Agentic Marketing: Ads for Whom?

How to cite

Jabarian, Brian; Imas, Alex; Bursztyn, Leonardo. 2026. "Agentic Marketing: Ads for Whom?." AEA RCT Registry. AEARCTR-0018404. https://doi.org/10.1257/rct.18404-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Brian Jabarian* , Alex Imas , Leonardo Bursztyn

* Primary investigator

Registry ID
AEARCTR-0018404
DOI
https://doi.org/10.1257/rct.18404-1.0
First registered
Last updated
Status
In development
Trial dates
2026-05-01 – 2026-05-15
Countries
Private
JEL codes
D91, C93, L15, D18
Registry record
http://www.socialscienceregistry.org/trials/18404

Abstract

We study whether common marketing techniques -- such as featured badges, default pre-selection, and information obstruction -- cause people to choose objectively worse options, and whether AI agents are similarly affected. Each participant makes 20 incentivized choices between pairs of simple lotteries, where one option strictly dominates the other. Half the choices use neutral formatting; the other half apply one of five marketing techniques to the dominated option. We recruit 200 participants on Prolific (US, English-speaking) and compare their choices with those of AI language models presented with identical stimuli. The primary outcome is whether the participant selects the dominated lottery. The primary specification tests whether marketing techniques increase dominated choices for humans relative to AI, using a within-subject design with pair and participant fixed effects.

Intervention

Within-subject experiment. Each participant makes 20 incentivized binary lottery choices. Ten choices use neutral formatting (control). Ten choices apply one of five common marketing techniques to one option (salience, highlight, confirmshaming, preselection, obstruction). The marketing technique-to-pair assignment and the block order (control first vs. marketing first) are randomized per participant.

Experimental design

Within-subject design. 200 participants on Prolific (US, English-speaking). Each participant sees 10 lottery pairs twice: once under neutral control formatting and once with a marketing technique applied. Items are organized into two blocks (control block, marketing block), with the block order randomized across participants. Items within each block are randomized. Marketing technique-to-pair assignment is randomized per participant (2 pairs per technique, balanced). Payment: $3.00 base + one randomly selected lottery resolved for a real bonus ($0.00-$2.60).

Design details

Randomization method
Computerized randomization within the oTree survey platform
Randomization unit
Individual participant (block order randomized between participants; item order and technique-to-pair assignment randomized within participant).
Clusters
200 participants (not clustered)
Observations
4,000 lottery choices (200 participants x 20 choices each).
Arms
Within-subject: all 200 participants see both conditions (2,000 control observations, 2,000 marketing observations).
Primary outcomes
Dominated option chosen (binary: 0/1). Whether the participant selects the strictly dominated lottery on a given choice.

Cited by

No citation data available for this registration yet.

Keywords