Anyone moving from the old version of Google Analytics to GA4 goes looking for average session duration first and cannot find it. It is not a bug and not a hidden menu. The metric no longer exists in its old form, and the reason for that is better than the annoyance it causes.
Why average session duration disappeared
The old duration was an estimate with a built-in hole. Analytics could only measure the time between two page views. The last page of a session had no second timestamp, so it counted as zero seconds. Someone who opened a page, read for four minutes and then left recorded a session of zero seconds. For pages people read to the end and then leave, average duration was therefore systematically too low, often by more than half.
GA4 measures something else: the time the tab was actually in the foreground. If your article sits in a background tab, the clock stops. That is more honest, and it makes the old and new numbers incomparable. A drop in duration after the switch is almost always a measurement effect, not a change in behavior.
What you look at instead: engagement time
The metric closest to the old duration is called average engagement time in GA4. It is not prominent, but it is there. The fastest way to it is the "Engagement" report under "Pages and screens": the table has a column for average engagement time per active user, and you can read it per page.
The reference point matters. GA4 has two similar values: engagement time per session and per active user. For the question "how long does someone spend with this page," you want per active user. Confuse the two and you will puzzle over numbers that differ by a factor of two.
Segment by landing page, or the number says little
A site-wide average across all pages is about as useful as the average temperature of a hospital. Engagement time only becomes interesting once you tie it to a single page and a search intent. Open "Pages and screens," sort by views descending and read engagement time row by row. How to reconstruct an actual behavior from these reports, instead of just staring at numbers, is in reading GA4 and in more depth in analyzing user behavior.
Only then can the number be read. Eight seconds on a page that hands over a phone number is a success. The same eight seconds on a guide that promised an explanation is a warning. Engagement time, like bounce rate, is not a grade but an observation you can only read together with the job of the page. Why that is so is in understanding bounce rate.
The quiet link to bounce rate
Engagement time and engagement rate are directly connected. A session counts as engaged in GA4 once, among other things, it runs longer than ten seconds in the foreground. Those ten seconds are the same threshold that decides whether a session counts as a bounce. Raise a page's engagement time and you almost automatically lower its bounce rate. Both values measure the same thing from two directions.
Frequently asked questions
Where do I find duration in GA4?
In the "Engagement" report under "Pages and screens." The column is called average engagement time per active user and replaces the former average session duration. There is no direct equivalent of the old metric anymore.
Why is engagement time lower than the old duration?
Because GA4 only counts the time the tab is active in the foreground, while the old version estimated time between page views and recorded the last page as zero seconds. The two numbers are not comparable; a drop after the switch is a measurement effect.
What is the difference between engagement time per session and per user?
Per session averages across all sessions, per active user across the individual people. For the question of how long someone spends with a page, engagement time per active user is the right metric.