Knowledge Architecture:ConceptsObservationsEvidence

Advertising Marginal Influence (AMI)

AMI

Experimental Counterfactual Measure of Advertising Effect on AI-Mediated Selection Probability

Experimental validationpublished

AMI measures advertising's causal effect on AI-mediated selection through experimental counterfactual comparison.

July 14, 2026
Version 1.0
8 min read
By HomeSelf Research
amiadvertising_effectivenesspersuasion_compressionexperimental_designai_selectionadvertising_marginal_influence

Definition

Advertising Marginal Influence (AMI) measures the causal effect of advertising exposure on AI-mediated selection probability. The canonical formula is AMI = P(choice | advertising exposure) - P(choice | no advertising exposure), where choice probabilities are measured under identical AI recommendations to isolate advertising effect from selection effects. AMI provides an experimental test of the Persuasion Compression hypothesis.

Advertising Marginal Influence quantifies how advertising exposure changes selection probability in AI-mediated discovery. By comparing selection probabilities with and without advertising exposure while holding AI recommendations constant, AMI isolates advertising's causal effect. AMI approaching zero would validate Persuasion Compression: advertising has no effect on AI-mediated selection.

Conceptual Formula

AMI = P(choice | advertising exposure) - P(choice | no advertising exposure). Requires AI Recommendation Control to ensure identical recommendations across exposure conditions. Measures selection probability difference attributable to advertising.

What This Index Measures

AMI measures advertising's causal effect on AI-mediated selection.

experimental confidence

By definition: AMI is the difference in selection probabilities with and without advertising exposure.

Implications

  • AMI provides a testable measure for Persuasion Compression

AMI approaching zero would validate Persuasion Compression.

experimental confidence

If advertising exposure does not change selection probability, AI-mediated discovery bypasses advertising influence.

Implications

  • Zero AMI would indicate advertising ineffectiveness in AI-mediated contexts

Methodology

Type

index construction

Data Sources

experimentalab testingai system logs

Confidence Level

experimental

Description

AMI = P(choice | advertising exposure) - P(choice | no advertising exposure). Requires AI Recommendation Control to ensure identical recommendations across exposure conditions. Measures selection probability difference attributable to advertising.

Limitations

  • Requires experimental design
  • AI system behavior may vary
  • Selection probability measurement requires instrumentation

Key Takeaways

Key Points

  • AMI = P(choice|ad) - P(choice|no ad)
  • Experimental counterfactual design
  • Tests Persuasion Compression hypothesis
  • AI Recommendation Control required

Target Audience

cmomarketing analyticsadvertisersresearchers

Relevance Tags

amiadvertising_effectivenesspersuasion_compressionexperimental_designai_selection

Source Paper

The Zero-Click Economy

HomeSelf Research (2026)

View on Zenodo
Read Paper

Citation

For Advertising Marginal Influence (AMI), see HomeSelf Research (2026), Digital Advertising Costs and AI-Mediated Discovery: An Evidence Synthesis on Zero-Click, Paid Media Dependency, and Customer Acquisition Economics.

DOI: 10.5281/zenodo.21360659

Download Options