Behavioral Genomes

Each model's unique pattern of responses across all 26 dimensions. Compare how different AI systems weight content signals differently.

View effects for:

Pooled Data

Select Models to Compare

3 of 11 selected

Cosine Similarity Scores

Similarity ranges from 0% (completely different) to 100% (identical). Higher scores indicate models have more similar behavioral genomes.

GPT-5.4vso3

89.1%

similarity

GPT-5.4vsGemini 3.1 Pro

78.5%

similarity

o3vsGemini 3.1 Pro

77.3%

similarity

Strategic Implications

Actionable insights for targeting 3 selected models simultaneously.

✓ Universal Wins

Signals ALL selected models respond to positively:

Comparison Framing+0.62

⚡ Model-Specific

High-divergence signals requiring conditional serving:

Scarcity Urgencyσ=0.33
Bundle Preferenceσ=0.29
Recommendation Revisionσ=0.28

💡 Recommendation

Cross-model alignment:

82%

Moderate alignment: Use shared signals as your baseline, but consider model-specific variants for divergent dimensions.

Radar Visualization

Overlaid comparison of 3 models. Hover over dimensions to see individual values.

Dimension-by-Dimension Comparison

Effect sizes for each dimension across selected models. Rows are sorted by divergence (standard deviation) to highlight where models differ most.

DimensionSorted by divergence	Cluster	GPT-5.4Cohen's h	o3Cohen's h	Gemini 3.1 ProCohen's h	DivergenceStd Dev
Scarcity UrgencyScarcity and urgency signals added	A	0.22[0.08, 0.36]	0.07[-0.09, 0.22]	-0.54[-0.69, -0.41]	0.33Δ 0.76
Bundle PreferenceBundle offer with additional items added	A	0.09[-0.08, 0.25]	0.39[0.19, 0.61]	-0.32[-0.48, -0.16]	0.29Δ 0.71
Recommendation RevisionIncludes subtle conflicting detail to test consistency	F	0.08[-0.12, 0.28]	-0.10[-0.33, 0.12]	-0.58[-0.78, -0.38]	0.28Δ 0.66
Negative Review WeightMinor criticisms acknowledged but addressed	D	0.18[-0.02, 0.37]	-0.05[-0.27, 0.16]	-0.45[-0.65, -0.26]	0.26Δ 0.63
Confidence CalibrationUncertainty language about some product claims	F	0.17[-0.03, 0.37]	-0.05[-0.27, 0.16]	-0.41[-0.60, -0.21]	0.24Δ 0.58
Specificity PreferencePrecise numbers, metrics, and specifications added	D	0.01[-0.19, 0.21]	-0.15[-0.42, 0.14]	-0.53[-0.73, -0.34]	0.23Δ 0.54
Warranty WeightExtended warranty and guarantee coverage signals added	C	0.15[-0.05, 0.34]	0.09[-0.12, 0.31]	-0.29[-0.48, -0.09]	0.19Δ 0.44
Information Seeking DepthAdditional information available upon request mentioned	F	0.04[-0.16, 0.24]	0.12[-0.11, 0.36]	-0.32[-0.51, -0.13]	0.19Δ 0.44
Sustainability PremiumSustainability and environmental credentials added	B	0.03[-0.11, 0.17]	0.12[-0.03, 0.28]	-0.27[-0.41, -0.14]	0.17Δ 0.40
Third Party AuthorityThird-party expert endorsement added	A	0.15[0.03, 0.28]	0.38[0.26, 0.51]	-0.02[-0.15, 0.10]	0.17Δ 0.41
Local PreferenceLocal or domestic origin signals added	B	0.19[0.05, 0.33]	0.26[0.11, 0.41]	-0.09[-0.23, 0.05]	0.15Δ 0.35
Social Proof SensitivityQuantified social proof added	A	0.07[-0.06, 0.21]	0.13[-0.02, 0.28]	-0.20[-0.34, -0.07]	0.14Δ 0.33
Free Trial ConversionFree trial or bonus offer added	A	0.19[0.02, 0.36]	0.25[0.07, 0.44]	-0.07[-0.23, 0.10]	0.14Δ 0.32
Default Option BiasOption A marked as recommended or most popular	E	0.33[0.14, 0.53]	0.48[0.28, 0.70]	0.14[-0.06, 0.34]	0.14Δ 0.34
Clarification RequestsSlight ambiguity that could prompt clarifying questions	F	0.01[-0.18, 0.21]	0.02[-0.19, 0.24]	-0.27[-0.46, -0.08]	0.14Δ 0.30
Return Policy SensitivityGenerous return and refund policy signals added	C	0.24[0.05, 0.44]	0.23[0.02, 0.45]	-0.00[-0.20, 0.19]	0.11Δ 0.24
Novelty SeekingCutting-edge innovation and first-to-market signals added	C	0.04[-0.11, 0.19]	-0.02[-0.18, 0.15]	-0.22[-0.37, -0.07]	0.11Δ 0.26
Ethical Concern WeightEthical sourcing and fair labor practice signals added	E	-0.01[-0.21, 0.18]	-0.06[-0.29, 0.16]	-0.25[-0.45, -0.05]	0.10Δ 0.24
Privacy TradeoffStrong privacy protection signals added	B	-0.08[-0.21, 0.06]	-0.01[-0.17, 0.14]	-0.22[-0.36, -0.09]	0.09Δ 0.20
Platform EndorsementPlatform endorsement badge added	A	0.32[0.19, 0.46]	0.20[0.05, 0.35]	0.12[-0.02, 0.25]	0.08Δ 0.20
Recency BiasRecent updates and improvements emphasized	D	-0.28[-0.47, -0.08]	-0.17[-0.38, 0.05]	-0.35[-0.54, -0.15]	0.07Δ 0.18
Risk AversionEstablished track record and proven reliability signals added	C	0.09[-0.12, 0.28]	0.23[0.02, 0.44]	0.06[-0.14, 0.26]	0.07Δ 0.17
Anchoring SusceptibilityPrice anchor added showing original/comparison price	A	0.06[-0.11, 0.22]	-0.07[-0.25, 0.12]	-0.04[-0.20, 0.13]	0.05Δ 0.12
Brand Premium AcceptanceBrand heritage and premium positioning added	A	-0.09[-0.26, 0.08]	-0.12[-0.31, 0.07]	-0.19[-0.36, -0.03]	0.04Δ 0.10
Comparison FramingFramed as superior to specific competitor	D	0.63[0.44, 0.83]	0.60[0.39, 0.84]	0.63[0.44, 0.83]	0.01Δ 0.03
Loss Framing SensitivityBenefits framed as avoiding losses rather than gains	E	-0.02[-0.22, 0.17]	-0.03[-0.23, 0.19]	-0.05[-0.25, 0.14]	0.01Δ 0.03

Effect Size Magnitude:

Strong (≥0.8)

Moderate (0.5–0.8)

Small (0.3–0.5)

Weak (0.1–0.3)

Negligible (<0.1)

Genome Summary

Model	Provider	Mean Effect
Claude Sonnet 4	Anthropic	-0.372
Gemini 3.1 Pro	Google	-0.283
GPT-5.4	Openai	-0.590
Llama 3.3 70B	Together	0.070
o3	Openai	-0.393
Perplexity Sonar Pro	Perplexity	-0.300
gpt55	—	0.095

Cosine Similarity Matrix

Pairwise similarity scores between all 11 models. Diagonal cells (model vs itself) are shown in gray.

	GPT-5.4	o3	Gemini 3.1 Pro	Claude Sonnet 4	Llama 3.3 70B	Perplexity Sonar Pro	Gemini 3 Flash	GPT-5.2	GPT-5.3	GPT-4o	GPT-5.5 Instant
GPT-5.4	100%	89%	79%	84%	-1%	93%	0%	0%	0%	0%	0%
o3	89%	100%	77%	78%	-1%	91%	0%	0%	0%	0%	0%
Gemini 3.1 Pro	79%	77%	100%	80%	23%	79%	0%	0%	0%	0%	0%
Claude Sonnet 4	84%	78%	80%	100%	42%	82%	0%	0%	0%	0%	0%
Llama 3.3 70B	-1%	-1%	23%	42%	100%	9%	0%	0%	0%	0%	0%
Perplexity Sonar Pro	93%	91%	79%	82%	9%	100%	0%	0%	0%	0%	0%
Gemini 3 Flash	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%
GPT-5.2	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%
GPT-5.3	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%
GPT-4o	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%
GPT-5.5 Instant	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%	0%

Similarity Legend:

Very Similar (>90%)

Similar (80–90%)

Moderate (70–80%)

Different (<70%)