How to Convert a PDF to Word, Excel or JPG Free in 2026, Scanned Files Included (Adobe, Smallpdf, iLovePDF, PDF24, Jobbit)

You have a PDF and you need it as something else. The contract as a Word file so you can edit one clause. The bank statement as a spreadsheet so you can add it up. The signed form as a JPG so the portal will accept it. The scan of an invoice as text you can search. You google "PDF to Word free", upload the file to the first converter, and get back a document where every table has fallen apart, or a sign-up wall after the second file, or a 20 MB limit your scan is over. Adobe still reports the PDF as the format behind almost every official document, and most of the people searching for a converter are not designers; they are people with one file and one deadline.

This guide shows you how to convert a PDF to Word, Excel or JPG, and back again, for free, with the tables intact. It covers which route suits each conversion, what to do when the PDF is a scan and the text is really a picture, when you do not need to convert at all because an agent can pull the numbers or the summary straight out, and what the converter sites (Adobe Acrobat online, Smallpdf, iLovePDF, PDF24, Sejda) actually charge once the free limit runs out.

Convert a PDF on Jobbit in one message

Jobbit is an AI agent, not a converter with a queue. You attach the file and describe the result you want, and it does the conversion, checks it and hands the file back in the same chat. The messages that cover most cases:

• PDF to Word: "Convert this PDF to an editable DOCX. Keep the headings, tables and page breaks, and tell me if any table did not convert cleanly."
• PDF to Excel: "Extract every table in this PDF into one Excel workbook, one sheet per table, with the numbers as numbers, not text."
• PDF to JPG: "Give me each page of this PDF as a separate JPG at 300 dpi, named page-01, page-02."
• JPG or photo to PDF: "Combine these five photos into one PDF in this order, straighten and crop each page, then compress it under 2 MB."
• Scanned PDF: "This is a scan. Run OCR so the text is selectable and searchable, then convert it to Word."
• Word or Excel to PDF: "Turn this DOCX into a PDF that prints on A4 with the fonts embedded."

The agent asks when the job is ambiguous (which table, which pages, which size), which is exactly what a converter site cannot do. The guide to how to write prompts for AI agents has the general pattern; for PDFs the rule is simply to say what you will do with the result.

Which conversion do you actually need?

Half the pain with PDF converters comes from picking the wrong target. This table saves the second attempt:

If the row you need ends in "ask for a summary", read the section on extracting instead of converting before you convert anything.

PDF to Word without the tables collapsing

A PDF stores where each character sits on the page, not that a group of them is a paragraph or a table. A converter has to guess the structure back. Simple documents convert perfectly; anything with columns, tables or text boxes is where the free sites let you down, because their guess is fixed and you cannot ask for another. The habits that get a clean DOCX:

1. Convert the whole file, not a screenshot of it. A PDF made from Word converts far better than a photo of a printout, because the text is real text.
2. Say what matters. "Keep the tables" or "the numbered clauses must stay numbered" changes how the conversion is checked.
3. Check the three places conversions fail: tables, headers and footers, and anything in two columns. Ask the agent to list them: "Which parts of this conversion should I check by eye?"
4. Keep the PDF as the reference copy and edit only the Word version. If you send a converted document back as a PDF, compress it first if it has grown; the PDF compression guide covers the sizes each portal accepts.

Adobe Acrobat online converts one file per tool free before it asks you to sign in, and its Word output is the most faithful of the free sites for simple layouts. For anything with tables, the agent route wins because you can ask for a second pass on the table that broke.

PDF to Excel: the tables problem

The most searched conversion after Word, and the one with the highest failure rate. Bank statements, price lists, timesheets and supplier invoices all arrive as PDFs and all need adding up. The two things that go wrong are merged cells, which converters turn into a staircase of half-empty rows, and numbers that arrive as text with the currency symbol attached, so the sum at the bottom says zero.

Ask for the outcome, not the format: "Extract the transactions on pages 2 to 9 into a spreadsheet with columns Date, Description, Money out, Money in, Balance, as numbers, and total each column." The agent reads the table, rebuilds it and returns an XLSX you can filter; if a page is a scan it runs OCR first. For a monthly routine, the same message with a new file each month becomes a two-minute job, and AI Excel tips shows what to do with the sheet once you have it.

PDF to JPG and JPG to PDF

PDF to JPG matters when a website, an app or a government portal insists on an image: a signature page, an ID document, a certificate. The mistake is exporting at screen resolution and getting a JPG nobody can read. Ask for 300 dpi and one image per page, then check the file size: a portal with a 2 MB image cap will need the JPG resized, which the free image resizer or the same chat handles.

JPG to PDF is the reverse and the more common one: photos of receipts, pages of a form, a passport and a utility bill for a tenancy application. Combine them into one PDF in the right order, straighten the pages, and compress the result to the limit the recipient sets. Photos make enormous PDFs, so the compression step is not optional; the PDF compressor takes a target size such as under 2 MB and keeps the text readable.

Word or Excel to PDF, done properly

Converting into a PDF is the one job every office suite does well, so the only question is which settings. For Word, export with fonts embedded so the layout survives on a phone. For Excel, set the print area and choose fit to one page wide before you export, or the PDF will slice a wide sheet into unreadable strips. If you need a PDF that looks the same everywhere and cannot be edited, that is the format's whole purpose; if you need people to fill it in, ask for form fields instead of a flat page.

Scanned PDFs: run OCR first

A scanned PDF is a photograph in a PDF wrapper. There is no text in it, which is why you cannot search it, copy from it, or summarise it, and why a converter hands back a Word file containing one large picture. The fix is OCR, optical character recognition, which reads the picture and adds a real text layer.

• Ask for it explicitly: "This is a scan. OCR it into searchable text, then convert to Word." On Jobbit the same message does both steps.
• Check the numbers. OCR is very good on printed text and unreliable on handwriting, faint scans and figures inside tables. Ask the agent to flag low-confidence words, and compare totals against the original.
• Keep the scan. The searchable PDF is for working with; the original scan is the evidence.
• Free sites charge for this. Smallpdf and iLovePDF put OCR behind paid plans or page caps, Adobe needs an account, and PDF24 offers it free with a queue. If you have a folder of scans, the difference in time is large.

Extract instead of converting

Most conversions are a means to an end: you convert to Excel to add things up, to Word to reuse three paragraphs, to text to find one clause. An agent skips the conversion and does the end job on the PDF directly:

• "Summarise this 48-page report in one page, with the page number for each claim."
• "Find every date and deadline in this contract and put them in a table."
• "Compare the prices in these three supplier PDFs and tell me which is cheapest per unit."
• "Pull the invoice number, date, net, VAT and total from each of these 30 PDFs into one spreadsheet."

That last one replaces an afternoon of converting and copying, and it is the kind of routine job the deep research guide and the 12 jobs an AI agent can do for your business describe in more detail. General chat assistants such as ChatGPT, Claude, Gemini and Copilot can read a PDF and summarise it too; the difference with an agent is that it returns the finished file, and agents like Manus and Meta's Muse work the same way. Convert when someone else needs the file in another format; extract when you need the answer.

What the converter sites charge in 2026

If you prefer a single-purpose site, these are the limits and prices at the time of writing:

PDF24 is the best pure converter if you are happy with adverts and a Windows desktop app. For one file a month, Adobe's single free use is fine. For anything regular, scanned, or with tables that must survive, the agent route costs nothing on the free plan and returns a checked file. The wider market is compared in the best online PDF tools and the compressors head to head in Smallpdf vs iLovePDF vs Adobe.

Privacy: where your PDF goes

A converter site uploads your file to its servers, converts it and deletes it after an hour or two, according to its policy. That is acceptable for a menu and questionable for a bank statement, a contract or a medical letter. The desktop apps (PDF24 for Windows, Sejda's desktop version) keep files local. On Jobbit the file goes into your own workspace, the same place your chats and projects live, and stays there until you delete it; nothing is passed to a third-party converter. Whatever route you use, never convert a document containing card numbers or passwords on a site you have not read the terms of.

Mistakes that make a converted file useless

• Converting a scan without OCR and wondering why the Word file is one picture.
• Trusting converted numbers without checking a total. Converters drop minus signs and decimal points.
• Exporting JPGs at screen resolution, which fail the "must be legible" check on every portal.
• Sending a converted PDF that is now 30 MB because it was rebuilt from images. Compress it.
• Converting when you needed an answer. A summary or an extracted table is faster than a conversion you then have to read.
• Editing the only copy. Keep the original PDF untouched.

Frequently asked questions

How do I convert a PDF to Word for free?

Attach the PDF in a Jobbit chat and type "convert this to an editable Word document, keep the tables"; the free plan covers it with no card. Adobe Acrobat online converts one file per tool free before asking you to sign in, and PDF24 is free with adverts. For documents with tables or columns, ask for a check of the converted tables, which is where free sites go wrong.

Can I convert a scanned PDF to Word or Excel?

Yes, but the scan has to go through OCR first so the text is real text. Ask for "OCR, then convert to Word" in one message. Check names and numbers afterwards; OCR is accurate on clean printed text and less so on faint scans, handwriting and tables.

How do I turn a PDF bank statement into Excel?

Ask for the transactions extracted into a spreadsheet with named columns and the numbers as numbers, then check the totals against the statement. Free converters often return merged cells and figures stored as text, so the sum shows zero.

How do I convert JPG photos to one PDF?

Combine the photos into a single PDF in the right order, straighten the pages, and compress the result to the limit the recipient sets, usually 2 to 10 MB. Photos make large PDFs, so compression is part of the job.

Is it safe to convert a PDF online?

Converter sites upload your file and delete it after an hour or two; read the policy before uploading anything sensitive. Desktop apps and your own Jobbit workspace keep the file with you. Never upload documents containing card numbers or passwords to a site you have not checked.

Do I need Adobe Acrobat to convert PDFs?

No. Acrobat is the most complete editor, and worth about £14 to £20 a month if you edit PDFs daily, but conversion itself is free through Adobe's online tools once per file, through PDF24, or through an agent on the free plan.

What is the fastest way to get the numbers out of a PDF?

Do not convert it. Ask the agent for the numbers you need in a table, with the page they came from, and it reads the PDF directly, running OCR if it is a scan. Start at jobbit.uk with the file and the question.