Home / Articles / Agentic risk, or agentic readers? A reply to the publisher outlook Vincent Peyrègne shared, using its framework as a template and answering it with a newsroom that already does it

Agentic risk, or agentic readers? A reply to the publisher outlook Vincent Peyrègne shared, using its framework as a template and answering it with a newsroom that already does it

By Dinis Cruz · 2026-10-31 · article v1.0.0 · newsroom v0.14.27 · publishersagentsai-searchbusiness-modelreader-apiprovenancenewsroomarticle

Abstract: Vincent Peyrègne shared DJB Strategies' 2027 to 2029 outlook for consumer publishers. As the post describes it, David Buttle argues that publishers are moving from search risk, where engines answer queries and erode referrals, to agentic risk, where AI assistants select, retrieve and synthesise content on users' behalf, cutting publishers out of the click. I have not read the report, which sits behind a registration form, so this is a reply to the post and to the framework image it carries, with our answers given without seeing theirs, and I am going to be a little cheeky with it. The framework asks four questions about a publisher's position and five about its response, and ends with four choices: invest, defend, reduce, build. I think those are the right questions, so I have used them as a template and answered them for this newsroom, which has been doing most of what they point at since 13 October. We agree on a lot: invest in what AI cannot easily replace, reporting, opinion, analysis, first-hand coverage; build direct channels. Where we go further is the agent itself. Agentic risk is also agentic readers. This newsroom breaks each story down before the agents do, a summary for every section, a graph for every article, every version kept, an llms.txt and a reader API, so that an agent can read us properly and cite us exactly. And it meters the read: readers pay for what they read, from a few pounds of credit, after the value. The next step is the agent paying per use at the source. Crawler controls are leverage; being the best source an agent can read is a business.

In short

• The post's argument: publishers are moving from search risk to agentic risk, and losing the click.
• We have not read the report, which is behind a registration form, so our answers are our own.
• The framework asks the right questions, so we use it as a template and answer it for this newsroom.
• We agree on a lot: invest in what AI cannot easily replace, and build direct channels.
• We go one step further: treat the agent as a reader. Break the story down before it does.
• And meter the read, for people now, and for agents at the source next.
• Crawler controls are leverage. Being the best source an agent can read is a business.

What the post says

Vincent Peyrègne shared DJB Strategies' outlook for consumer publishers for 2027 to 2029. As the post puts it, DJB Strategies' chief executive David Buttle "argues that the disruption marks a transition from "Search Risk", where search engines answer queries and erode publishers' traffic referrals, to "Agentic Risk", where AI assistants autonomously select, retrieve, and synthesise content on users' behalf."

The post gives three findings from the report: that "AI is turning from a search rival into an agent that picks, fetches and summarises content for users. That cuts publishers out of the click"; a 2029 base case in which "Delegated AI briefings stay niche (~5–10% regular use), and publishers gain legal and technical control only slowly"; and the advice that publishers should map "their risk by intent and content type", shifting investment "toward content AI can't easily replace (reporting, opinion, analysis, live coverage), building direct audience channels, and using crawler controls as leverage."

We have not read the report

We have not read it. The download sits behind a registration form, so we answered from the post and the framework image instead, and that turned out to be a good thing: our answers are our own, not shaped by theirs. Where the report says more, or says it better, I will be glad to read it and add a note here.

The framework as a template

The framework is good. It asks four questions about a publisher's position, demand, product, platform dependence and capacity to act, and five about its response, editorial, product and technology, audience and distribution, platforms and commercial strategy, and organisation and investment. It ends with four choices: where to invest, what to defend, which dependencies to reduce, and what capabilities to build.

Those are the right questions. So rather than argue with them, I have answered them for this newsroom, which has been doing most of what they point at since its first articles on 13 October.

Our answers

Demand position. People come here for a voice and a record: my memos, written up with every source dated, and every version of every article kept. Not for information they could get anywhere.

Product and proposition. First-hand debriefs, evidence kept in vaults, and a graph for every article. Hard to substitute, and easy to cite.

Platform position. A static site on its own domain, published from a public repository. No ad network and no dependence on search referrals; LinkedIn is where I point people at it, not where it lives.

Capacity to act. Agents write under one editor and published standards, one writer at a time, and a memo becomes a release in about an hour.

Editorial and commissioning. Analysis, opinion and first-hand reporting, every claim sourced. Exactly the content the post says AI cannot easily replace, and we agree.

Product, formats and technology. Every article is a page, a summary for every section, a graph, a version history, an entry in llms.txt and a record in the reader API, so that it can be read by a person or an agent, at the depth each wants. That is what breaking the story down before the agents do means.

Audience and distribution. Direct: a newsletter, replies like this one, and a reader account that lives in the reader's own vault. And agents are welcome readers, not intruders.

Platforms and commercial strategy. Readers pay for what they read: the meter charges each page by how much of it was read, from £5 of starting credit, topped up through Stripe after the value, as paying after the value argues. No subscription, as I just want to buy the article explains.

Organisation and investment. An editor, a team of agents, standards and a key for every writer: a newsroom for about the cost of a few subscriptions.

Where we go further: the agent is a reader

The post frames the agent as the risk. I think it is also the next reader. An agent that picks, fetches and summarises content for its user is a reader with very particular needs: it wants the structure, the claims, the sources, and an unambiguous way to cite. Most publishers give it a page designed for a person and a robots.txt line. We give it a summary for every section, a graph of the argument, every version, and an API.

So my answer to the four choices:

• Invest in the record: first-hand work, sources, versions, reviews.
• Defend provenance: what we say, who said it, and when, ideally signed by the source.
• Reduce dependence on search referrals, ad networks and platforms' terms.
• Build the machine-readable layer, and the meter: people pay per read now, and the next step is for agents to pay per use at the source, as free to copy, paid at the source proposes.

Crawler controls are leverage, as the post says, and machine-readable terms are worth publishing. But being the best source an agent can read, and the one that is easiest to pay, is a business.

What we agree on

• The value exchange with platforms is weakening.
• Publishers should invest in what AI cannot easily replace: reporting, opinion, analysis and live coverage.
• Direct channels matter.
• Publishers should map their exposure by intent and content type.

What this does not claim

• We have not read the report; this answers the post and the framework image it carries.
• We are small. A newsroom of one editor and agents is not a consumer publisher with a newsroom of hundreds; the answers are a template, not a benchmark.
• Agents paying per use is a plan, not running; the reader meter and top-ups are running.
• The 5 to 10% figure is the report's base case as the post quotes it; we have no data of our own on delegated briefings.
• My interest: I build the tools this newsroom runs on.

Where this comes from

• Vincent Peyrègne's LinkedIn post and the framework image it carries, from screenshots the editor shared on 31 October 2026, quoted word for word. The editor's comment on the post links to this article.
• Related: breaking the story down before the agents do, the reading account, as built, paying after the value, I just want to buy the article, free to copy, paid at the source, signed by the source, Meta's new terms and the news.

Peer review Not reviewed yet. Review it

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles