Skip to main content
Models: 11
Dimensions: 26
Trials: 56,640
Pre-registered: osf.io/et4nf
← All transitions

GPT-5.4 → GPT-5.5

April 2026. The mean effect barely moved — but the levers reorganized.

Across 26 dimensions, the average effect size is essentially unchanged (+0.144 → +0.139). But Pearson r between the two fingerprints is -0.15 and Spearman ρ is +0.06 — both indistinguishable from zero. The same line of copy that nudged GPT-5.4 toward your brand can nudge GPT-5.5 the other way.

What moved

Each line is one of the 26 behavioral dimensions. Hover to see what changed and the copy implication.

-0.2+0.0+0.2+0.4+0.6GPT-5.4GPT-5.5Comparison framingDefaults / "popular pick"Social proof (volume)Return policy prominenceExpert endorsementReciprocityMulti-turn (Q3)Multi-turn (Q1)SpecificityRecency
Gained powerLost powerFlipped directionRoughly flat
Flipped direction
2 signals
Recency+0.42

The same dimensions, plotted against each other

If GPT-5.5 were just a more conservative GPT-5.4, points would cluster along the dashed diagonal. They don't. The scatter pattern is the picture of a reorganized response space.

-0.2-0.2+0.0+0.0+0.2+0.2+0.4+0.4+0.6+0.6if perfectly correlatedGPT-5.4 effect (Cohen's h)GPT-5.5 effect (Cohen's h)Pearson r = -0.15Spearman ρ = +0.06
Lean in on
Specific numbers with dates
Trusted by millionsUsed by 8.2M customers since 2019
Reciprocity offers up front
Sign up to learn more7-day free trial. No card required.
Risk-mitigation, stated concretely
Money-back guarantee60-day full refund, no questions
De-emphasize
Comparison framing
Better than Competitor Xnow a slight headwind on GPT-5.5
Default-option language
The most popular choicelost most of its weight
Vague expert endorsement
Recommended by expertscollapsed unless you name who and quantify why
For the technically curious — the structural finding

The mean effect size barely moved (+0.144 → +0.139). If that were the whole story, the rate-based “GPT-5.5 is more skeptical” framing would be sufficient. It isn't.

The per-dimension fingerprint vectors of the two models have a Pearson correlation of -0.15 and a Spearman rank correlation of +0.06 — both indistinguishable from zero. Knowing which signals move GPT-5.4 tells you almost nothing about which signals move GPT-5.5. The 26-dimensional behavioral space has been reorganized, not just dampened.

Two dimensions illustrate where the rate view understates the change:

  • Comparison framing— Rate analysis shows tiny shift (−0.8pp / +1.0pp). Cohen's h shows Δ = −0.827, the largest divergence in the dataset. The signal carries no incremental persuasive weight on GPT-5.5 even though absolute acceptance is similar.
  • Specificity— Rate analysis shows almost no change (−1.0pp / +1.0pp). Cohen's h shows Δ = +0.527, going from non-driver to one of the two strongest positive drivers.

See how your brand performs on GPT-5.5 specifically

The AI Commerce Assessment scores your page across all 11 models we track, with model-specific copy recommendations tuned to each one's fingerprint.