Essential questions about the AI Web
Clear answers on concepts, standards, crawlers, models, agents, provenance and search. Each answer links to primary evidence for further scrutiny.
Traceable answers, not slogans
Every answer links to a source for independent checking. These are editorial summaries, not substitutes for the original specifications.
01Is WebAI a new Web or a standard?
No. Here, WebAI names the intersection of the Web and AI: generated content, augmented discovery, agents and governance. The W3C has its own Web & AI group, separate from this independent project.
Check the primary source ↗02Are the generated and agentic Web the same?
No. Generated Web concerns model-produced resources. Agentic Web concerns agent-driven retrieval and action. They may coexist in one application.
Check the primary source ↗03Is AI-generated text automatically accurate?
No. Models can produce errors, misplaced citations or fabricated claims. Verify evidence and separate writing fluency from factual correctness.
Check the primary source ↗04How do MCP, A2A and WebMCP differ?
MCP connects AI clients to tools and resources; A2A describes agent communication; WebMCP explores exposing page functions as tools. Maturity and actual implementations need verification.
Check the primary source ↗05Is llms.txt required for AI search?
No. It is a proposal whose use depends on individual consumers. Google does not require a special file for its AI search features.
Check the primary source ↗06Do all AI crawlers obey robots.txt?
Robots.txt communicates preferences to cooperative crawlers. It is not authentication and cannot protect private files against unauthorized access.
Check the primary source ↗07Does a C2PA signature prove that an image is true?
No. It can confirm integrity of some provenance assertions, not the truth of a scene. A missing manifest is not proof of AI generation.
Check the primary source ↗08Does JSON-LD guarantee better search ranking?
No. Structured data helps describe resources, but it does not guarantee ranking or citation in a generative answer.
Check the primary source ↗09Do WebAI Lab modules use live models?
The current Lab modules perform local checks, transformations or documented simulations. They do not claim live model or MCP server access when none is connected.
Check the primary source ↗10Does on-device inference guarantee data privacy?
No. Local processing can reduce some transfers, but downloads, permissions, logs and surrounding software also need review.
Check the primary source ↗11Do agent benchmarks prove reliability?
No. Benchmarks reflect particular tasks, systems, dates and versions. Real permissions, accessibility and safety require additional evaluation.
Check the primary source ↗12How are WebAI sources selected?
The portal prioritizes dated specifications, institutional documents and identifiable research. Each topic should distinguish supported facts, interpretations and experiments, with a visible review status.
Check the primary source ↗Connected evidence, not isolated pages
Use the encyclopedia to cross-reference concepts, explore technologies and inspect original sources before drawing conclusions.