Measured reference, not a feature tour

Organize PDF Pages: Reorder, Remove and Shuffle

Drag pages, delete them, shuffle them, or type an exact order, and know in advance which of those quietly drops pages you never asked to remove. Everything here comes from 19 runs of the live tool: the order box's real parsing rules, the duplex and booklet sequences that work, the 14 messages it can produce, and the byte-level cost of the export.

No upload Copied, not re-rendered 20-step undo
By Karan Kumar Updated November 1, 2026 25 min read
What this page covers

01Three Moves: Load, Arrange, Export

Everything in this tool is one of three actions. Nothing else changes your file.

01

Load the PDF

Drop the file on the zone, press Browse Files, paste with Ctrl+V, or pull it from Google Drive / Dropbox. The tool reads the page tree and starts rendering thumbnails one at a time as you scroll, so a 400-page file is usable before every thumbnail exists.

Upload dropzone in the AFFLIGO organize PDF tool
02

Change the order

Drag a card, or use one of the five buttons above the grid: Reverse, Shuffle, Odd First, Even First, Undo. Each card also carries rotate, duplicate and delete controls. There is no "apply" step - the grid is the document from that moment on, and every change pushes onto a 20-step undo stack.

Page card grid with the reverse, shuffle, odd and even buttons
03

Organize, then download

Organize builds a fresh PDF, copies each page in the order you set, and hands you yourfile_organized.pdf. The progress line stays honest about what it is doing - "Organizing page 7/12" - because that copy loop is literally what is running.

Organize button, download button and processing status after organizing the PDF

Two things surprise people on the first run. First, the buttons act on positions in the grid, so after a shuffle "delete page 3" means the third card you can see, not the page that used to be number 3. Second, the export does not edit your file - it copies pages into a new document, which is fast and lossless but drops anything that lives at document level rather than page level. That second point gets its own section, because it is the thing that catches people out.

If you only need to fix one page's orientation, the Rotate PDF tool handles it in one pass without the grid. This tool earns the extra step when the order itself is wrong.

02What the Export Keeps and What It Drops

Organize does not edit your file in place. It builds a new PDF and copies pages into it, which is why some things survive untouched and others disappear completely.

This is the part most guides leave out. I built a six-page file with a deliberately awkward structure: document title, author, subject and keywords all filled in, a two-entry bookmark outline, one fillable form field, one link annotation, and a single page that already carried a 90 degree rotation. Then I ran it through Organize with the order left alone and read the output back field by field.

PropertyBeforeAfter one Organize run
Page count and order6 pages6 pages, in the order shown in the grid
Per-page rotationpage 3 at 90°still 90°, and it travels with that page when you move it
Page content (text, images, vectors)original streamssame streams, no re-encoding
Link annotations1 link on page 2kept, and it follows the page it belongs to
Title, author, subject, keywordsall four setall four empty
Bookmarks (outline)2 entriesnone
Form fields1 fillable field0 fields: the widget is left behind as a dead annotation
Open action, page mode, page labelsnot setnot set, because nothing writes them back

The pattern is simple once you see it: everything that lives on a page is copied with the page, and everything that lives on the document is not. Your text, your images, a stamp you added earlier, a hyperlink someone embedded: all fine. The title that shows in a browser tab, the sidebar bookmarks, the fillable client-name field: gone the moment you press Organize.

Bytes: usually identical, sometimes 11x bigger

Copying pages sounds like it must re-encode something. It does not. A 12-page scan where every page carried its own JPEG came back 72 bytes larger than the input (1,498,520 to 1,498,592 bytes), with all twelve content streams and all twelve image streams byte-for-byte identical to the originals. The 72 bytes are the new file's cross-reference table and a short page index. Nothing was re-compressed, so the export cannot make your scan blurrier.

There is one case where the output grows a lot. A second 12-page file pointed all twelve pages at a single shared 125,791-byte image (129,726 bytes total on disk, because the image was stored once and referenced twelve times). Output: 1,515,500 bytes, about 11.7 times the input, because the copy loop writes that image into the new document once per page. This is not a bug you can work around with settings; it is what "copy each page into a fresh document" means. If you hit it, compress afterwards and the duplicate copies merge back down.

Before you press Organize

If a form field still needs filling, fill it first. If the bookmarks matter, note the chapter pages before you start, because you will be rebuilding them. If the file is a signed contract, keep the original: the signature panel and document metadata are the first things a rebuild drops.

03Watch a Full Run, Start to Finish

About twenty seconds, the real sequence: load, drag one page to the front, shuffle, delete, then organize and download.

SIMULATED SCREEN RECORDING • 0:20
Ready — hit play0:00 / 0:20

04Removing Pages, At One Page or Two Hundred

Deleting pages is the job people open a page organizer for. There are two ways to do it here, and one of them refuses to work when it should.

The slow way is the small X on a page card: one click, one page gone, and Undo brings it back. That is fine for a cover sheet or a stray blank. It stops being fine somewhere around page forty of a 200-page scan, which is where the Delete Pages box earns its place. Type the positions you want gone and it removes them in one action.

You type (6-page file)What actually happensToast
3Page 3 removed, five pages leftDeleted 1 page(s).
3,7-9Only page 3 removed; 7, 8 and 9 do not exist and are ignoredDeleted 1 page(s).
2,4Both removed, four pages left, original order keptDeleted 2 page(s).
1-6Refused. The grid is untouchedCannot delete all pages.
9-3Reversed to 3-6 and removed, so four pages remainDeleted 4 page(s).
emptyNothing happensEnter pages to delete.
0,-2,abcNothing happensNo valid pages found.

Two behaviours there are worth knowing before you rely on them. A range typed backwards (9-3) is silently swapped to 3-9 rather than rejected, so a fat-fingered range deletes more than you meant. And a file cannot be emptied: deleting every page is blocked, which is a guard rail rather than a limitation, because a zero-page PDF is not a file you can open again.

The refused case is also handled properly, which is not automatic. In my test I rotated page 1 to 180°, then tried to delete all six pages. The delete was refused and the rotation was still there afterwards, because the tool rolls back the change it did not complete instead of leaving the grid half-edited. Same rule for everything else on this page: nothing is written until you press Organize.

Positions, not page numbers

This trips up almost everyone, so it gets its own heading. Both boxes read positions in the grid as it looks right now. Reverse a six-page file first and the order becomes 6,5,4,3,2,1. Now type 1 in the Delete box and you do not delete the original page 1: you delete whatever is sitting first, which is the original page 6. Output: 5,4,3,2,1. The same applies to the order box: after that reverse, typing 1,2 gives you pages 6 and 5, not 1 and 2.

Practically, that means one operation at a time with a look at the grid between them. If you shuffle for a worksheet and then delete "page 2", you are deleting a page you have not looked at. Undo is the way out: twenty steps of it, covering reorder, delete, rotate and duplicate.

Finding the pages worth deleting

Blank pages from a duplex scanner arrive in pairs, and the same sheet of paper shows up twice. Scroll the grid in order and note the positions on paper first, then type them as a single comma list. On a 200-page file that is one action instead of two hundred clicks, and because the parse ignores anything that is not on the page list, an out-of-range position costs you nothing. If you are deleting to shrink a file rather than to clean it, deleting pages helps far more than compressing does, and the compress guide has the measured numbers for why.

05The Page Order Box, Exactly As It Parses

Typing a sequence is faster than dragging once a file passes about thirty pages. Here is what the parser does with every shape of input, including the ones it accepts without complaint when it probably should not.

The box takes comma-separated numbers and ranges. Anything that is not on the resulting list is dropped, which is the single most important thing to understand about it: you are not saying "put page 5 first", you are saying "the new document is these pages, in this order".

You type (6-page file)New orderWhat it tells you
1-3,51, 2, 3, 5Applied custom order: 4 pages. Pages 4 and 6 are dropped, not appended
3,1,23, 1, 2Applied custom order: 3 pages. Explicit order is respected exactly
1,1,11, 1, 1Duplicates are allowed: one page, printed three times
9-33, 4, 5, 6Applied custom order: 4 pages. A backwards range is swapped, not rejected
2,4-unchangedNo valid pages found. A dangling dash voids the whole input
0,99,abcunchangedNo valid pages found. Out-of-range and non-numeric entries produce nothing
 2 , 4 2, 4Applied custom order: 2 pages. Spaces around commas are trimmed
emptyunchangedEnter page order.

Notice the difference between the two failure styles. A partially valid list still runs, so 1-3,5,88 quietly gives you four pages and drops the 88. A list where nothing resolves is refused outright and your grid stays as it was, which is the safer half of the behaviour. The toast is your pre-flight check either way: read the page count in it against the file you expected before you hit Organize.

Recipes worth keeping

Two limits on the box. It reads the grid, not original page numbers, so a recipe stops matching once you have dragged or shuffled. And it has no memory: the box keeps your text for reference, but pressing it again applies the same list to the grid as it stands now, which after the first run is a different document. If you need to try two arrangements, run one, look at the toast, then press Undo before typing the second.

06Two-Sided Scans: Rebuilding the Real Order

A sheet-fed scanner that does not duplex gives you every front, then every back. Here is how to turn that back into a readable document without retyping forty page numbers by hand.

This is the most common mess I see, and it is worth being precise about which version you have, because the fix is different for each. Dump the grid and look at page 21 or so.

Build the list in a spreadsheet, paste it once

The order box needs an explicit list, and it will not interleave for you. Generating that list is a thirty-second job with two columns. For the 40-page two-pass case (fronts 1 to 20, backs 21 to 40 in reading order):

CellFormulaWhat you get
A1 down to A201, 2, 3, and so onThe front pages
B1=A1+2021, the back that follows front 1
C1, filled down=A1&","&B11,21, then 2,22, and so on
One cell=TEXTJOIN(",",TRUE,C1:C20)The whole string, ready to paste

For the flipped-stack version, replace B1 with =41-A1, which counts 40, 39, 38 downward, so back 40 lands after front 1. Same C column, same paste. Paste the result into the Page Order box, press the button, and read the toast before exporting: it should say 40 pages. If it says 20, the join cell only picked up one column and you have just deleted every back.

The odds-and-evens trap

Odd First and Even First reorder, they do not filter. On a six-page file, Odd First gives 1, 3, 5, 2, 4, 6 and Even First gives 2, 4, 6, 1, 3, 5. Both keep all six pages. That is exactly right for manual duplex printing, where you print the whole file twice with a different first page, and exactly wrong for a two-pass scan, where you need fronts and backs alternated one at a time rather than grouped.

When the count is odd

A two-pass scan of an odd number of sheets always ends with a blank back, and a reversed-stack version ends with that blank in the middle of the interleave rather than at the end. Rather than hunting for it by eye, delete it before you interleave: it is the last page of the file if the backs ran forwards, and the first back if they were flipped. Undo exists, so the cheaper test is to run the interleave, scroll to the seam in the grid, and fix the one page that looks wrong.

If your scanner writes each side as its own file instead of one PDF, stop here and merge first: Merge PDF in the same order you scanned, then come back to this page. The merge guide covers what merging does to page order, which matters for the same reason the grid reads positions rather than page numbers.

07Booklet and Imposition Sequences

Printing a folded A5 booklet from A4 sheets is a page-order problem, not a page-edit problem. The grid handles it, but only if you know the sequence.

When you fold sheets in half and staple the spine, page 1 and the last page end up on the same side of the same sheet. A printer does not know that. If you send it a document in reading order with "2 pages per sheet, duplex" ticked, you get a stack of unrelated spreads. What it needs is the physical order in which the pages pass through the tray: outermost spread first, innermost spread last.

For a document of n pages, the sequence is: n, 1, 2, n-1, n-2, 3, 4, n-3 and inward from both ends. Here it is worked out for the sizes people actually print. Type the whole string into the Page Order box, press the button, and check that the toast still reports the original page count before exporting.

PagesSheetsPaste this into the Page Order box
414,1,2,3
828,1,2,7,6,3,4,5
12312,1,2,11,10,3,4,9,8,5,6,7
16416,1,2,15,14,3,4,13,12,5,6,11,10,7,8,9
20520,1,2,19,18,3,4,17,16,5,6,15,14,7,8,13,12,9,10,11
24624,1,2,23,22,3,4,21,20,5,6,19,18,7,8,17,16,9,10,15,14,11,12,13

Every page appears exactly once, so nothing is lost and the toast count matches the input. After you export, the imposed file is no longer a document you can read on screen: page 1 of the PDF is the outside cover spread. Keep the reading-order original and treat the imposed copy as a print artefact. If you need to go back, Undo in the grid, or re-upload the original, because the imposition itself is just a page order.

Printing it

Send the imposed file to the printer with 2 pages per sheet, duplex, and scale to fit turned on. The one setting that varies by printer is which edge it flips on: half of them need flip on short edge for a landscape sheet, the other half want long edge. Print the first sheet only, fold it, and look at it before committing twenty sheets of card. If the backs are upside down, change that one setting and print the same sheet again.

Two limits worth stating plainly. This tool reorders pages, it does not impose them: the 2-up layout is the print driver's job, which is why the sequence above is a flat list rather than pairs. And n has to be a multiple of four, because a folded booklet always has four panels per sheet. A 14-page file cannot be saddle-stitched without two blank panels, so either add padding pages in another tool or print the final sheet separately.

Before you fold anything

Add page numbers before imposing, not after. Once the pages are in imposition order, a numbering pass counts the imposed sequence, so sheet 1 gets "1" even though it is physically the back cover. Number first, impose second.

08The Same Job in pdfarranger, Acrobat and Preview

Three desktop options, what each one is actually good at, and the one job the browser grid does better.

If you reorder pages twice a year, a browser tab is enough. If it is part of a weekly workflow, a desktop tool earns its place, mostly because it keeps the file on disk and can be scripted. Here is the honest split, not a sales page.

CapabilityAFFLIGO (browser)pdfarrangerAcrobatPreview (macOS)
CostFree, no accountFree, open sourcePaid subscriptionFree with macOS
InstallNoneNeeds installNeeds installAlready there
Drag to reorderYesYesYesYes, in the sidebar
Type an explicit page orderYesNoNoNo
Delete, rotate, duplicateAll threeAll threeAll threeDelete and rotate
Undo20 stepsEdits stay in memory until you saveYesYes
Imposition presetsTyped sequences (section 07)Booklet layout built inPrint dialog onlyNo
Also does OCR, redaction, formsNo, other toolsNoYesNo

pdfarranger is the closest free equivalent and the only one on this list with a real booklet layout feature, so if you print folded booklets every month, install it and skip my sequence table. Its weakness is the other end: it is a pure drag interface, so restacking a 200-page scan means dragging or cutting and pasting page by page, and there is no order box to type into. It is also the tool I would reach for if a file had to stay on a machine with no network.

Acrobat is worth the subscription only if you need what the subscription actually sells: OCR on scans, redaction that removes text rather than covering it, form authoring, and preflight for print. Page reordering is the free tier of that skill set, and paying $19.99 a month on an annual plan ($29.99 month to month) to drag a page out of a PDF is hard to justify. Standard sits at $22.99 month to month. Prices are Adobe's own listings, and they move, so check before you subscribe.

Preview on macOS is the one people already have open. Drag a thumbnail in the sidebar, press Backspace to delete, Cmd+S to save. It is genuinely fast for fixing one page in a ten-page file. What it will not do is accept a typed sequence, which means anything from section 06 or 07 turns into an hour of manual dragging.

The one case where the browser wins outright

Typed sequences. A 24-page saddle-stitch imposition is 24 numbers pasted once, or 24 drags in a desktop app where a single misdrop shifts everything after it. No desktop option on this list has an equivalent.

09Limits, and the Ones That Actually Stop You

There is no page-count limit and no account limit. What exists is a set of guards sized to your device, and they behave differently from the numbers most pages quote.

Four gates run in a fixed order every time you drop a file. Knowing which one fired tells you what to do next, because three of them can be worked around and one of them cannot.

GateTriggerWhat you see
Not a PDFAny other file typeToast: Please select a PDF file. Nothing loads
Device capFile larger than your tier allowsModal: File too large for this device, with your RAM, cores, tier and the safe limit, plus a Close button
Hard ceilingOver 1 GBToast: exceeds 1GB, too large
Page-count confirm300 pages or more on a computer, 150 or more on a phoneBrowser confirm dialog. Decline and the file is removed from the organizer

The device cap is the one people hit, and it is not a fixed number. It comes from a tier table: 30 MB, 60 MB, 100 MB, 200 MB or 300 MB per file, chosen from your reported RAM, your core count, and how full the browser's JavaScript heap currently is. That last signal is why the same laptop can accept a 150 MB file in the morning and refuse it in the evening after forty tabs: using more than about 70 percent of the heap drops you to the lowest tier regardless of hardware. The line under the dropzone that reads "Max N MB per file on this device" is not decoration, it is the exact figure the guard will enforce.

The 1 GB ceiling is a backstop, not a working limit. Every tier cap sits far below it, so in normal use you will never see that message; it exists for the case where the tier detection gives up entirely. In the same way, 300 pages is a confirm rather than a refusal, because thumbnails render only for the cards you scroll to: a 600-page file is perfectly usable provided you do not scroll it end to end at speed.

What is actually consuming memory

The PDF itself is the small half. A 40 MB scan renders around forty canvas thumbnails at preview width, and those bitmaps, not the file, are what fill the tab. That is why the guard is per file rather than per queue, and why the page-count dialog exists on top of it. If a large file feels sluggish, the fix is to work in passes: organize the first hundred pages, export, then do the rest, rather than loading everything at once.

Undo is capped at twenty steps, covering reorder, delete, rotate and duplicate. On a file you have shuffled five times, the twenty-first action quietly pushes the oldest state out of the stack, so the thing to undo is always the most recent mistake rather than the one from ten minutes ago.

Nothing is uploaded

All of the above happens in the tab: the PDF is read with pdf-lib and PDF.js from a local file handle, and the export is assembled in memory before the download starts. There is no request carrying your document, which is also why the limits are your device's limits and not a server quota.

10Message Decoder

Every toast and dialog this tool can produce, what triggered it, and the shortest way past it.

These are copied from the code rather than paraphrased, because "No valid pages found" and "Cannot delete all pages" look similar on screen and mean completely different things.

MessageWhat produced itWhat to do
Please select a PDF file.A file that is not a PDF, or a renamed oneCheck the extension and the file itself; a DOCX renamed to .pdf is still a DOCX
Failed to load PDF. Try a different file.pdf-lib parsed the header but the structure is brokenTruncated download, or a file exported by a broken generator. Re-export the source
File too large for this deviceThe tier cap in section 09The modal names your RAM and the safe limit. Split or compress the file first
Large document: N pages. Continue?300+ pages on a computer, 150+ on a phoneDeclining removes the file. Accept and scroll carefully
Removed large file from organizer.You declined that dialogNothing to fix; drop the file again if you changed your mind
Enter page order. / Enter pages to delete.Empty boxType a list first. Typing into the box does not apply anything by itself
No valid pages found.Nothing in the input resolved to a real positionAlmost always a dangling dash (2,4-) or numbers beyond the page count
Applied custom order: N pages.Success, and N is your new page countRead N. If it is lower than expected, pages were dropped rather than moved
Deleted N page(s).Success, counted after out-of-range numbers were ignoredIf N is smaller than the list you typed, part of it did not exist
Cannot delete all pages.The delete list covered the whole documentRemove the last remaining page by hand instead, or keep it
First click the Page Order or Delete Pages box, then click pages.You clicked a card expecting it to fill a box that was never focusedClick into the box, then click cards to append their positions
Page N already added.The same position clicked twice into the same boxHarmless. Each position is only listed once by clicking
Cancelled. Your file and page order are kept.Cancel pressed during the export loopNothing was lost. Press Organize again when ready
PDF Organization FailedAn exception during the copy loop, with the reason underneathRead the small line under the title. A password prompt cancelled mid-run lands here

One message deserves a note of its own: a password-protected PDF usually never reaches any of these, because the tool asks for the password in-page first and unlocks a copy in memory. Cancelling that prompt is the most common cause of the generic failure card, which is why the two are listed separately above.

Organizing pages from several files

This tool takes one PDF at a time, and the grid is the reason: each page card tracks a source page in one document. For a mixed document, merge first, then organize the merged file, then compress once at the end. Doing it in that order means one compression pass instead of three, and the merge guide explains what merging does to page order before you touch the grid.

11How This Was Tested

Every number and behaviour on this page came from running the tool and reading the output back, not from the documentation.

The method matters if you want to trust the rest of the page, so here it is in full. I built test PDFs with pdf-lib, drove the live tool in a headless browser, and then re-opened each download and measured it rather than judging it by eye.

The fixtures

The runs

Nineteen UI cases through the real interface: identity, Reverse, Odd First, Even First, Shuffle twice, Undo, two rotate-plus-export combinations, six page-order strings and three delete strings, plus the two "does it act on positions or page numbers" tests. Each run was a fresh page load and a fresh upload, so no state carried over.

Every download was then reopened and checked for: page count, page order (via the text-width trick), per-page rotation angle, annotation count per page, document metadata, outline presence, form-field count, file size, and whether the content and image streams were byte-identical to the source.

What was measuredResult
Identity run output size3,273 bytes from a 3,800-byte fixture
Reverse, six pages6,5,4,3,2,1 with the 90 degree page following its own page number
Shuffle, two runs6,5,1,4,2,3 and 3,2,6,5,1,4, so the order really is random per run
Page-order string that works1-3,5 produced four pages, not six
Page-order string that refuses0,99,abc left the grid untouched
Delete that is blocked1-6 refused, and a 180 degree rotation survived the refusal
Positions vs page numbersReverse then delete 1 removed the original page 6
Scan with one image per page1,498,520 bytes in, 1,498,592 bytes out, streams identical
Scan with one shared image129,726 bytes in, 1,515,500 bytes out

What I did not test

Being explicit about the gaps is more useful than pretending there are none. I did not test files above 200 MB, because the results would describe one machine's memory rather than the tool. I did not test a 1,000-page document end to end. I did not test every PDF flavour: files from LaTeX, InDesign and a few scanner apps behaved the same as the fixtures, but a heavily structured PDF with tagged accessibility content may drop more than the table in section 02 lists. And I did not test the password path beyond confirming that a correct password unlocks and a cancelled prompt lands on the failure card.

Two claims in earlier versions of this guide did not survive the testing, and both are corrected below rather than quietly deleted. That is the honest trade of publishing measurements: the numbers can contradict you.

12Corrections to Earlier Versions

Six things this page used to state that the measurements contradicted, left visible instead of quietly edited out.

A guide that never corrects itself is either very new or not checking. The previous version of this page was written from the feature list, which is how the usual claims get made: limit is 1 GB, metadata is preserved, invalid input is ignored. Running the tool showed four of those to be wrong and two to be true only in a narrow case.

What this page used to sayWhat the tests showed
Bookmarks and document metadata survive organizing, and links may not remapTitle, author, subject and keywords come back empty, and the outline is removed entirely. The one link annotation I planted did survive, so the old sentence had the two halves the wrong way round
Each file can be up to 1 GBYour device tier caps it at 30, 60, 100, 200 or 300 MB, and that cap is enforced with a modal. 1 GB is a backstop above every tier, so the limit you actually meet is a fraction of it
Invalid or out-of-range page numbers are ignored with a notificationTrue only while at least one entry resolves. A list with nothing valid in it is refused and nothing changes. The bigger correction is that the order box is a whitelist: pages you do not list are deleted, not appended
Odd First and Even First help with duplex-scanned documentsBoth buttons sort by position parity and keep every page: a six-page file becomes 1,3,5,2,4,6. That is right for manual duplex printing and wrong for a two-pass scan, which needs a typed interleave
File size barely changes after organizingTrue when each page carries its own image: 1,498,520 bytes became 1,498,592. False when one image is shared across pages: 129,726 bytes became 1,515,500, about eleven times larger
Deleting pages removes them from the arrangementCorrect, and worth restating because it is the surprising half. Nothing touches the file until Organize, and a refused delete rolls the whole change back rather than leaving a half-edited grid

The previous version also carried seven sections explaining what a page organizer is, why people need one, which operations exist, and what the tool's own feature list contains. None of that survived the rewrite, because none of it is something a reader cannot get from the first paragraph of the tool page. The space went into the measurements above instead.

If a claim on this page is wrong, the test that produced it is reproducible: build a six-page file with a unique text width per page, run it, and read the widths back. That is how every number here was produced, and it is the standard I would want applied to any guide that quotes figures.

13Frequently Asked Questions

Nine questions this tool actually gets asked, answered from the code and from the runs in section 11 rather than from the feature list.

Drop one PDF into the Organize PDF tool, then drag a page card by its grip or type positions into the Page Order box and click Apply. Reverse, Shuffle, Odd First and Even First each rearrange the whole document in one click. Nothing is written until you press Organize, so you can try a dozen arrangements before exporting one.
Quality does not change, because pages are copied rather than re-rendered: text stays text, images keep their original resolution, and rotation is written as page metadata. File size can surprise you. When every page carries its own image the output lands within a few bytes of the input, 1,498,520 to 1,498,592 in my test. When one image is shared across many pages, the rebuild has to duplicate it per page and the file grew from 129,726 bytes to 1,515,500, about eleven times larger.
Metadata does not. Title, author, subject, keywords and the bookmark outline all come back empty, so a downloaded file loses its document properties and its bookmark tree. Page-level content is a different story: the link annotation I planted survived and travelled with its page, and a page carrying a 90 degree rotation kept that rotation. If navigation matters, check the outline after export and rebuild it in a reader.
Click the X on a page card, or type positions into the Delete Pages box as 3, 7-9 and press Delete Selected Pages. It acts on positions in the current grid, not on printed page numbers, so delete after you reorder rather than before. The tool refuses to delete every page and rolls the whole change back, which also restores rotations that were part of that change. Deleted pages are simply left out of the export. If you want the kept pages as separate files instead, use the Split PDF tool.
Yes, and it is the fastest route on long documents. The Page Order box is a whitelist: 1-3,5 produces four pages, so anything you do not list is dropped from the export. Ranges work, a backwards range like 9-3 is silently swapped rather than refused, and a repeated number like 1,1,1 duplicates that page. When nothing in the box resolves to a real position it says No valid pages found and leaves the grid untouched.
Press Undo. The stack holds twenty steps and covers reorder, delete, rotate and duplicate together, so a mistake made ten actions ago is still recoverable. If the delete was refused rather than applied, nothing was lost and the grid is already back to its previous state. Closing the tab loses everything, because no state is saved anywhere, so work from the downloaded file rather than the open page.
Type an interleave, do not use the parity buttons. Odd First and Even First sort by position and keep every page, which is right for manual duplex printing and wrong for a two-pass scan. A six-sheet stack scanned all fronts then all backs needs 1,7,2,8,3,9 and so on, which you can build in a spreadsheet column and paste into the Page Order box. Section 06 has the formula for both stack orientations.
Your device sets the size cap, not a single site-wide number: 30, 60, 100, 200 or 300 MB per file depending on available RAM and core count, and the dropzone prints your exact figure before you upload. Anything above it gets a modal naming your limit rather than a silent failure. Page count is a soft gate: 300 pages on a computer or 150 on a phone triggers a continue prompt, and declining removes the file.
No. Parsing, thumbnails and the export all run in the browser tab through pdf-lib and PDF.js, so there is no upload step and no server copy to delete. It also means the tool keeps working offline once the page and its libraries are cached, and it means closing the tab discards the work.

Organize Your PDF Pages Now

Drag, delete, shuffle or type an exact order in one browser tab. The file is never uploaded and nothing is re-encoded, so the pages you keep come out exactly as they went in.

Open Organize PDF Tool

Key Takeaways

Karan Kumar, Founder of AFFLIGO

Karan Kumar

Founder & Web Developer at AFFLIGO

Karan built AFFLIGO's Organize PDF engine on pdf-lib, including the grid, the undo stack and the export loop. Every figure on this page comes from a fixture he built and a download he reopened, which is also how two claims in the earlier version of this guide were caught and corrected.

Last reviewed: November 1, 2026