Clarify end-of-session event guidance for sendBeacon() - #45081

Preview URLs (1 page)

• /en-US/docs/Web/API/Navigator/sendBeacon

(comment last updated: 2026-08-12 18:22:28)

Reading this back I think we also ought to expand a bit on pagehide in the previous section - with the following section removed, we are left with just one mention of pagehide, which appears kind of hanging. We could perhaps rename this section to something like "Legacy events", and add a paragraph at the end saying something like:

The pagehide event is not as problematic as unload or beforeunload, because it is compatible with the bfcache. However, it suffers from the same reliability problems as the other two events, so we recommend that developers should avoid it as well, and use only visibilitychange.

I wonder if we also ought to acknowledge the problem with visibilitychange: that it will fire in situations that are not what we might think of as the session end, such as the user switching to a different tab.

I'd also be happy if @philipwalton would like to confirm that this is still the best guidance for developers.

Thanks, this makes the guidance clearer. I updated the previous section to note that visibilitychange also fires when the user switches tabs, and added an explicit explanation that pagehide is bfcache-compatible but still unreliable for end-of-session analytics. I kept the existing heading so the published #avoid_unload_and_beforeunload fragment does not break. The branch is rebased and the focused pre-commit, URL/xref, formatting, Markdown lint, front matter, and spelling checks pass.

I think my version of the last paragraph here is better.