Introduction

Cortex predicts how a paid-social ad will perform against your brand's own reference library — using a model that maps video and audio into predicted neural activity.

What you get

• A final score for each new candidate ad.
• A brand winner-likeness score based on your own winning and losing references.
• General attention, video, and element-level diagnostics.
• Six brain-network metrics tracked second-by-second through the ad.
• A divergence heatmap when that artifact is available.
• A narrative explaining what works and what doesn't.

The basic workflow

Every Cortex workspace follows the same structure:

1. Create a brand.
2. Add historical reference ads and label them as winners or losers.
3. Review the reference library and adjust weights when a reference is weak, ambiguous, or mislabeled.
4. Build a brand signature.
5. Upload a candidate ad and score it against the latest ready signature.
6. Read the result, including the final score, recommendations, playback timeline, network metrics, and confidence.

The signature is brand-specific. A candidate that scores well for one brand is not automatically a winner for another brand because each brand has its own reference bank, creative history, and response pattern.

What Cortex is and is not

Cortex uses model-predicted brain-response patterns. It does not recruit a new live viewer panel for every scoring run, and a score should not be read as a guarantee of media performance. The score is a structured prediction based on the relationship between your references and the candidate.

Use Cortex before spend is committed, during creative iteration, and when you need a consistent way to compare many candidate edits against the same brand standard.

First API shape

The app and API use the same product model. API automation starts by uploading a video, completing the upload, then creating a scoring run:

# 1. identify your workspace credential
export CORTEX_TOKEN="ctx_live_..."

# 2. create a scoring for an uploaded candidate video
curl -X POST https://app.cortex.ad/api/scorings/create \
  -H "Authorization: Bearer $CORTEX_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"video_id":"VIDEO_ID"}'

Most teams start in the browser so they can see reference health and calibration, then automate repeated candidate scoring through the API.

Concepts