What is really happening when the change does not show
You publish a price, correct some text or replace an image. You see the new version, but a customer still sees the old one. First confirm you are opening the same address and that the change is published; then check whether a cache is serving an older copy.
Between your server and your customer's screen there are several intermediate copies. Each stores a version of your site for minutes or hours to serve it faster. We call them cache layers, usually four: the visitor's browser cache, the CDN or proxy in front of your server, the CMS cache (WordPress and its cache plugins) and, sometimes, a service worker, a browser program that can store responses even if the website has not been installed as an app.
Distinguish stored copies of content from DNS responses, which tell the browser which server to contact. After a DNS change, some visitors may still reach the former server while retaining an older response. This is often called propagation: the symptom may look similar, but purging a page cache does not change visitors’ DNS responses.
The principle: before purging at random, look at which layer is serving the old version.
Before touching anything: reproduce the problem properly
The first temptation is to hit the «clear everything» button. Hold on. If you purge without knowing which layer was serving the old version, you lose the trail and, when it happens again, you will be just as blind. Before acting, separate three views: yours, your customer's and an anonymous visitor's.
- Open the affected page in a private or incognito window, not signed in. The closest thing to seeing your site as a stranger does.
- Try another device and network. A different result may involve the browser, session, DNS or responding CDN server. It does not prove the fault is on the customer’s device.
- Ask the customer for the exact URL and a screenshot with time and browser. «The change does not show» is ambiguous; «on
/pricesI see 29 EUR at 10:14 on Chrome on Android» is useful. - Write down the publication time, the URL and what changed (a text, an image, a price, the whole page).
With that small record, diagnosis stops being guesswork. The same preserve-before-acting principle we use in other incidents: a sober note saves half an hour of going in circles later.
The browser cache: the first suspect
Your browser keeps copies of files it has downloaded (HTML, CSS, JavaScript, images) so it does not ask the server on every visit. That browser cache is the first layer to rule out because it lives on the visitor's device.
The classic test is the hard reload: on Windows and Linux press Ctrl + F5; on Mac press Cmd + Shift + R. That tells the browser to download the files again, ignoring its local copy. If the change appears after a hard reload, record the difference. Caching or response revalidation may be involved; compare the other checks before confirming the cause.
If site data needs clearing, distinguish the cache from cookies, sessions and local storage. Clearing everything may sign you out or remove saved work. Your team can inspect the cache through browser tools; there is no need to clear data for every website.
What decides how long the browser keeps each file is a header called Cache-Control, configured by whoever maintains your website on the server. Values such as max-age=3600, no-cache or must-revalidate change the behaviour, and touching them is work for whoever administers your server, not from the CMS panel. For each directive, the MDN reference for Cache-Control is where technical teams go back to.
| Test | What it tells you if it works |
|---|---|
Normal reload (F5) | The server returns the current version and the browser uses it. |
Hard reload (Ctrl + F5) | The local copy was outdated. Likely a browser cache issue. |
| Private window | If the versions differ, compare session state, extensions and browser data. The difference guides the investigation but does not establish the cause on its own. |
| Another device | A different result helps comparison. The same result does not rule out local caching or identify a specific cause. |
If these tests make the problem disappear only for you, do not close the case yet: your customer may keep seeing the same thing until a higher layer is purged.
CDN, proxy and edge: the layer between your server and the street
A CDN (content delivery network) is a layer of servers spread around the world that keep copies of your site close to each visitor. Cloudflare, Fastly, Akamai or the host's own edge are common examples. A reverse proxy such as Varnish or Nginx does something similar in front of your server: it serves cached responses before bothering the origin.
The key point: even if you press Ctrl + F5, if the CDN has stored the previous version, that is what it returns. You can reload a thousand times and keep seeing the old version while the CDN does not serve the new one.
Whoever maintains your website can inspect response headers and hosting settings to identify intermediaries. A header such as cf-cache-status or x-cache may indicate whether that layer found a copy (HIT) or requested a response from the next layer (MISS). Another cache may sit behind it; a via header alone does not show whether content was reused.
Once you confirm there is a CDN, do not flush the whole site. Ask for a per-URL purge of the page that changed. A full purge forces the CDN to request every resource from the origin again and can spike server load at the worst moment. References such as the web.dev guide on HTTP caching explain why.
Some CDN services include filtering or other protections, depending on the plan and configuration. For the wider picture, how HTTPS, WAF and CDN combine to protect your site.
Your platform's own cache: WordPress and cache plugins
If your site runs on WordPress, Prestashop, Drupal or similar, there is another common layer: the CMS cache. The CMS generates the HTML of each page from the database when the visitor arrives; because that work is expensive, the result is stored as a static page to serve it fast next time.
On WordPress this is handled by plugins such as WP Rocket, LiteSpeed Cache, W3 Total Cache or WP Super Cache, and many specialised hosts add their own cache on top (SiteGround, Kinsta, Hostinger and others). It is normal to have several stacked caches: the plugin, the host, an object cache and sometimes PHP's OPcache. Purge only one and the others may keep serving the old version.
Some cache configurations exclude signed-in users. In that case, you may see a change through your account while a signed-out visitor receives the former copy. Confirm whether your website has this exclusion before attributing the difference to it.
Where is the purge button?
- On the WordPress admin toolbar a button of the active cache plugin usually appears («Clear cache»).
- In your hosting panel, under performance, there is usually another purge option worth running too.
- For small changes, look for a per-URL purge instead of flushing the whole site.
Many of these symptoms appear right after a CMS or plugin update. If that was your case, the guidance on what to do when WordPress stops working after an update may help.
Service workers and installed PWAs: the most confusing case
Some websites use a service worker: a browser program that can return stored copies. It can exist even if you have not installed the website as an app. Its storage is not automatically shared with another phone or a fresh private session.
If the issue only affects one browser profile, your team can inspect “Application → Service Workers” in Chrome and check which version controls the page. If it also appears on other devices, investigate other layers too; an old program on one phone cannot explain every result.
Updates must coordinate code and caches to avoid mixing versions. skipWaiting() is not a universal fix: it can activate new code while an old page remains open. The service worker lifecycle guide explains that risk. Ask whoever maintains your website to test the update process.
When it is not caching: DNS propagation, domain and email
Sometimes the cause is not caching. If you recently moved hosting, moved the domain or updated DNS, what you see may be propagation.
DNS connects your domain to the server that should answer. DNS responses are cached for a period indicated by their TTL. After a record change, some visitors may retain an older response. There is no universal 48-hour deadline: the changed records, previous values and resolvers all matter.
After a DNS change, some people may see the new site while others see the former one. Cached responses or incorrect configuration may explain the difference. The team should compare records before recommending that you wait.
The same happens with email if you changed MX records: some messages may land in the old mailbox for hours because the sender still resolves the previous MX.
Simple checks whoever maintains your website can run:
dig yourdomain.comornslookup yourdomain.comfrom a terminal, to see which IP answers right now.- Online tools such as whatsmydns, which query the DNS from several countries and show where the change has propagated.
After a DNS change, ask your team to compare authoritative records with responses reaching different visitors before deciding to wait. Incorrect configuration does not fix itself with time. The guide to assessing a hosting move covers checks after a move.
The order to purge and what to ask whoever maintains your website
With several stacked layers, purging at random has two problems: it hides the real cause and, with a CDN, it can spike server load. A reasonable order, not a universal recipe, usually works:
- Confirm the change is published. Check the page was saved and the content in the editor is the new one. Sometimes the draft was not published or an approval flow stayed halfway.
- Purge the CMS cache for the affected URL. First in the cache plugin, then in the hosting panel if it has its own.
- If you use a CDN, request a per-URL purge and give it a few minutes. Do not flush the whole site unless you need to.
- Re-check in a private window on another device. If the new version appears, retain which caches were purged and the response now received. The change appearing does not by itself establish which layer had retained it.
- If it persists, ask whoever maintains your website to check whether the website uses a service worker and which version is active. The website does not need to be installed as an app to use one.
- If you moved hosting or changed DNS, confirm the records first. Purging page caches does not change DNS responses stored by other resolvers.
To get help, collect the exact address, the published change and two screenshots showing the difference. Whoever maintains your website or our team can then compare which version is served at each point and agree a specific intervention.
Our performance page explains how we review loading speed, Core Web Vitals and the way your website responds.
If you want us to walk you through it
If the old version keeps appearing, tell us which page changed and what differs. We can help locate the layer involved and agree a change with whoever maintains your site.
If your intention is not a one-off case but to understand how speed, caching and server configuration interact before hiring anyone, the performance area gathers the related pieces, including whether changing hosting will speed up your site.