Connect the ideas behind the AI Web

The AI-generated Web is more than model-written text.

Creation, adaptation and publication with generative models.

Retrieve documents to inform a generated answer.

Agents, tools, permissions and control of actions.

Connect AI applications to external tools and resources.

Enable independent agents to collaborate.

Expose structured capabilities from a Web application.

Describe the origin and some edits of a media asset.

Use the GPU to accelerate computation and interfaces.

Understand model families and their limitations.

Visibility in conventional search and answer interfaces.

Structure links between concepts, sources and technologies.

Interfaces usable by keyboard, readable and understandable.

Potential intersection of immersive Web and generative AI.

Tests, partial success, safety and reproducibility.

Keep authorized instructions separate from untrusted data.

Discover and evaluate content inclusion in AI answer interfaces.

Understand the distinct roles of indexing, search and training crawlers.

Avoid treating an optional convention as a universal ranking standard.

Make interfaces easier to interpret without degrading human experience.

Understand protocol evolution without conflating adoption with universal maturity.

Understand agent cards without oversharing internal information.

Study the experimental API and the exact status of the community draft.

Design approval gates before consequential actions.

Separate authorized instructions from untrusted retrieved Web content.

Structure risk assessment rather than relying on promises.

Assess translation, summarization and generation APIs offered by some browsers.

Understand a Web API designed for inference on user devices.

Separate GPU computation from model-level AI logic.

Identify errors at each stage of retrieval-augmented generation.

Understand relationships between semantics, metadata and AI systems.

Avoid conflating signatures, attribution and factual truth.

Read transparency requirements without generalizing beyond applicable law.

Apply WCAG to dynamic content and AI interactions.

Choose model architecture proportionate to tasks and devices.

Move from passive acceptance of answers to evidence-led inquiry.