The short answer is that no quality loss was measured when Toolit merged the ordinary PDF pages in this test. Think of it as placing existing sheets into a new binder rather than photocopying and reprinting them. Searchable text and embedded image data remained unchanged in the tested documents.
Other services may work differently. Results can change if a merger rasterizes a page—turning the whole document page into one pixel image—or applies image compression at the same time.
To check the actual Toolit path, we created three PDFs: a text-and-vector page, an image page, and a page containing both text and an image. We compared dimensions and searchable text, plus the pixel dimensions, data size, and SHA-256 hash (a digital fingerprint used to confirm identical data) of embedded images. We then copied the completed result into a new PDF five more times to look for accumulated change.
No quality loss was detected under the tested conditions
Both 1536×1024 images kept the same data size and digital fingerprint after the first merge and five additional merges.
Extracted searchable text and all three source page dimensions matched the merged document.
The result was 1,253 bytes smaller than the source total, while the image streams themselves remained identical.
We checked internal data, not only screenshots
- 1
Using pdf-lib 1.17.1, we created an A4 text-and-vector page, a landscape A4 page containing a 1536×1024 PNG, and a Letter page containing both text and the same PNG.
- 2
We followed the Toolit merger path: load each source, copy its page with copyPages, and save the three-page result with useObjectStreams enabled.
- 3
We compared source and output page dimensions and used PDF.js to verify that the same searchable strings could be extracted from each page.
- 4
For every Image XObject (image data stored inside the PDF), we compared pixel width and height, data size, and the SHA-256 digital fingerprint to detect recompression.
- 5
We copied all three output pages into a new PDF five additional times and repeated the page, text, and image-stream checks.
The three source PDFs
The test ran on August 24, 2026 with Node.js 24.14.1 and pdf-lib 1.17.1, the version pinned in Toolit. We generated the files specifically for the test so it can be repeated without personal documents or third-party copyrighted samples.
The image-only and mixed pages each contained one copy of the same 1536×1024 PNG. That gave us two source image streams that could be matched directly with the two streams in the output.
| Source | Page content | Page dimensions | File size |
|---|---|---|---|
| Text and vector | Searchable text, lines, rectangle | 595.28×841.89pt | 1,162 bytes |
| Image | One 1536×1024 PNG | 841.89×595.28pt | 2,097,784 bytes |
| Mixed | Searchable text and the same PNG | 612×792pt | 2,098,041 bytes |
Measured results before and after merging
The three source files totaled 4,196,987 bytes. The merged result was 4,195,734 bytes—1,253 bytes, or about 0.03%, smaller. That difference alone does not indicate compression or quality loss. Saving a new PDF can rebuild document packaging such as the catalog, object numbering, cross-reference data, and object streams.
The stronger evidence came from the image data stored inside the PDF. Before and after merging, each PNG-based Image XObject was 1536×1024 pixels and 2,096,814 bytes, with an identical SHA-256 digital fingerprint. Byte-for-byte equality means this path neither recompressed nor reduced the resolution of the test images.
| Check | Merged result | Meaning |
|---|---|---|
| Page count | 1+1+1 → 3 pages | All selected pages included |
| Page dimensions | All three matched | Portrait A4, landscape A4, and Letter preserved |
| Searchable text | Extracted strings matched by page | Text was not flattened into one page image |
| Embedded images | Dimensions, length, and SHA-256 matched | No re-encoding of the test image streams |
| Total file size | 1,253 bytes smaller | Document-overhead change, not image degradation |
Did five additional merges accumulate loss?
We copied the first merged result into a fresh PDF five times in succession. Every output remained 4,195,734 bytes, all page dimensions and image hashes matched, and the final document yielded the same extracted text.
Under this path, the embedded images were not repeatedly encoded like a JPG saved generation after generation. This is an observation about the generated samples and pdf-lib 1.17.1 copyPages behavior, not a guarantee for every editor or service that combines merging with compression.
Why a merged PDF can still look softer
A workflow that renders each source page to an image before building a new PDF can soften small text and lines depending on render scale and image compression. Quality may also have been lost before the merge when a source was converted to JPG, or after the merge when a separate file-size reduction step was applied.
Viewer zoom and screen antialiasing can make the same data appear different. Compare documents in the same viewer at the same zoom, then check text selection and searchability as well as image pixel dimensions before blaming the merge itself.
- Check whether the merger copies PDF pages or recreates them as images.
- Look for an enabled option that combines merging with compression.
- Inspect the source scan resolution and JPG quality before merging.
- Do not infer quality solely from a larger or smaller output file.
Document features need a separate check even when pages look identical
Matching appearance does not prove that every PDF feature survived. Bookmarks, document-level tagging, internal destinations, forms, and attachments can depend on structures outside the visible page and should be verified separately in the new document.
Digital signatures prove document integrity, so editing or rebuilding a signed PDF requires particular care. Keep the originals and recheck signature status, forms, bookmarks, links, and accessibility structure in the PDF viewer when those features matter.
Official documentation used for this analysis
Separate ordinary page merging from image recompression
In Toolit’s August 24, 2026 test, merging three different PDF page types and remerging the result five more times did not change page dimensions, searchable text, or embedded image data. No quality loss was detected in these samples with the current page-copy path.
The label ‘merge PDF’ does not reveal every service’s implementation. If an output looks softer, distinguish page rasterization (turning a whole page into a pixel image), separate image compression, and low source quality first. For specialized PDFs, verify signatures, forms, bookmarks, links, and accessibility structure independently of visual quality.