Website Change Explainer

How it works

Give a "before" and an "after" version of a web page (fetch a URL now or paste saved HTML/text). The tool extracts readable blocks, title, meta description, headings, links, prices and dates, diffs them block by block and word by word, and ranks meaningful changes above noise like whitespace, tracking parameters, timestamps or copyright years.

• Each side is parsed (tolerant HTML tree builder; plain text is split into lines). Text is collected per block-level element (p, li, h1–h6, td, …); scripts, styles and navigation-less noise like <svg> are skipped.
• Blocks are aligned with a longest-common-subsequence diff over whitespace-normalised text; adjacent removed/added blocks with ≥ 40 % word overlap (Dice) are paired as "modified" and diffed word by word.
• A change is trivial when it disappears after masking whitespace/case, clock times, "N minutes ago", copyright years and long hex/random tokens; links differing only in tracking parameters (utm_*, gclid, fbclid, …) are trivial.
• Meaningful categories and scores: price 100, legal/terms keyword 90, contact (email/phone) 75, link target 70, title 65, date 60, heading 55, other number 45, meta description 40, link removed/added 35/30, other text 20 (+ up to 15 for size). Changes are ranked by score, then document order.
• Every change lists the rule ids and matched tokens that produced its category.

Rule-based comparison of static HTML. Pages that build content with JavaScript may show fewer blocks than a browser; always read the actual wording of legal changes.