Back/forward cache: the free speed win many sites block
The back/forward cache restores pages instantly when a visitor hits Back. Here is what blocks it, how to test it in Chrome, and how it touches Core Web Vitals.
Most speed advice is about making a page load faster. The back/forward cache skips the load entirely. When a visitor presses Back, the browser can restore the page exactly as they left it, in a blink. Many sites quietly switch this off without knowing.
What the back/forward cache is
The back/forward cache, or bfcache, is a browser feature. When someone navigates away, the browser keeps a frozen snapshot of the whole page in memory, including its JavaScript state. Press Back or Forward and it unfreezes that snapshot instead of fetching and rendering the page again.
This is different from the HTTP cache. The HTTP cache stores files, but the browser still has to parse, run scripts, and paint. The bfcache stores the finished page. Nothing reloads.
It is common in practice. Think of a shopper who opens a category page, clicks a product, then goes back to the list. With bfcache they land on the same scroll position with filters intact. Without it, the list reloads and they lose their place.
Why it matters for SEO
Be careful about the claim here. Google has not said that bfcache eligibility is a ranking factor. The link to SEO is indirect, and you should treat it that way.
- Visitors who go back to a result list and pick another page have a better experience when your list restores instantly.
- Restores are near-instant navigations, so they can pull down your slow-page numbers in field data. How each tool counts them varies, so check the documentation for the one you use.
- Faster return trips mean fewer abandoned sessions on browse-heavy pages like categories, search results, and article lists.
In other words, this is a user experience win that may help your field data. It is not a shortcut to rankings. If your Core Web Vitals problem is a slow first load, bfcache will not fix it.
What blocks the cache
A page is only eligible if the browser can safely freeze it. Several common patterns say no. The exact rules differ between browsers and change over time, so test rather than assume.
- An unload event listener. This is the most common blocker, often added by old analytics or chat scripts. Use pagehide instead.
- A Cache-Control: no-store header on the main page. Chrome has been working on allowing some of these cases, but it is still a frequent cause of failures.
- Open connections such as WebSockets, WebRTC, or an in-flight fetch the browser cannot pause.
- A page that holds a reference to another window through window.opener.
- Third-party iframes or scripts that use APIs the browser treats as unsafe to freeze.
Note that no-store is often added by default on pages with a login or cart. Sometimes that is right. Often it is applied to every page by a blanket server rule, and a plain article gets caught in it.
How to test your pages
- Open the page in Chrome and open DevTools.
- Go to the Application panel and find Back/forward cache in the sidebar.
- Click Test back/forward cache. Chrome navigates away and back, then reports whether the page was restored.
- If it fails, read the listed reasons. Each one names the blocker.
- Fix the top reason, then run the test again. Repeat until the page restores.
Test your real templates, not just the home page. Run it on a category page, an article, and a product page. The blockers often come from one script that only loads on some templates.
Chrome also exposes a not-restored-reasons field in its performance API, which lets you log failures from real visitors. This is a Chrome feature, so treat the data as a sample of Chrome users, not everyone.
Fixes that usually work
Start with the unload listener. Search your code and your third-party tags for it. Swap it for pagehide, or ask the vendor for an updated script. If a tag has no fix, weigh whether it earns its place.
Next, review your caching headers. Keep no-store for pages that show private data. For public pages, a normal Cache-Control value is enough. Then close connections when the page is hidden and reopen them when it is shown again.
One catch: a restored page is old. Data like a cart count or login state may have changed while the visitor was away. Listen for the pageshow event and refresh anything time-sensitive when the page was restored.
What to do next
Pick your three highest-traffic templates and run the DevTools test on each this week. Remove any unload listeners first, since they are the easiest win. Then audit your no-store headers and keep them only where the page truly needs them. It costs little and helps every visitor who hits Back.