As users stop going to publisher-driven platforms for information and begin asking AI agents to bring those search results to them, the power structure and value proposition of the publishing industry changes, argues Michael Cairns.

By Michael Cairns

For most of history, publishers have exercised control by managing supply. Publishers decided what was published, when and where it appeared, how it was structured, and how it could be read. This model governed research and professional content, where scarcity, prestige, and distribution barriers established the publisher’s control. In this world, users adapted to the content, and their workflows and processes were governed by the methods employed by the publisher. Increasingly, the opposite situation is developing as AI tools gain acceptance and become embedded.

AI is shifting value away from static units of content and toward dynamic use. This isn’t happening overtly, of course, but that is the effective reality. Researchers, professionals, and students increasingly expect to interrogate, remix, summarize, and contextualize information on their own terms. Instead of navigating publisher‑defined products—journals, books, databases—users are defining their own workflows and asking systems to pull insight from content wherever it lives. Prompt engineering becomes both art form and practical need. The locus of control moves from supply to demand, from what is published to how it is used.

The impact on publishers is already evident. Traditional value propositions built around access, distribution, and format are under pressure. When AI tools can synthesize thousands of articles into a single response, the journal issue or monograph becomes less visible, even if the underlying content, structure, and method remains critical.

This is not simply a discoverability problem; it is a power shift. Control based on gatekeeping access weakens when users no longer need to “enter” publisher platforms to extract value.

This creates a real risk of disintermediation. Publisher brands can be reduced to upstream data sources powering someone else’s interface, with little direct relationship to the end user. At the same time, users gain new leverage. They increasingly decide which content integrates cleanly into AI‑mediated workflows and which does not. What content is useful and what is fluff.  (We will find out there’s a lot of fluff). Content that is difficult to discover, poorly structured, or legally ambiguous is or will be simply bypassed. Demand, rather than supply, defines relevance.

‘A Particular Kind of Power’

Many publishers are already responding to this shift.

Elsevier’s development of AI‑enabled products such as Scopus AI reflects a move away from passive hosting toward active participation in researcher workflows—answering questions, mapping concepts, and accelerating discovery. Wolters Kluwer has embedded AI across its professional platforms to support decision‑making rather than document retrieval, recognizing that its users value outcomes over exhaustive research and reading. Springer Nature’s partnerships around AI‑assisted writing and review tools similarly acknowledge that interaction now matters as much as publication.

These examples illustrate an important point: publishers are not necessarily losing power, but they are losing a particular kind of power. Control no longer comes from restricting access or enforcing rigid usage models. Instead, influence comes from being indispensable within workflows defined by users and mediated by AI. The value proposition in metadata, taxonomy, or simple organization, which may have been obscure, may now attain greater recognition.  In the past, content owners banked on friction to support their value  prop: Power now shifts from gatekeeping to enablement.

A critical enabler of this transition is metadata. In an AI‑driven environment, content that is not well described is effectively invisible. Rich, consistent, machine‑readable metadata allows systems to understand what content is about, how it relates to other work, and why it should be surfaced in response to a particular question. Metadata provides semantic meaning rather than surface description, enabling discovery based on intent rather than keyword coincidence. Taxonomic management should be prioritized.

Persistent identifiers, structured abstracts, domain‑specific taxonomies, and clear rights metadata ensure that content can travel across platforms without losing context, attribution, or provenance. As discovery increasingly happens outside publisher‑owned environments—within AI assistants, enterprise systems, and institutional tools—metadata becomes the primary mechanism through which publishers maintain visibility,  relevance, and ownership. In this sense, metadata is no longer back‑office infrastructure; it is strategic capital.

‘Answers’

Metadata also plays a central role in trust. In a landscape increasingly populated by synthetic and unverified material, signals such as peer‑review status, editorial oversight, versioning, and correction history influence what AI systems choose to surface. Users may not see these signals directly, but the systems acting on their behalf do. Publishers who expose trust markers clearly increase both discoverability and credibility at precisely the moment when reliability is questioned.

The balance of power that emerges may be asymmetrical but will not zero‑sum. Users now dominate the how of interaction: how content is queried, combined, and applied. Publishers retain influence over the quality of what circulates: accuracy, provenance, editorial accountability, and long‑term stewardship. What has changed is that authority must now be earned at the point of use, not assumed at the point of publication and preeminence is not assured.

This reality requires concrete change. Content must be structured and enriched at the source, not retrofitted. Licensing models need to articulate clearly how content can be indexed, summarized, cited, or embedded by AI systems. Publishers must design for modular, question‑based discovery, recognizing that users increasingly want answers rather than documents.  Publishers may have to eliminate content silos and/or consider broader partnership arrangements.

High‑quality content remains essential fuel for AI, and publishers are uniquely positioned as creators, mentors and custodians. But relevance in an AI‑driven world no longer flows from owning supply. It flows from aligning with demand—supporting how knowledge is actually accessed, used, and interpreted.

Publishers who recognize this shift, and adapt accordingly, can remain central to the ecosystems they helped build. Those who cling to a supply‑led model risk discovering that control, once lost, is not replaced by negotiation, but by irrelevance.

This piece was originally published on Michael Cairn’s blog, Personanondata.

Michael Cairns has served as President, CEO and CRO of several brand name publishing, technology and content-centric businesses. His career includes a wide range of publishing and information products, services and B2B categories and his operating and consulting experience has largely been with brand-name companies such as American Psychological Association, Reed Elsevier, Wolters Kluwer, PriceWaterhouseCoopers, Macmillan, Inc., Berlitz International, AARP, R.R. Bowker and IEEE. Mr. Cairns has also served on boards including the Association of American Publishers (AAP), the Book Industry Study Group (BISG) and also served as Chairman of the International ISBN Executive Committee.