Transparency, not promise.
VibeScore
Which emotions does this text carry?
Several AI models read a piece of text across eight emotions and return a profile and a 0 to 1000 score, with where they agree and split. It reads the content, not the people who see it.
Vibe-Score
Measure the Moments that Matter.
Receipt saved · score ready
Eight emotions. Four models. One number.
We measure the message as written. Not the messenger, not the event, not human worth.
Three steps, one reading.
- 01 · In
Content
A headline, a message, a script, a product description. VibeScore reads the words, not the person who wrote them or the people who receive them.
- 02 · Read
Several models, eight emotions
Several AI models read the same text. Each gives a strength for each of the eight emotions, and each reading is kept separate.
- 03 · Out
A profile and a record
All eight emotions with their strength, a 0 to 1000 score, where the models agree and split, and a record of what produced the reading, so it can be compared with the next version.
The eight emotions.
Every reading covers the same eight, in the same order, so two readings can be set side by side.
- Joy
- Trust
- Fear
- Surprise
- Sadness
- Disgust
- Anger
- Anticipation
Agreement is consistency, not truth.
When the models agree, the reading is consistent. That does not make it correct. When they split, the split is shown, not averaged away.
How closely the models agree with each other, and with labels written by people, is measured on Research, with the low numbers included.
What it reads, what it does not.
What it reads
- Text.The words of a piece of content, as written.
- Emotional associations.Which of the eight emotions the words carry, and how strongly.
- Where readers differ.Where the models agree on a reading and where they split.
What it does not
- People.A reading describes the text. It never diagnoses the writer or says how a reader feels.
- Forecasts.No forecast of audience reaction or sales.
- Images, web pages, video.It reads text. Other formats have no validation yet.
Guardrails and limits.
Guardrails
- Words, not people.Read-outs describe the text and take no side.
- No accuracy talk in readings.Read-outs do not state percentages or objective claims. Test results live on Research, with their limits.
- Colour means something.A reading's colours are computed from its own emotions, never added as decoration.
Limits
- Preliminary.The October 2026 tests are an early pass on a draft method, and will be run again once it is locked.
- One set of models.The tested figures describe the models used in October. A reading made with a different set is not covered by them.
- Trust reads lowest.The models agree least on trust. Treat a trust finding as lower confidence.
Use it.
The live instrument runs at vibe-score.ai. This site does not run readings.
Request a free introduction
Twenty minutes to talk through what you're working on and whether Ad+Verb can help. If there's a fit, we'll agree on a written scope before work begins.