How to Reduce PDF File Size and Keep Text Readable
Reduce a PDF's size by checking what makes it large, removing unneeded pages and choosing a suitable approach for the images inside it. Compare the saved output with the source, especially small text and scanned images. A strict size target may require fewer pages or a better source rather than stronger compression. The iLikePDF Compress PDF page rebuilds the file structure but does not re-sample images, so it works best as a first step.
Every upload limit tells a story: an email cap, a portal that accepts 2 MB, a form that wants 500 KB. Meeting it is easy when you know where the bytes are. This guide gives a workflow you can repeat, and it points to the detailed article on why a compressed PDF may still be too large.
Understand why PDFs become large
Most of a PDF's size is images. Scanned pages and photographs dwarf text. Embedded fonts, attachments and unused objects add smaller amounts. A ten-page text document may be a few hundred kilobytes, while ten scanned pages can be tens of megabytes. That is why compression that only tidies internal structure achieves little on scans, and why reducing images is where the real savings are.
Read the destination upload rules
Before compressing, find the actual rule: the maximum size, the accepted formats, whether a page count is limited and whether text must remain searchable. Write it down. A vague goal like "make it smaller" leads to over-compression. A precise goal, such as "under 2 MB, readable at 100% zoom", tells you when to stop.
Choose a compression approach
Work from least destructive to most.
- Export again from the source with a "minimum size" or "reduce file size" option. This is usually the best result.
- Remove pages you do not need with Split PDF.
- Run the structural rebuild with Compress PDF. It copies pages into a new document and saves it with object streams; it never lowers image quality.
- Shrink the images before they become pages. If the PDF comes from photos, use Resize Image or Compress Image, then rebuild with JPG to PDF.
- Rescan at a lower setting, for example 150 to 200 dpi in greyscale.
- Rasterise and rebuild as a last resort: PDF to JPG, compress the images, then JPG to PDF. Text becomes part of the picture.
Compare size and readability
After every step, compare the new size with the previous one and open the file. Zoom to normal reading size and check small print, numbers, stamps and photographs. Compression damage appears first on fine text, thin lines and flat colours. Check every page, not only the first, and compare page counts.
Preserve searchable text and forms
Structural rebuilds keep real text as text. Rebuilding from images does not, so the result cannot be searched or read by screen readers. If searchability matters, avoid the rasterise route, or add text recognition afterwards; see scanned PDF versus text PDF. Forms and links may not survive a rebuild, so test them. Flatten a form first if a fixed appearance is acceptable; see what flattening does.
Troubleshoot files that will not shrink
| Symptom | Likely cause | What to try |
|---|---|---|
| Size barely changes | Image data dominates | Reduce images at the source |
| Size grows | Re-encoding at higher quality | Keep the original; try another route |
| Text looks blurry | Over-compression | Use a higher-quality setting, or fewer pages |
| Limit still exceeded | Too many pages | Split into parts if allowed |
| Scans are huge | 600 dpi colour | Rescan at 200 dpi greyscale |
Some limits simply cannot be met by compression alone. If a portal wants 200 KB for a twenty-page scan, ask whether fewer pages, a summary or separate uploads are acceptable.
Document a repeatable size test
Keep a short note of what worked: the source, the steps, the settings and the result. For a recurring task such as monthly invoices, a written routine means you never have to rediscover it. A useful format is: original size, after removing pages, after each step, final size, and the limit. If you compress images, also record the dimensions you used. This is the kind of small record that turns a frustrating afternoon into a five-minute job next time.
For image limits in KB, see how to reduce image size to 50 KB, 100 KB or 200 KB and the image size guide.
A worked example
A small company has to submit a twelve-page scanned tender document through a portal that accepts files up to 5 MB. The scan is far larger. The team follows the ladder in order. They remove two blank pages and a fax cover with Split PDF. They run Compress PDF and see only a small change, which confirms that images dominate. They rescan the pages at 200 dpi in greyscale using the office scanner. The new file is well under the limit and the small print is still readable at normal zoom. They open every page, check the page count and upload. They record the scanner setting in a shared note for next time.
The lesson is that the decisive step was changing the source, not applying stronger compression to a heavy file. The same pattern holds for photographs and presentations.
Quick decision table
| What you have | First thing to try |
|---|---|
| A PDF exported from a document | Export again with minimum size |
| A PDF made from phone photos | Resize or compress the photos, rebuild with JPG to PDF |
| A scanned paper document | Rescan at 150 to 200 dpi greyscale |
| A long PDF where only part is needed | Extract the pages with Split PDF |
| A PDF near the limit | Try the structural rebuild with Compress PDF |
| A PDF far over the limit | Reduce images, or send in parts if allowed |
Habits that prevent oversized PDFs
- Scan at a sensible resolution from the start. Higher is rarely better for reading.
- Crop photos to the page before building a PDF.
- Use greyscale for text-only pages.
- Keep source documents and export fresh PDFs rather than re-processing old ones.
- Avoid attaching files inside PDFs unless necessary.
- Name files by purpose and date so that you know which version is the smallest final copy.
Small habits like these keep files light and make later size limits far less stressful. If a file must also be numbered, labelled or signed, do those steps after the size is settled, and check the size again at the end, since stamping adds a little.
Frequently asked questions
They often store page-sized images rather than compact text objects. Resolution, color depth, page count and image encoding all contribute to size.
Record original and output bytes, then calculate the reduction as the difference divided by original size. Report it only alongside readability and output-function checks.
Read the destination’s maximum size, accepted format and page requirements. Remove unnecessary pages before sacrificing the quality of content you must keep.
No. Source documents differ substantially. Describe results from specific tested samples instead of presenting one percentage as universal.
Their contents may already be stored efficiently. Rewriting them may save little or even add overhead, so compare the actual results.
Process copies of the same source under identical conditions. Compare size and inspect representative text, photos, diagrams and searchable content.
Zoom in on fine print, handwriting, checkboxes, barcodes and identification numbers. A small file is not useful if the recipient cannot read necessary details.
Recognition generally benefits from a clear source. Keep the original scan, verify OCR output and only then optimize a reviewed copy if needed.
It can be when the recipient accepts multiple files or only needs selected pages. Do not split a submission that explicitly requires one complete document.
Lower resolution removes detail, but usefulness depends on the content. Choose the lowest setting that still preserves the required reading and printing quality.
Often it offers little benefit and may add degradation. Start from the original for each trial rather than compressing a previous lossy output repeatedly.
Keep the smaller valid version. Growth indicates that the applied method was not effective for that source, not that the larger file must be used.
Finishing content and size adjustments first is usually simpler. Apply protection to the final copy and verify access afterward.
Use a non-confidential sample, record its characteristics, date, browser, settings and observed output. Label the result as a sample test rather than a guarantee.
Re-export the source more efficiently, remove optional pages or ask about an accepted alternative submission route. Do not send unreadable content merely to satisfy a byte limit.