Guide · 7 min read
How to Rotate or Remove Pages from a PDF
Straighten sideways scans, cut unwanted pages, and reorder a PDF in your browser — then add page numbers as the finishing touch.
Last updated: 2026-10-03
Tools for this guide
Survey the document before changing it
Page surgery goes wrong when people operate on the wrong pages, so start by reading the document rather than editing it. Open the PDF, note the total page count, and write down which pages are sideways, which are unwanted, and what the final order should be. Working from that short plan — rotate these, remove those, move this one — turns three anxious operations into one calm pass each.
The tools help with the survey: after dropping in a file (one PDF per run, up to 100 MB), they display the page count and render thumbnail previews for the first 60 pages, so you can point at pages visually instead of guessing numbers. Everything runs locally in the tab, which is convenient twice over — large files process without uploading, and sensitive documents never leave the machine while you rearrange them.
Keep a backup habit for the ten seconds it costs: duplicate the original file before your first operation and leave it untouched until the final result is verified. Every tool here creates a new download rather than overwriting, but a separate pristine copy protects against your own mistakes, such as deleting page 7 when you meant page 9.
Straighten sideways and upside-down pages
Sideways pages are the classic scanning artifact: a landscape receipt fed through a portrait feeder, or a phone photo taken in the wrong orientation. The Rotate PDF tool fixes these by writing rotation into the file itself, so every app displays the correction — unlike viewer-only rotation that resets when the file is reopened elsewhere. Choose a clockwise turn of 90, 180, or 270 degrees; the angle adds to each page's existing rotation, so a page already turned 90 degrees needs only one more 90-degree pass to stand upright.
Control is per-page through the pages field, which accepts individual numbers and dash ranges such as 2,5-7. Leaving the field empty rotates all pages at once — useful for a uniformly upside-down scan batch — while typing a range targets only the offenders in a mixed document. Entries are validated as you type, so a typo like a page number beyond the document's length shows an error instead of producing a surprise file.
The download arrives with -rotated appended to the original filename, keeping it distinct from the source. Verify by reopening: flip through the previously broken pages and confirm text lines run horizontally and navigation thumbnails look right. If one page still leans, run just that page again — rotation composes cleanly across passes because each run simply adds to the stored angle.
Delete the pages you do not want to keep
Deletion day arrives with every long document: blank filler pages, duplicate scans, draft appendices, or the cover sheet the sender should have removed. The Delete Pages tool presents the document as a clickable thumbnail grid with a live count of how many of the total pages are selected, plus a text field accepting comma-and-dash ranges like 1-3, 7 that stays in sync with the grid — click thumbnails, type numbers, or combine both. The action button names the stakes explicitly, reading Delete N pages for your exact selection.
Two guardrails keep this safe. First, the tool refuses to remove every page — a selection covering the whole document explains that at least one page must be kept, which catches the classic mistake of selecting pages to keep while in delete mode. Second, the result is a genuinely new file: only the kept pages are copied into it, so removed pages are gone from the download rather than hidden. The filename gains a -cleaned suffix, distinguishing it from both the original and any extraction.
Know the sibling operation for the inverse job: where Delete removes the selected pages and keeps the rest, Extract keeps the selected pages and drops the rest, producing a file suffixed -extract instead. Removing three bad pages from a fifty-page report calls for delete; pulling one chapter out of the same report calls for extract. Choosing correctly the first time avoids a second pass.
Put the surviving pages in the right order
Wrong-order pages plague double-sided scanning, merged meeting packs, and any stack that fell off a desk. The Reorder tool shows the same thumbnail grid with drag handles and arrow controls: drag pages into place with the pointer or nudge them with the keyboard, and the interface tracks whether the order actually changed — the apply button stays disabled with a move-at-least-one-page note until you alter something, so accidental no-op runs cannot happen.
Think in moves rather than in final positions: pull the misplaced signature page down to the end, swap the two transposed chapters, slide the appendix after the index. Each change is visible in the grid before anything is written, which beats editing blind page numbers. Applying the order rebuilds the document in the new sequence and downloads it with a -reordered suffix, leaving the source file untouched for comparison.
A frequent scenario is the interleaved duplex scan, where odd pages run 1,3,5 and evens follow 6,4,2 — reordering weaves them back into 1,2,3,4,5,6 in a single session. For very long documents, combine techniques: reorder the section you can see in the previews, then verify the result's full page sequence before sharing, since one misplaced page in a hundred is easy to miss at thumbnail scale.
Add page numbers as the finishing touch
A rearranged document without page numbers invites confusion — references like see page 12 stop working the moment pages move. The Add Page Numbers tool stamps printed numbers onto the finished file as real added text, which means the document's original text stays selectable and the numbers themselves behave like text rather than burned-in image pixels. Pick from formats showing a bare number, the word Page before it, or a current-of-total style; choose any of six positions from top-left to bottom-center; set the starting number for front-matter offsets and adjust the font size to suit the layout.
Run numbering last, after rotation, deletion, and reordering are all settled — numbers must reflect the final sequence, and renumbering after another reorder wastes a pass. Then perform the full verification flip: every page upright, no unwanted pages remaining, order matching your plan, numbers running unbroken from first to last. Open the result in a different viewer or on a phone if one is handy, since a second renderer catches orientation and layout issues familiarity hides.
Archive deliberately when the document passes: keep the verified final alongside the pristine original with filenames that state what each is, and share only the final. Future edits start from the final for small fixes or from the original for structural rework — either way, nobody reconstructs your rotate-delete-reorder session from memory.
Fix mixed-orientation batches without touching clean pages
Mixed-orientation documents punish blanket operations, so isolate the offenders with surgical targeting. Consider a thirty-page field report where pages 6, 7, and 19 arrived sideways from a portable scanner while the remainder reads upright: rotating everything would straighten three pages and topple twenty-seven. Drop the single PDF onto the Rotate tool — one file per run, up to 100 MB, processed locally — and wait for the page count to appear with the Pages field hinting Leave empty to rotate all N pages alongside a ranges example such as 2,5-7. Type 6,7,19 into that field and choose a clockwise step of 90, 180, or 270 degrees; entries validate as you type, so a transposed digit or a number beyond the document length triggers an inline error before anything executes rather than emitting a mangled file.
Understand the rotation arithmetic and the workflow stays predictable across passes. Each run adds the chosen angle to the stored rotation of every targeted page instead of overwriting it, which means a page already carrying 90 degrees needs only one further 90-degree pass to reach 180, and a uniformly upside-down ten-page scan batch is repaired by leaving the field empty and applying 180 once. Viewer-only turns in desktop readers create the classic false fix — upright on your screen, sideways on the recipient machine — whereas this tool writes rotation into the file itself so every renderer honors it. Downloads arrive with -rotated appended to the original stem, keeping generations distinct without manual renaming, and reopening the result for a thumbnail scan confirms text lines run horizontally before you proceed.
Handle the awkward middle cases with staged passes rather than heroics. Duplex scans where odd leaves face north and even leaves face south resolve into two runs: rotate the evens 180 with a range covering those pages, verify, then address leftovers individually. Phone photographs interleaved with exported pages often need 90 clockwise on some sheets and 270 on others — separate runs per direction, since one angle per run keeps the mental model clean. For documents beyond the 60-page thumbnail preview window, typed ranges still address any page by number, so page 87 rotates as reliably as page 8 even when its thumbnail never renders. When a single sheet still leans after a batch, re-run just that number; composed angles accumulate cleanly, and the pristine source remains untouched for comparison until the straightened copy passes inspection.
Combine delete, extract, and reorder into one tidy pass
Deletion and extraction are mirror twins that newcomers constantly interchange, so choose by intent before clicking. The Delete Pages interface presents a clickable thumbnail grid with a live N of total selected counter plus a text field accepting comma-and-dash expressions like 1-3, 7 that stays synchronized with the grid — click faces, type numerals, or blend both techniques while the action button narrates the stakes as Delete N pages. Removing filler, duplicates, draft appendices, or a mistaken cover sheet from a fifty-page report calls for delete: the kept remainder flows into a fresh file suffixed -cleaned, with removed sheets absent rather than concealed. Pulling chapter four out of that same report for standalone circulation calls for extract instead, which keeps the selection, discards the rest, and downloads a -extract file. Naming the job correctly on the first attempt spares an entire second pass.
Respect the guardrails and the pass stays boring in the right way. Selecting every page trips an explicit refusal explaining at least one page must be kept, which intercepts the classic polarity error of marking keepers while in delete mode. Selections of zero pages simply leave the action disabled, so stray clicks cannot manufacture empty documents. Both operations copy kept pages into a newly built file without re-rendering content, meaning surviving text stays selectable and vectors stay crisp at any magnification — weight leaves through omitted sheets, never through degraded survivors. Keep the untouched original archived until the trimmed download opens cleanly at full magnification, then re-run with corrected numerals if page 9 vanished when page 7 was meant to go.
Reordering closes the triage sequence for documents whose sheets survived but shuffled. The Reorder view shares the thumbnail grid vocabulary with drag handles plus keyboard-accessible arrow controls, and the apply control stays disabled behind a move-at-least-one-page notice until the sequence genuinely differs — accidental no-op runs cannot occur. Think in discrete moves: drag the orphaned signature leaf to the terminus, swap transposed chapters, slide the appendix behind the index, watching each relocation reflect in the grid before committing anything to disk. Applying rebuilds the document in the displayed sequence as a -reordered download while the source file lingers for side-by-side comparison. Finish with the survey ritual: every sheet upright, no strays remaining, order matching the written plan, and for numbered documents a final pagination pass afterward so printed numerals track the settled sequence rather than the chaos you inherited.
Repair a shuffled meeting pack from survey to handoff
A canonical mess — the quarterly board pack scanned double-sided in a hurry — exercises every tool in this guide in one sitting and rewards methodical sequencing. The stack arrives as forty-odd sheets where the agenda sits behind the treasurer annex, two resolution pages face sideways, a duplicate catering invoice pads the middle, and the attendance folio hides near the terminus. Begin with the survey ritual on paper: record total count, circle sideways numerals, cross out discards, and sketch the intended collation of agenda, minutes, financials, resolutions, annexures, and sign-off. Duplicate the source file into an archive folder first, since every downstream operation mints fresh downloads and the pristine scan remains the court of final appeal when a folio goes astray.
Execute rotation first while thumbnails still mirror the original chaos. Drop the file into the Rotate interface, confirm the count, and straighten only the sideways resolutions with a targeted range while upright minutes stay untouched; remember angles accumulate onto stored orientation, so an already-quarter-turned sheet needs a single further quarter rather than a full revolution. Next open the Delete view and mark the duplicate invoice plus blank separator leaves through the synchronized grid and range field, watching the live selected-of-total tally and the Delete N pages narration before committing to a -cleaned download. Then weave the survivors into collation order with drag moves and arrow nudges — recto before verso, resolution before ballot, appendix after index — until the grid mirrors the paper plan, and apply the sequence into a -reordered file whose pagination pass stamps numerals last so references like see folio 12 survive circulation.
Handoff verification separates a repaired pack from a reshuffled one. Flip the finished file end to end at comfortable magnification checking orientation, continuity, and unbroken numerals; open it on a handset or secondary viewer where unfamiliar rendering exposes lingering skew; confirm recto-verso spreads pair correctly for duplex printing and staple placement avoids obscuring marginal folio marks. Share only the verified derivative with a filename stating meeting, date, and version, retain the archived source beside it, and log the transformation — rotated sheets, discarded numerals, relocation moves — in a one-paragraph cover note so the secretariat can audit the surgery without replaying it. Future amendments start from the repaired file for minor corrections or from the source for structural recuts, and nobody reconstructs the session from recollection.
Work confidently beyond the thumbnail preview window
Length changes tactics once documents outgrow the sixty-page preview strip, so shift from visual pointing to numeric navigation with deliberate verification. Catalogues, theses, deposition transcripts, and annual reports routinely run into triple digits, and their thumbnails render only the opening stretch while typed ranges address every folio by numeral regardless of visibility. Build a written manifest first: upright chapters, sideways exhibits, redundant dividers, transposed appendices, each noted as numerals and dash expressions such as 88-94 rather than circled thumbnails. Process rotation numerically against that manifest — empty field never, targeted ranges always — and confirm each committed angle by reopening the -rotated derivative and sampling the repaired folios at magnification instead of trusting memory of the click.
Deletion and extraction scale gracefully under the same discipline because both accept synchronized grid-or-typed selection with live tallies and explicit narration. Excise printer calibration leaves, duplicated signature blocks, superseded errata, and placeholder dividers through typed expressions, heeding the refusal that blocks whole-document erasure and the disabled state that prevents zero-page runs. For anthology building — assembling a course reader from scattered chapters or a diligence bundle from disparate filings — extract each source separately into -extract derivatives, then merge those vetted intermediaries rather than wrestling one gargantuan selection string across fragile originals. Retain every source file until the consolidated volume passes the continuity read described below, since storage costs pennies while reconstruction costs afternoons.
Verification at scale demands sampling strategy plus full-sequence machinery. Complement the thumbnail skim with extraction spot checks: run the finished file through text extraction with page markers and scan the transcript for vanished headings, duplicated numerals, or orphaned footnotes that thumbnails conceal. Cross-reference the manifest line by line — every planned rotation present, every condemned sheet absent, every relocation reflected — and countersign the manifest with verifier initials and date for collaborative matters. When pagination crowns the workflow, stamp numerals only after structural edits settle, then audit the run for skips and doubles across the entire extent rather than the previewed prefix. Archive the manifest beside the source and the derivative so future custodians inherit evidence of diligence instead of an inscrutable stack of suffixed filenames.
Frequently asked questions
Will rotating pages reduce their quality?
No. Rotation rewrites each page's rotation value in the file rather than re-rendering content, so pixels and vectors stay exactly as they were — every viewer simply displays them turned.
Can I rotate just one page of a long document?
Yes. Type the page number or a range such as 2,5-7 in the pages field; leaving it empty rotates every page. Invalid ranges show an error before anything runs.
What if my PDF has more than 60 pages?
Thumbnail previews render for the first 60 pages, but typed page ranges accept any page in the document — so rotation and deletion by page number work on longer files regardless of previews.
I deleted the wrong pages — can I undo it?
The original file is never modified; deletion builds a new file containing only the kept pages. Keep the original until you verify the result, then simply re-run with the corrected selection.
Should I rotate pages before deleting the ones I do not need?
Either order produces a correct file, but surveying first, then rotating the keepers, then deleting the rest keeps the thumbnail grid readable throughout. Numbering always comes last so printed numbers match the final sequence.
Try it now — free, no signup
Fix sideways scans in seconds. Files stay on your device.
Open Rotate PDF