The Science Behind the Scores

Gustoso uses Semantic Similarity Rating (SSR), a peer-reviewed methodology that achieves 90% of human panel reliability — without recruiting a single respondent.

Based on Maier et al., 2025 — developed at PyMC Labs

Why not just ask AI to rate things?

Traditional consumer panels take weeks and cost tens of thousands of dollars. But simply asking AI to “rate this product 1 to 5” doesn’t work either — language models default to safe, middle-of-the-road answers that tell you nothing useful.

SSR solves both problems. Developed by researchers at PyMC Labs, it extracts genuine preference signals from AI by measuring what they say, not what number they pick.

How SSR Works

Four steps from creative upload to actionable scores

Step 1Generate Personas
Step 2Elicit Reactions
Step 3Embed & Compare
Step 4Convert to Ratings

1. Generate Personas

We create AI consumers that match your target demographics — age, income, gender, region. Each persona responds independently, giving you a full panel of 300 unique viewpoints.

2. Elicit Free-Text Reactions

Instead of forcing a 1-to-5 rating, each persona writes an open-ended reaction to your creative. This captures nuance that numeric scales miss — enthusiasm, hesitation, specific objections.

3. Embed & Compare

Each response is converted to a mathematical representation and compared against five anchor statements ranging from “I would definitely not purchase this” to “I would definitely purchase this.” This is the SSR breakthrough — it measures semantic similarity, not arbitrary numbers.

4. Convert to Ratings

The similarity scores become a probability distribution across the five-point scale. The result isn’t a single average — it’s a full picture of how your audience splits between strong intent and strong rejection.

What You Get

Purchase Intent Distributions

See exactly how your audience splits across the 5-point scale, not just an average.

Demographic Breakdowns

Slice results by age, income, gender, and region to find which segments respond strongest.

Qualitative Themes

Surface the specific language your audience uses — objections, praise, and unexpected reactions.

Where It Works Best

SSR accuracy correlates with how much real consumer language exists for a category.

High confidence

CPG, personal care, food & beverage, consumer electronics, apparel

Abundant consumer language online

Medium confidence

B2B SaaS, luxury goods, healthcare

Less consumer discussion data to draw from

Low confidence

Industrial equipment, agricultural inputs

Limited online consumer discourse

What This Isn’t

Gustoso is not a replacement for large-scale quantitative research. It’s a fast, affordable way to screen concepts and creative before you commit budget. Think of it as a taste test — directional enough to kill bad ideas early and double down on good ones.

Ready to taste-test your next campaign?

Get purchase intent scores from 300 AI personas in minutes. Start with a free trial — no credit card required.