How AFileFix's Document Comparison works

Upload two versions of a document — PDF, DOCX, or one of each. AI analyzes both versions and identifies what changed: sections that were added, sections that were removed, and sections that were modified with the wording reworded, numbers updated, or terms swapped.

Results render on-screen as a list of changes you can read and review. A downloadable DOCX report contains the full analysis structured for sharing.

Processing takes 20-40 seconds for typical documents. Both files are forwarded to Anthropic's Claude API for comparison and deleted from our server within minutes after processing. Anthropic retains API data for up to 30 days for safety review, then permanently deletes it; data is never used to train AI models.

What counts as a change?

Document Comparison classifies every detected change into one of three categories:

Added — Content present in the revised version but not in the original. New sections, new paragraphs, new clauses. Useful for: identifying what was inserted, catching scope additions, spotting new terms a counterparty quietly added.

Removed — Content present in the original but absent in the revised version. Deleted sections, removed paragraphs, dropped clauses. Useful for: catching what was cut, identifying scope reductions, spotting terms that disappeared.

Modified — Content present in both versions but with reworded text, changed terms, updated numbers, or restructured phrasing. The before/after wording shows what specifically changed. Useful for: catching subtle revisions, identifying number changes (dollar amounts, dates, percentages), spotting term-of-art swaps.

Sections that appear unchanged in both versions are counted but not detailed individually. The summary at the top tells you how many sections were analyzed and how many had changes.

How the comparison report is structured

The on-screen results and the downloadable DOCX report share the same structure:

• Summary — Counts of added, removed, and modified sections at the top
• Added sections — One card per addition with the new content and where it appears
• Removed sections — One card per removal with the deleted content and where it appeared
• Modified sections — One card per modification with before and after text side-by-side
• AI explanations — Each change includes a plain-English summary of what changed and why it might matter

The explanations are descriptive, not prescriptive. AI tells you “the indemnification cap changed from $50,000 to unlimited” — it doesn't tell you whether that's good or bad for you. Interpretation is your job.

Mixed-format comparison

AFileFix handles mixed-format pairs: original as DOCX and revised as PDF, or vice versa. Internally the AI normalizes text content from each format and compares at the content level, so formatting differences don't get flagged as content changes.

This is useful because real-world workflows often mix formats — your original draft might be DOCX (still being edited), while the version you got back is PDF (signed or finalized). You don't need to convert formats first; upload them as-is.

A consequence worth knowing: layout-only changes (font, spacing, page breaks, margins) are not detected. The comparison focuses on text content. If you need to detect visual layout changes, this isn't the right tool.

What types of documents work best?

Best results: Structured prose documents with clear sections. Contracts, policies, proposals, terms of service, manuals. Text-heavy documents where content matters more than visual design.

Modest results: Documents with significant tables, formulas, or numbered lists. The AI can usually detect content changes within these structures, but accuracy varies based on how tables and lists are encoded in the source files.

Not supported well:

• Documents with heavy formatting where layout carries meaning (designed marketing collateral, infographic-style reports)
• Scanned PDFs where text is stored as images — Document Comparison doesn't OCR
• Mixed-language documents (the AI normalizes for English; non-English content extracts poorly)

What if my document is very long?

Documents over approximately 150 pages get truncated for comparison. The first portion is fully compared; content beyond that point isn't analyzed. The DOCX report includes a warning when truncation happens.

If you need to compare full content of long documents, the workaround is to split the documents into sections first using AFileFix's Split PDF or by exporting each chapter as a separate file, then compare each pair separately. It's more work but it's the only way to ensure full coverage.

The 150-page threshold isn't arbitrary — it reflects how much content the AI can hold in a single comparison context while still producing useful analysis. Pushing past it would degrade quality across the entire comparison, not just the tail end.

What Document Comparison doesn't do

Worth being explicit about a few things this tool isn't:

• Not a side-by-side visual diff. Tools like Beyond Compare or git's diff render documents side by side with colored highlights. AFileFix produces a structured change list, not a visual diff.
• Not Microsoft Word track changes. The output is a new DOCX report describing changes, not an annotated version of the source documents with track-changes markup.
• Not a 3-way merge. Compare two versions per run. For three versions (A, B, C), compare A vs B, then B vs C, then synthesize manually.
• Not in-document comparison. This tool compares two separate documents. If you want to compare two versions of the same content within a single document, you'd need to extract each version first.
• Not legal redline output. The plain-English explanations summarize changes for human readers; they're not formatted as a redline for legal review.

Frequently asked questions

For AI-processing details and Anthropic's API retention terms, see our Privacy page.