Skip to main content
Models: 9
Dimensions: 26
Trials: 56,640
Pre-registered: osf.io/et4nf

Machine Likeability Intelligence

Empirical measurement of how AI agents evaluate content for purchase recommendations.26 dimensions measured across 9 frontier models.

Key Findings
85%

of signals work cross-model

Universal optimization is viable

r = .03

correlation with human behavior

AI ≠ human psychology

±22 pts

model-specific swing

When models diverge, it matters

26
Dimensions
9
AI Models
57K
Trials
0.75
Peak Effect (h)
Filter by context:
Total TrialsPooled Data
56,640

A/B preference comparisons

Strongest Effect
h = 0.75

Ethical Concern Weight

Model Similarity
9399%

Genome cosine similarity range

What is APIS?

The AI Purchase Intelligence System (APIS) is a pre-registered research study measuring how frontier AI models respond to different content signals when making purchase recommendations. Using a forced-choice A/B methodology, we quantify the causal effect of 26 distinct content dimensions on AI preference.

This research provides empirical guidance for businesses optimizing content for AI-mediated commerce. As AI agents increasingly influence purchase decisions, understanding what makes content "machine likeable" becomes critical for digital success.

Machine Likeability Score Calculator

Score any product URL against all 26 dimensions and get optimization recommendations.

Model Genomes

Explore unique behavioral profiles for each AI model across all dimensions.

26 Dimensions

Deep dive into each psychological dimension and its effect on AI recommendations.

Cluster Breakdown

Six thematic clusters organizing 26 dimensions. Mean effect size across all clusters: h = 0.072

A: Evidence-Based
B: Value-Based
C: Risk & Assurance
D: Information Processing
E: Choice Architecture
F: Agentic Behaviors