changed before export. When the first row is a set of headings, it can be made bold and frozen. Turn that option off when row one is ordinary data. Column widths are measured from the extracted values rather than left at Excel's narrow default.

The cells themselves remain editable in this tab. That matters for the cases geometry alone cannot settle: a centered title may appear to be a new column, a footnote may share a row anchor with the last record, and a visually merged cell may arrive on its own baseline. Fixing one value here changes exactly that value in the download. The original PDF is never rewritten, and building the workbook does not clear the candidates, so a second corrected export does not require another upload or another analysis.

Draw around a table the detector missed

Automatic suggestions are deliberately conservative. A page of two-column prose also has repeated horizontal positions, and turning every such page into a table would produce a long list of confident-looking mistakes. If the table you can see is absent, select its page and drag a rectangle from one corner of the table to the other. Text inside that region is evaluated separately and added to the same candidate list, with the same picture and editable grid.

The rectangle is especially useful on reports where a table shares a page with a heading, explanatory paragraphs and a footer. Whole-page extraction cannot know which of those blocks belong together; your selection supplies the missing boundary. It cannot, however, create characters that are not stored in the file. A manual selection over a scanned page still has no text to place in cells, and the tool says so instead of returning an empty workbook.

Scans, logos and check marks are different data

A scan is a picture of a table. Reading it requires optical character recognition, where a model guesses letters and numbers from pixels. This page does not perform OCR and labels pages with no usable text layer before export. That avoids the worst possible result: a valid Excel file whose empty cells look like a successful conversion. If you only need to establish whether the document has selectable text, the PDF to text tool gives a page-by-page text-layer verdict without trying to reconstruct a grid.

Logos, certification marks, handwritten ticks and symbols drawn as paths create a similar boundary. They remain visible in each table picture, but they are not silently named or invented as cell values. A label drawn with text may be recovered; a check mark drawn as two lines will not. If those graphics carry business meaning, use the picture during cell review and enter the value you want Excel to hold.

What survives the conversion

The workbook preserves the reviewed text matrix, the order of the selected tables, the sheet names and the optional heading-row treatment. It does not recreate the PDF's fonts, borders, cell fills, exact spacing, formulas or page layout. Those belong to a drawing format, while this export is structured data. Merged cells are not guessed either; when the anchors are irregular, the candidate is marked for review and shown as ordinary editable cells.

That makes this a different starting point from CSV to Excel. A CSV already states which values belong to which rows and columns, so that converter can focus on types, dates, leading zeroes and encoding. A PDF supplies none of those boundaries. Here the work is deciding which visual regions are tables and checking the grid recovered from their coordinates. Once the workbook exists, Excel to CSV can flatten any chosen sheet for a system that expects plain delimited data.

The PDF and workbook stay in this tab

PDF.js opens the source document in browser memory, draws the page previews and exposes the text positions. The candidate analysis runs against those positions locally. When you press the final button, the same tab assembles the XML parts of the .xlsx archive and offers that memory object as a download. There is no upload endpoint in the path and no account is required. Password-protected PDFs are opened with the password in this tab; the password is not stored in the workbook.

This local route is useful precisely because tables often carry customer records, invoices, clinical data or internal financial figures. If you are comparing it with a hosted PDF service, the Smallpdf privacy review separates encryption in transit from the separate question of whether a document leaves the machine at all. Here the receipt states the relevant number directly: uploaded, zero bytes.