Ad refresh

The research question

This study tested four ways of loosening the Standard’s current ad-refresh rule: refreshing after an unfilled ad bid, refreshing to a different ad size, refreshing every in-view ad unit rather than just one, and combining all three. Each was compared against the current rule using a live A/B test of real user behaviour — rather than a panel reacting to mock webpages — to see whether any of these changes affects publisher revenue or the user experience.

Method and sample

Approximately 9.9 million ad-blocking users were assigned to each of five groups (one control and four experimental) in a live A/B test run from 1 May to 1 June 2026, covering more than 320 million aggregate page views. The four experimental groups each tested a different relaxation of the current ad-refresh rule: refresh after an unfilled ad bid, refresh to a different ad size, refresh of every in-view ad unit, and all three changes combined.

Main findings

• —Across all four tested groups, no group showed a statistically significant increase in ad-blocker uninstalls or in ad-related issue-report submissions, compared with the control group.
• —The strongest revenue effect came from combining all three changes — refreshing after an unfilled bid, refreshing to a different ad size, and refreshing every in-view ad unit — into a single “all changes” configuration: a statistically significant gain of +16.5% in publisher revenue, robust across multiple statistical tests. Refreshing every in-view ad unit alone produced a smaller but still statistically significant gain of +4.5%. The two remaining single-change configurations — refreshing after an unfilled bid alone (+0.6%) and refreshing to a different ad size alone (+2.1%) — produced smaller gains; the ad-size change was itself statistically significant but neither cleared the report’s own 4% threshold for a practically meaningful effect.
• —The report’s own recommendation is preliminary, not final: it requests Committee approval for the all-in-view and all-features variants, and requests more data before any decision on the size-flexibility change alone. It records no recommendation to change the Standard for the unfilled-bid trigger, finding no deviation from the current rule.
• —The Committee approved this change on August 20, 2026, following the public comment period, which closed September 20, 2026 without objection. The change is now in effect; the Committee will continue to monitor its effects for six months, through March 20, 2027, to ensure no unanticipated user impact.

These are the report’s own findings and its own preliminary recommendation, not by themselves a Committee decision. The Committee separately reviewed and approved the change, as described above.

Limitations and context

• —Page views per session were measured only as an aggregated, per-website-day average rather than at the user level, so they could not be tested for statistical significance; the report treats this figure as descriptive only.
• —AA opt-out survey responses could not be linked to individual A/B test groups during this trial, so opt-out reasons are not broken out by group.