Compare two PDFs by text

Provide the extracted text from each document as one string per page, and this tool reports exactly which page numbers differ. Each changed page receives a short, deterministic summary with word counts and the words added or removed. It also identifies pages that occur in only one version. The comparison runs without network access, preserves page order, and rejects either document when it contains no extracted text.

Prepare page text that preserves document structure

Start with text already extracted from both PDF files. Supply each document as an ordered array in which the first string is page one, the second string is page two, and so on. Keeping page boundaries intact is essential because the result reports differences by page number rather than treating the document as one continuous block. Text extraction tools may produce repeated spaces, tabs, or line breaks, so the comparison normalizes runs of whitespace before deciding whether page text changed. This avoids flagging a page merely because one extractor wrapped a sentence differently. It does not perform optical character recognition, open PDF files, or infer missing page boundaries. If a scanned document has no text layer, run OCR first and pass its page-level output here. Review extraction quality when fonts, columns, or tables produce unusual reading order. Both inputs must contain at least one page with non-whitespace text; an empty array, a missing input, or pages containing only whitespace produce an input error instead of a misleading equal result.

Understand how changed pages are summarized

The tool compares corresponding page positions after normalizing whitespace. Identical normalized text produces no difference entry. When a page changes, the summary tokenizes both versions into words, compares case-insensitive word frequencies, and reports how many words were removed and added. A brief preview lists up to five removed and five added words, making changes such as an updated amount, date, party name, or clause term easy to spot. Counts preserve repeated occurrences, so deleting the same term twice registers two removals. If the words are the same but their order, capitalization, or punctuation differs, the page is still marked as modified and the summary explains that word order or punctuation changed. This is a concise diagnostic rather than a character-by-character patch. It deliberately avoids guessing semantic importance or claiming that two differently written provisions mean the same thing. Use the listed page numbers to focus a human review, especially for contracts, financial statements, regulated documents, and layouts where a textual change may affect meaning.

Interpret missing pages and use the result safely

Documents do not always have the same length. When one input has additional page positions, each extra page is reported with a status indicating that it exists only in PDF A or only in PDF B, together with its extracted word count. The top-level result includes both page totals, an ordered list of differing page numbers, a detailed entry for every difference, and an equal flag for quick automation. This makes the output useful in release checks, approval workflows, archival validation, and document ingestion pipelines. A true equal flag means the normalized extracted text matched page by page; it does not prove that images, annotations, signatures, colors, fonts, metadata, hidden layers, or geometric placement are identical. Conversely, extraction differences can reflect changes in the PDF encoding rather than an intentional editorial revision. For high-stakes review, treat this report as a reliable way to locate textual differences, then inspect the original pages and apply the legal, financial, or operational judgment appropriate to the document.