Watching a page for changes, visually and in text

Capture a competitor's pricing page or a supplier's terms on a schedule, diff the Markdown to know what changed, and keep the image as proof.

Pricing pages, terms, status pages and feature lists change without announcements. Watching them by hand means remembering, which means not doing it.

A scheduled job with two requests does better.

DAY=$(date +%F)

curl -s -o "watch/${DAY}-pricing.png" \
  "https://api.screenshotline.com/take?url=https://example.com/pricing&full_page=true&cache=false" \
  -H "X-Access-Key: $KEY"

curl -s -o "watch/${DAY}-pricing.md" \
  "https://api.screenshotline.com/extract?url=https://example.com/pricing&cache=false" \
  -H "X-Access-Key: $KEY"

diff "watch/$(date -d yesterday +%F)-pricing.md" "watch/${DAY}-pricing.md"

Diff the text, keep the image

Image diffs tell you that something moved four pixels, which is usually a carousel. Text diffs tell you the price changed from $29 to $34, which is the thing you actually wanted to know.

So diff the Markdown and alert on that, and keep the image as the record of how it looked when it changed.

Making the diff quiet

Consent banners and ads are removed by default, which removes most of the noise before you start.

hide_selector anything that rotates: a testimonial carousel, a live counter, a "last updated" timestamp.

Cache off, because the point is today's version.

Daily is usually enough. Pricing pages do not change hourly, and a tighter schedule mostly buys you rate limits and noise.

Be a good citizen

You are fetching someone else's page on a schedule. Keep it to the pages you genuinely need, run it once a day rather than every minute, and expect that a site that does not want automated traffic will show you a bot wall. If it does, you will get a picture of the bot wall rather than a silent empty frame, which at least tells you to stop.