AI Taxonomy Methodology

Current methodology version: 3.4.

The AI Coin Map taxonomy describes a project's role in the AI crypto industry, not its size or the future performance of its token. It supports comparable segments, market views, and AI Score. Classification is conservative, based on public evidence, and is not a technical audit, endorsement, or investment recommendation.

Four taxonomy dimensions

AI Relevance — the role of AI in the project

  • Core AI Protocol
  • AI-Native Layer-1 / Blockchain
  • AI Infrastructure
  • Applied AI / Agentic Applications
  • AI-Enabled / Peripheral
  • Non-AI / Remove

AI Relevance is mandatory. Non-AI / Remove is an internal exclusion state; these projects should not appear as a standalone category in the public AI catalogue.

AI Industry — application domain

  • DeFi / Finance
  • Gaming & Entertainment
  • Enterprise & Business Tools
  • Healthcare & Biotech
  • Media / Content Generation
  • DeSci / Research
  • Cybersecurity & Defense
  • Social / Creator Platforms
  • Robotics & Physical AI

AI Stack — position in the technology stack

  • Decentralized Compute / GPU DePIN
  • Inference Networks / Model Serving
  • Model Training & Fine-tuning Networks
  • Data Marketplaces & Labeling
  • AI Agents & Autonomous Platforms
  • Developer Tools / AI Middleware
  • Indexing, Search & Retrieval (RAG)
  • Privacy, ZKML & Secure AI
  • Consumer & Application Layer AI

AI Technique — verified AI approaches

  • Large Language Models (LLMs)
  • Multimodal AI
  • On-chain AI / Verifiable AI
  • RAG / Retrieval-Augmented Generation
  • Agentic AI / Autonomous Agents
  • Decentralized Training & Federated Learning
  • Generative AI
  • ZKML & Cryptographic AI

New terms are not invented for individual projects. Two to four verified tags are usually sufficient; more tags do not automatically indicate a stronger AI component.

Sources and review process

Official documentation, whitepapers, technical materials, and GitHub repositories receive the highest weight. Official websites, blogs, Medium posts, and other publications are secondary sources. An aggregator description, marketing post, partnership, or roadmap promise is not sufficient technical evidence.

  1. Fact extraction: verifiable claims are extracted from the available sources.
  2. Classification proposal: facts are mapped only to terms in the controlled vocabularies.
  3. Critical review: a separate stage challenges inflated AI relevance and weak evidence.
  4. Final decision: disputed signals are resolved in favour of the more conservative category.
  5. Human review: required when confidence is low, evidence conflicts, or sources are incomplete.

Key classification safeguards

  • AI-Native Layer-1 / Blockchain requires AI-oriented architecture for execution, training, inference, data, or verifiable AI. Partnerships and roadmap claims are not enough.
  • Applied AI / Agentic Applications requires meaningful first-party AI logic. A simple interface to a third-party model is usually AI-Enabled / Peripheral.
  • Agentic AI requires autonomous planning, action execution, workflows, or an agent framework, not merely the word “agent”.
  • Robotics & Physical AI is used only for direct work in robotics, embodied AI, vision for physical systems, drones, or physical automation.
  • GPU DePIN with documented AI workloads may qualify as AI Infrastructure even when the project does not develop its own model.
  • When technical evidence is absent, the project receives Non-AI / Remove and its other AI tags are cleared.

Data, descriptions, and the public catalogue

Descriptions contain durable facts about purpose and architecture. Price, market capitalisation, supply, exact capacity figures, partners, roadmaps, and GitHub counts are not embedded in permanent copy; they change and are presented separately through updated structured data.

Only published, valid AI projects contribute to public lists and aggregates. Segment pages with fewer than five published projects are marked noindex and excluded from the sitemap to prevent thin pages from entering search indexes.

Limitations

Public sources may be incomplete, outdated, or contradictory. A classification reflects the evidence available at the latest review and may change. It does not establish project safety, team quality, tokenomics, or future returns. See the Disclaimer for additional terms.