AMP session-based metrics

Single sessions across multiple AMP pages

Accelerated Mobile Pages (AMP) is a standard for creating content that can be cached and served from any domain. For example, an article can originate on and also be served in an AMP viewer on

Because AMP content can be (and usually is) served on multiple domains, a user can have a single engagement with a business across multiple domains: for example, a user might first read an article from the AMP viewer on, and then click to another article on the site.

Cookie-based data collection for this 1 user consuming content behaves in this fashion:

  • Sets 2 cookies (1 on, 1 on
  • Collects 2 users (1 on, 1 on
  • Collects 2 sessions (1 on, 1 on

Learn more about Client IDs in AMP pages.

How AMP impacts measurement in Analytics

Any session-based metrics that you collect in Google Analytics from these pages are impacted by the distribution of content across domains and by the types of interfaces in which the content is displayed.

When a user has continuous interaction with two pieces of content from the same publisher but that continuous interaction is split across multiple domains (e.g., one piece of content viewed on and one piece viewed on, that one virtual session is recorded as multiple separate sessions, one for each domain. This results in:

  • Artificially higher user counts, session counts, and bounce rate
  • Artificially lower values for pageviews per session and session duration

For example:

  • A user interacts with 3 pages from the same publisher and spends 1 minute on each page
  • The first page is an AMP page on one domain like
  • The latter 2 pages are non-AMP pages on another domain like
  • Analytics records:
    • 2 sessions (1 on each domain)
    • 1.5 pageviews per session
    • 0.5 minutes per session (0 minutes for the first page since there’s no additional pageview on that domain, and 1 and 0 minutes, respectively, for the pageviews on the second domain, for a total of 1 minute)

What really happened was 1 user conducting 1 session that included 3 pageviews, with a session duration of 2 minutes (remember: 0 minutes for the 3rd page since there’s no subsequent pageview).

What you see in your data are 2 different users conducting 2 different sessions, with 1 and 2 pageviews respectively and an average session duration of 0.5 minutes . In addition, there is a bounce attributed to the first session since it contains just one pageview.

Was this article helpful?
How can we improve it?