Key takeaways
- Every page monitor runs the same loop: fetch, extract, normalize, compare, judge, notify.
- Normalization removes tokens, clocks and ad slots, which are the differences that carry no meaning.
- A threshold or a judge decides whether a real change is worth a notification, so the rule behind it matters more than the diff format.
Tools detect web page changes by running one loop: fetch the page, keep a copy, compare the new read against the old one, then decide whether the difference deserves a ping. Every product in this category runs that loop. The disagreements start at the compare step and at how much judgement sits between a difference and a notification.
What changed is that the fetch can run a real browser and the compare step can describe the delta in a sentence instead of handing you a line of HTML.
The loop, stage by stage
- Fetch. A plain HTTP client is cheap. A headless browser costs more and sees what a person sees, including prices built after load.
- Extract. Hash the document, pull one element by selector, or read the rendered text. AyeWatch takes the last route, which is why a rule can be written in ordinary words.
- Normalize. Session tokens, rotating ad ids, cache busters, server clocks and view counters create differences that mean nothing.
- Compare. A hash is exact and says nothing about what moved. A text diff names the lines. A DOM diff names the node. Screenshots compare pixels. The artifact is what you pay for.
- Judge. A threshold, a model, or both.
- Notify. Email, chat, push, webhook, or an entry in a feed.
Where the noise gets in
| Difference in the page | Common cause | What a naive diff does |
|---|---|---|
| A random string changes on every load | session or cache token | fires on every check |
| A clock in the footer moves | server time | fires on every check |
| An ad slot renders differently | third party script | fires at random |
| The price line changes | a real edit | the one you wanted |
Three of those rows are noise, and no regex removes all of it across thousands of sites.
A threshold is a policy, not a setting
Set a pixel threshold low and a carousel rotation wakes you. Set it high and a digit change in a price slips past. Text and element diffs carry the same trade-off in different units.
A judge moves the question from how much changed to whether this is the kind of change you asked about. AyeWatch's version is a plain-English rule plus a summary that names the delta, and the rule is the policy. Say "tell me when the price changes" and a moved footer stops mattering.
Test one awkward URL
Put two tools on the same page, change nothing, and note which one fires. Then edit a page you control and read the alert. One that names the line saves you a click; one that says the page changed hands the work back.
Build or buy
A cron job with curl and diff is a real option if you control the network and can live with HTML level noise. The recurring cost is not the script. It is the headless browser, the retries, the selector that breaks, and the inbox nobody reads.
AyeWatch is the hosted version of the loop, with paid plans from $9 a month billed in monitoring credits and a 7 day trial on the web. Push and the feed come with every plan; email, Slack, Discord, webhooks and the API need a paid one. None of them will say whether the change matters to your business, which is why the alert text is the part worth judging.
FAQ
Why does a monitor show an empty page?
Usually because the content is built in JavaScript and the tool fetched the raw HTML. A headless browser fixes it when the site allows one.
Is hashing a page enough?
It detects that something changed and says nothing about what. Fine as a checksum, poor for a person reading the alert.
Do I need to write a selector?
Not with AyeWatch, which takes a rule in plain English. Selector tools still win for one exact element inside a complex app.
Frequently asked questions
Why does a monitor show an empty page?
Usually because the content is built in JavaScript and the tool fetched the raw HTML. A headless browser fixes it when the site allows one.
Is hashing a page enough?
It detects that something changed and says nothing about what changed. Good for a checksum, poor for a person reading the alert.
Do I need to write a selector?
Not with AyeWatch, which takes a rule in plain English. Selector tools still win when you must track one exact element inside a complex app.
Sources
- AyeWatchAyeWatch
- Web page change monitoring: tools comparedAyeWatch
- AyeWatch pricingAyeWatch
