# Internet Archive Change Explorer

> See how a web page changed in the Wayback Machine: capture timeline plus the changes that matter. Lists the distinct Wayback Machine captures of a URL, picks the captures closest to two dates (or exact timestamps), fetches both raw archived versions and diffs them, ranking meaningful changes — prices, headings, legal/policy terms, dates, contacts, links — above noise. Paste CDX rows and two HTML versions to analyse offline.

Part of usefulweb.site (https://usefulweb.site/). Results are deterministic calculations from the inputs and the listed assumptions; the API returns exactly what the web page shows.

## Use it
- Human interface: https://internetarchivechangeexplorer.usefulweb.site/ (interactive tool at https://internetarchivechangeexplorer.usefulweb.site/tool)
- JSON API: POST https://internetarchivechangeexplorer.usefulweb.site/api/v1/calculate with a JSON object of inputs (schema: https://internetarchivechangeexplorer.usefulweb.site/api/v1/tool)
- OpenAPI: https://internetarchivechangeexplorer.usefulweb.site/openapi.json
- MCP (Streamable HTTP): https://internetarchivechangeexplorer.usefulweb.site/mcp - tool `internet_archive_change_explorer`

## Inputs
- url (url): Page URL
- beforeDate (date): Before date - The capture closest to this date is used. Default: one year ago.
- afterDate (date): After date - Default: today (latest capture).
- beforeTimestamp (text): Exact "before" capture (14-digit timestamp)
- afterTimestamp (text): Exact "after" capture (14-digit timestamp)
- maxCaptures (integer): Captures shown in the timeline
- maxChanges (integer): Max changes listed
- pastedCdx (textarea): Pasted CDX output (offline) - output=json rows or default text lines.
- pastedBefore (textarea): Pasted "before" HTML (offline)
- pastedAfter (textarea): Pasted "after" HTML (offline)

## Method
- Capture list: Wayback CDX API GET https://web.archive.org/cdx/search/cdx?url=…&output=json&fl=timestamp,original,statuscode,digest,length,mimetype&filter=statuscode:200&filter=mimetype:text/html, one capture per day (collapse=timestamp:8) — or per month for spans over 900 days. A new "version" starts whenever the content digest (SHA-1) changes.
- Capture selection: exact 14-digit timestamps if given, otherwise the successful HTML capture closest to each date, taken from the timeline query (from = before − 30 days, to = after + 30 days); only if that is empty does the tool fall back to CDX closest=…&sort=closest&limit=1.
- Archived pages are fetched raw (without the Wayback toolbar or URL rewriting) from https://web.archive.org/web/<timestamp>id_/<original URL>. Identical digests skip the fetch.
- The two versions are compared with the Website Change Explainer engine: block-level LCS diff, word diffs, and rule-based categories (price, legal/policy terms, contact, link target, title, date, heading, number) ranked above trivial noise (whitespace, timestamps, copyright years, tracking parameters).
- Similarity is the Dice coefficient over the words of the extracted page text (htmlToText).
- Caveats: digests also change for http/https/www variants, embedded tokens or ads; the archive may miss periods; content loaded by JavaScript after page load is usually not in the raw HTML.