Brand Function Registry

Machine-readable brand specifications for the AI era.

A Brand Function is a structured JSON file that describes how a brand behaves across eight perceptual dimensions: Semiotic, Narrative, Ideological, Experiential, Social, Economic, Cultural, and Temporal. It lives at a known URL on a brand's website (.well-known/brand.json) so AI systems can discover it and build accurate perceptions rather than hallucinating from scattered training data.

Why This Exists

AI shopping agents -- ChatGPT, Gemini, Claude, DeepSeek, and others -- are becoming the default tool consumers use to compare brands. These models systematically collapse brand perception:

Dimension	AI Weight	Baseline	Status
Experiential	18.8	12.5	Over-weighted (150%)
Semiotic	14.8	12.5	Over-weighted (118%)
Economic	14.3	12.5	Over-weighted (114%)
Social	7.8	12.5	Collapsed (62%)
Narrative	10.5	12.5	Collapsed (84%)
Temporal	8.1	12.5	Collapsed (65%)
Cultural	7.3	12.5	Collapsed (58%)
Ideological	8.2	12.5	Collapsed (66%)

This is not a quirk of one model. Cross-model cosine similarity is .977 across 24 architectures from 7 training traditions (21,350 API calls, 10 experimental runs). The collapse is structural.

The Brand Function provides the structured information AI systems need to perceive brands accurately. It is the schema.org of brand identity.

Research:

• Zharnikov (2026v), "Spectral Metamerism in AI-Mediated Brand Perception." 21,350 calls, 24 LLMs, 7 traditions. Zenodo | Dataset
• Zharnikov (2026x), "AI-Native Brand Identity." Behavioral specification framework. Zenodo
• Zharnikov (2026ac), "Spectral Immunity: Why Brand Portfolio Interference Disappears for AI Observers." 9,925 obs, 40 brands, 13 models, 7 traditions. Supersedes R8 (2026q) + R20 (2026ab). Zenodo | Dataset
• Zharnikov (2026ad), "Restoring Perceptual Separability After Coherence Shocks: A μ > λ Threshold Inequality in Brand Perception." Monte Carlo: gap 1.10 vs .02 (52x); IRF half-life 1.4 vs 13.1 months. Zenodo

Registry Contents

This registry contains Brand Functions for 26 brands across categories:

Category	Brands
Luxury	Hermes, Louis Vuitton, Rolex
Sportswear / Outdoor	Nike, Patagonia, North Face, Lululemon
Automotive	Tesla, Toyota, Mercedes-Benz, Volvo, Rivian
Tech / Media	Apple, Samsung, Huawei, Netflix
Retail / FMCG	IKEA, Zara, Uniqlo, Starbucks, Coca-Cola, Dove, Trader Joe's, Whole Foods
Specialty	Erewhon, Emirates

Empirical Evidence

Brand Functions were tested in a controlled experiment (Run 12b): 26 brands from this registry were rated by 4 cloud LLMs (624 API calls) with and without their Brand Function specification.

Result: Brand Functions produce dimensional redistribution, not uniform collapse reduction. The aggregate DCI change is not statistically significant (p = .427), but individual dimensions shift meaningfully:

Dimension	Mean Shift	Direction
Economic	+2.5	Toward baseline (good)
Cultural	+1.6	Toward baseline (good)
Narrative	+1.3	Toward baseline (good)
Experiential	-1.2	Further from baseline
Ideological	-0.9	Further from baseline

71% of brands (15/21 with sufficient data) showed DCI improvement. The effect is a structured intervention with measurable but uneven impact -- not a guaranteed fix. Brands should monitor per-dimension shifts after deployment.

File Structure

brands/
  nike/
    brand.json      # The Brand Function specification
    README.md       # Human-readable summary
  coca-cola/
    brand.json
    README.md
  ...
schema/
  brand-function-v1.schema.json   # JSON Schema for validation

Brand Function Format

Each brand.json contains:

{
  "brand": "Brand Name",
  "version": "1.0",
  "dimensions": {
    "semiotic": {
      "score": 9.0,
      "positioning": "What the brand's visual identity communicates.",
      "key_signals": ["observable evidence 1", "observable evidence 2"]
    },
    "narrative": { ... },
    "ideological": { ... },
    "experiential": { ... },
    "social": { ... },
    "economic": { ... },
    "cultural": { ... },
    "temporal": { ... }
  }
}

Required fields: brand, version, dimensions (all 8), and positioning per dimension.

Optional fields: score (0--10 intensity scale), key_signals (evidence array), source, updated.

See the JSON Schema for the full specification.

How to Use

Deploy for your brand

1. Write your Brand Function using the format above (or copy a similar brand's file as a starting point)
2. Save it as brand.json
3. Host it at https://yourdomain.com/.well-known/brand.json
4. AI crawlers that index your .well-known directory will discover it

Run an AI Brand Audit

Measure how AI currently perceives your brand before and after deploying a Brand Function:

git clone https://github.com/spectralbranding/sbt-papers.git
cd sbt-papers/r15-ai-search-metamerism/experiment
# Edit brand pairs, set API key, run
python ai_search_metamerism.py --live --runs 3

Full audit methodology: The $0.80 AI Brand Audit.

Validate a Brand Function

pip install jsonschema
python -c "
import json, jsonschema
schema = json.load(open('schema/brand-function-v1.schema.json'))
bf = json.load(open('brands/nike/brand.json'))
jsonschema.validate(bf, schema)
print('Valid.')
"

Contributing

Brand Functions are community-maintained. To add or update a brand:

1. Fork this repository
2. Create brands/brand-name/brand.json following the schema
3. Add brands/brand-name/README.md with a brief human-readable summary
4. Open a pull request

Guidelines:

• One brand per directory
• Directory names: lowercase-hyphen
• All eight dimensions required; use your best judgment for scores
• positioning text should describe behavior, not appearance
• key_signals should be verifiable, observable evidence
• If you work for the brand: indicate this in your PR description

Claim Your Brand

If you represent a brand in this registry and the specification is inaccurate, open an issue or PR. We will prioritize corrections from brand owners.

If your brand is not in the registry, add it yourself or request it.

Related

• Spectral Brand Theory -- the research program behind the 8-dimension framework
• SBT Framework -- open-source measurement toolkit
• AI Brand Perception series -- practitioner articles on Substack
• R15: AI Search Metamerism -- dimensional collapse study (21,350 calls, 24 models) | Dataset
• R16: AI-Native Brand Identity -- behavioral specification framework
• R21: Spectral Immunity -- spectral immunity study (9,925 obs, 40 brands, 13 models, supersedes R8 + R20) | Dataset
• R22: Spectral Gap Restoration -- μ > λ threshold inequality for cohort separability after coherence shocks; Monte Carlo demonstration with Dove design parameters
• .well-known/brand.json specification -- the deployment standard

License

Brand Function data: CC BY 4.0. JSON Schema: MIT.