How do I compare AI referral behavior across devices in GA4?

The Citead team, updated

In short. A free-form exploration compares AI referral behavior by applying a session acquisition segment and grouping session metrics by Device category.

Definition: Comparing AI referral behavior across devices in Google Analytics 4 (GA4) means applying a session segment for selected AI assistant acquisition sources, then breaking down its metrics by Device category in a free-form exploration. A session segment selects visits that meet its conditions.

Key facts

  • As of 17 October 2026, Google's Segment builder documentation describes session segments as subsets of sessions, including sessions originating from a particular campaign.
  • Google's Tech details documentation defines Device category as the type of device on which activity originates, with values including desktop, mobile and tablet.
  • Google's free-form documentation provides separate settings for segment comparison, table rows and metric values.
  • Google's engagement documentation defines engagement rate as the percentage of sessions that qualify as engaged sessions.

Which acquisition segment belongs in the comparison?

An acquisition segment selects data using information about where visits originate. For this comparison, its inclusion conditions select the recorded sources identified as AI assistant traffic. Google's traffic-source scope documentation explains that Session source describes acquisition for new sessions, including those of returning users. First user source describes the source that initially acquired a user and keeps its value when that user returns.

The proposed setup therefore uses a session segment with Session source conditions. The source values come from the property's recorded acquisition data. Google's segment builder supports dimension conditions with a comparison operator and values; alternative source conditions can use OR logic within a condition group. Its example shows that a user segment can include activity from several visits, while a session segment returns activity from qualifying sessions.

How is the device table configured?

Google's free-form instructions start at Explore, under Start a new exploration, with the Free form template. Table is the default visualization. The following configuration applies the selected acquisition segment to a table whose rows represent device categories:

Explore > Free form
Visualization: Table
Segment comparison: selected AI referral session segment
Rows: Device category
Values: Sessions, Engaged sessions, Engagement rate, Key events

Device category and the selected metrics are added to Variables and placed in the corresponding Tab Settings areas. Segment comparison receives the AI referral session segment. The same free-form documentation supports nested rows, so Session source can be added beneath Device category to inspect the acquisition sources within each device group.

Google's Explorations guide places the date selector at the top of Variables. A custom date range remains fixed when the exploration is reopened. The Export data control provides CSV export for the data available in the selected visualization.

What do the device differences measure?

Sessions supplies the visit count behind each row. Engagement rate expresses the share of engaged visits, allowing a comparison alongside those counts. Google's engagement criteria qualify a session when it lasts longer than 10 seconds, contains a key event, or contains at least 2 screen or page views.

A key event is an event marked as important to the business. Google's Tech details report defines Key events as the number of times users trigger such an event. That metric counts occurrences; a device group with more sessions can also accumulate more occurrences. The report defines engagement rate as engaged sessions divided by total sessions.

FAQ

Can I compare individual assistant sources within each device category?

Yes. You can add Session source as another row dimension. Google's free-form documentation supports nested rows for grouping dimensions hierarchically.

Does this table follow someone from a phone to a desktop?

The table groups recorded activity by device category. Google's traffic-source scope documentation explains that identifying an individual user depends on the property's reporting identity and the user's browser or device settings.

Sources

  1. Google, "Get started with Explorations", accessed 17 October 2026
  2. Google, "Segment builder", accessed 17 October 2026
  3. Google, "Scopes of traffic-source dimensions", accessed 17 October 2026
  4. Google, "[GA4] Free-form exploration", accessed 17 October 2026
  5. Google, "Tech details report", accessed 17 October 2026
  6. Google, "[GA4] Engagement rate and bounce rate", accessed 17 October 2026

Check how a page reads to answer engines. Open the GEO score analyzer

Test a page and see what to fix first

The analysis usually takes less than 10 seconds and needs no account.

400,000+ pages scored in the public benchmark