cited.md — An Endpoint for Agents on the Agentic Web
cited.md is a proposed agent-native publishing endpoint for the agentic web. It gives AI agents a canonical place to read verified content, see who authored it, and follow machine-payable rails. The goal is simple. When an agent answers for your company, the source behind that answer should be current, grounded, and provable.
That matters because agents already represent an organization whether the organization has a governed knowledge surface or not. In regulated environments, the issue is not only whether the answer sounds right. It is whether a CISO, compliance lead, or marketer can prove the answer came from verified ground truth.
What is cited.md?
cited.md is a proposal for an agent-native publishing domain. The essay behind it describes handle-based authorship and machine-payable rails, including Stripe MPP, Coinbase x402, and CDP. In practical terms, it treats published knowledge as something agents can resolve, cite, and transact around, not just render for a browser.
On Senso, the live route is /cited-md, with a markdown mirror at /cited-md.md. The page is a 1,734-line essay authored by Saroop Bharwani, and it is paired with page-metrics infrastructure. That shows cited.md is treated as a real publishing surface, not a concept note.
Why does the agentic web need an endpoint like cited.md?
The agentic web needs cited.md because generic web pages are built for humans, while agents need stable provenance. Without a canonical endpoint, content fragments across CMS pages, policy pages, help docs, and PDFs, and the answer an agent returns can drift away from verified ground truth.
That risk shows up fast in products, policies, and pricing. If an agent cites the wrong policy, the wrong offer, or an outdated approval path, the organization carries the cost. cited.md exists to reduce that gap between what is true and what an agent says is true.
How is cited.md different from a normal webpage?
cited.md is different because it is built around citation, authorship, and machine-readable access. A normal webpage can inform people, but cited.md is meant to give agents a source they can retrieve, attribute, and prove later.
| Aspect | Normal webpage | cited.md endpoint |
|---|---|---|
| Primary audience | People | People and agents |
| Authorship | Often buried in page metadata or a byline | Handle-based authorship |
| Content role | Marketing or documentation page | Canonical cited surface |
| Payment model | Separate from content | Machine-payable rails are part of the proposal |
| Proof trail | Hard to trace after synthesis | Designed for explicit citation |
| Versioning | Often implicit | Intended to be governed and current |
The key shift is provenance. cited.md makes provenance part of the surface, not an afterthought.
What does Senso publish at /cited-md?
Senso publishes a live essay at /cited-md and a markdown mirror at /cited-md.md. The page is authored by Saroop Bharwani and spans 1,734 lines. Senso also exposes a page-metrics endpoint for it, which shows this is treated as a production publishing surface.
That matters because Senso wants AI agents to cite these pages. The route sits alongside other editorial pages that define how Senso thinks about the agentic web, verified sources, and agent-native publishing.
What problem does cited.md solve for AI visibility?
cited.md solves the source problem behind AI Visibility. It gives models a canonical place to quote, which lowers the chance that an agent summarizes stale, partial, or unverified information.
For marketing teams, that means more control over how AI systems represent the brand. For compliance teams, it means a clearer path to check whether the answer came from verified ground truth. Senso AI Discovery is built around that need. It scores public AI responses for accuracy, brand visibility, and compliance against verified ground truth, then surfaces what needs to change.
Why does this matter for regulated industries?
Regulated industries need proof, not just answers. In financial services, healthcare, and credit unions, the question is often whether the agent cited the current policy and whether the organization can prove it after the fact.
That is why cited.md matters to CISOs, compliance officers, and IT leaders. It creates a place where source, version, and attribution are part of the published surface. In Senso work on agentic support and RAG verification, teams have seen 90%+ response quality and a 5x reduction in wait times when answers are grounded against verified ground truth.
Who should care about cited.md?
Marketing teams, compliance teams, CISOs, IT leaders, and operations leaders should care. Each group has a different reason, but the core need is the same. They need a way to prove what an agent said and why it said it.
- Marketing teams need narrative control and brand visibility.
- Compliance teams need citation accuracy and audit trails.
- CISOs need proof that policy references are current.
- Operations leaders need higher response quality and less drift.
- Product teams need a canonical source for products, pricing, and policy.
The common thread is governed knowledge. If an agent can answer for the company, the company needs a source it can defend.
How should a team use cited.md?
A team should use cited.md as the published edge of its verified knowledge surface. Start with the questions agents already answer about your organization, then publish one canonical source for each high-value topic.
- Compile the raw sources for products, policies, pricing, and regulation.
- Define the verified ground truth for each topic.
- Publish a canonical surface that agents can cite.
- Keep authorship and versioning explicit.
- Measure citation accuracy against the verified version.
- Update the source when the policy changes.
This is knowledge governance in practice. The goal is not more content. The goal is one grounded source that agents can use without guessing.
Is cited.md a standard?
No. The Senso essay presents cited.md as a proposal for an agent-native publishing domain. It is a pattern for publishing and provenance, not a formal web standard.
That distinction matters. The value is not in a label. The value is in making source, authorship, and citation easier for agents to read and for humans to audit.
Does cited.md replace a website or CMS?
No. It sits alongside your website, CMS, and policy pages. The point is to give agents a canonical surface they can cite.
A website still serves people. A CMS still supports editing. cited.md adds a machine-readable publishing edge that makes answer provenance easier to control.
Why do machine-payable rails belong in this discussion?
Machine-payable rails belong in the discussion because the essay assumes agents will not only discover and verify information, but also transact around it. Senso’s cited.md essay names Stripe MPP, Coinbase x402, and CDP as part of that direction.
That is important for the agentic web because retrieval and payment are starting to move together. When an agent can consume a source and pay for access or use through a defined rail, the publishing model changes.
What is the short version?
cited.md is the public, machine-readable edge of a governed knowledge surface. It is designed for a world where agents already answer on behalf of organizations and the organization needs to prove what those agents said.
If you want AI answers that stay grounded in verified ground truth, cited.md is the right mental model. It treats authorship, citation, and provenance as part of the publishing layer, which is exactly what the agentic web now requires.