People in a Facebook field experiment spent 6% more time on the platform when a browser extension hid 82% of the ads they would otherwise have seen, researchers report.
The experiment involved 1,810 users and also tested what happened when targeted advertisements were replaced with untargeted ones. It offers evidence about a particular group and platform, not a universal rule for every social network or advertising format.
The research, presented at a 2026 National Bureau of Economic Research conference, examines the amount of advertising people see as one cost of using a service funded by ads.
Two different changes to advertising
George Beknazar-Yuzbashev, Rafael Jiménez-Durán, Andrey Simonov and Mateusz Stalinski studied participants who installed a browser extension. One experimental condition hid nearly all ads; another replaced personalized ads with untargeted ones.
Personalized advertising uses information about a person or their activity to select an ad. Untargeted ads are selected without the same individual matching. The authors report that replacing targeted ads reduced ad clicks and longer-run engagement.
The comparison is relevant to marketers because showing fewer ads and changing which ads are shown are different interventions. A campaign may reach the same number of people but differ in whether its message matches their interests.
What time on the platform can and cannot tell us
The 6% change is an average result from the experiment. It does not mean that users enjoy every advertisement, that businesses should increase ad frequency or that all audiences react similarly. The experiment also does not establish the sales return from an individual campaign.
The researchers say that roughly two-thirds of variation in advertising load occurred across users. They interpret the pattern as consistent with platforms presenting different amounts of advertising to different users. The observed variation alone does not identify every reason behind a person’s feed.
The authors also use simulations to compare advertising-funded and subscription models. Such results depend on the model and assumptions; they are not direct observations of what would happen if a platform switched its entire business model.
What advertisers can measure
Businesses can compare frequency, clicks and subsequent customer actions rather than relying on impressions alone. A click is an expression of interest, not necessarily a completed purchase, and targeting may bring privacy and consent obligations that differ by jurisdiction.
Our earlier reporting on repeated marketing tests examined a separate provider and found that responses could decline with repeated messages. The two studies address different settings, making local testing more informative than assuming a single ideal advertising frequency.
