Wednesday, October 07, 2026
Wednesday, October 07, 2026
Mike Klein:
AI search is building a dominant position as people and organizations seek insights and information. But the platforms and strategies for measuring success and maximizing impact in the space are nowhere near settled.
That gap – discussed at length in Muck Rack’s Generative Pulse Summit in London this week is the most significant structural opportunity in the communication world right now. It is also where the interests of brand communicators, PR agencies, and established publishers are converging in ways that none of them have fully mapped.
Muck Rack’s “What Is AI Reading?” research – released in UK-specific form for the first time at the Generative Pulse Summit week – puts a number on something that many practitioners suspected but couldn’t quantify. Approximately 82% of AI citations come from earned media. Paid content registers barely at all. Owned content accounts for around 15%.
That 82% figure for third-party earned content is the one that changes how strategy should be built. It means the single most effective lever for influencing what AI says about an organization is what other credible sources say about it.
Not what the organization says about itself.
What it earns through journalism, through expert commentary, through community discussion, through the long tail of citations that accumulate in crawlable formats across the open web.
For communication professionals, this should feel like a structural advantage. For paid media teams, it is a structural disruption.
The complication is that “earned media” is not a monolithic category.
It encompasses everything from Financial Times investigations to Reddit threads to consumer testing reports from the British consumer magazine “Which.” And those sources are not equivalent in AI’s eyes – or, more precisely, in the eyes of the specific LLMs that billions of people are using to find information.
Here is where the strategic picture gets complicated.
Several of the world’s most prestigious publications – The Times, The Telegraph, The New York Times, currently block AI crawlers entirely.
These publications receive zero citations in AI responses as a result. The Financial Times has done partial deals and receives some citations, but far fewer than comparable outlets that allow open crawling.
This is not a permanent situation. These publishers are holding out for commercial agreements with the major LLM platforms – the kind of licensing deal that would compensate them for the content that AI systems were trained on and continue to retrieve. Some version of those deals will eventually be struck.
But “eventually” may be a year or two away. In the meantime, a placement in The Times – which remains enormously valuable for its human readership and its institutional signaling – contributes nothing to what AI says about a brand or organization.
This creates a genuine strategic tension for communications teams. Do you allocate resources toward the outlets that AI currently favors, even if they carry less traditional prestige? Or do you maintain relationships with the blockade-holding tier-one publications against the day when their deals are done and their content floods back into the AI citation pool?
The honest answer is both, but the balance has shifted. Building a pitching strategy exclusively around traditional prestige outlets is no longer sufficient if AI visibility is a genuine objective.
The Muck Rack data introduced one finding that stops practitioners mid-sentence: “Which,” the UK consumer product testing publication, emerges as the single most-cited UK authority in AI responses.
Not the Guardian. Not the BBC. “Which.”
The reasons are structural rather than editorial. “Which” allows AI crawlers. It publishes in formats that LLMs can easily extract and repurpose – structured product evaluations, comparison tables, clear verdicts. It carries genuine topical authority in the domains AI users frequently query. And it has not – yet – done the kind of blocking or deal-seeking that has removed other authoritative voices from the AI citation ecosystem.
The third dimension of this picture is measurement, and it is the one where the field is most visibly under construction.
At the Generative Pulse Summit, Shell’s Kyle Mason pushed back on the temptation to treat citation count as the new headline metric. Being cited is not the same as the right thing being said. The relevant question is whether the content AI is surfacing matches what an organization actually wants communicated to the audience that matters.
That is a harder metric to build than citation volume. It requires ongoing prompt-testing across multiple LLMs, comparison against organizational messaging objectives, and the kind of rapid course-correction capacity that most communications teams have yet to build.
The European Central Bank’s Martin Zwiebelberg captured the institutional version of this challenge. The ECB has seen 30-40% traffic decline to its website as AI search takes over a portion of what used to be direct source visits. Google AI Overviews are drawing heavily on ECB content and distributing it accurately – but without the attribution that would bring users back to the institution. The organization is generating influence without generating engagement, which is a fundamentally new problem for a body that depends on public trust and market credibility.
What the ECB experiences at an institutional level, brands experience commercially. AI may be surfacing accurate product information – and users may never click through to the brand site that produced it. Attribution in the AI search layer is not resolved. The metrics to capture it do not yet exist at scale.
Which brings us to a question that surfaced in the Generative Pulse Summit’s final session and deserves extended treatment.
If earned media dominates AI citations, and if the commercial market for AI visibility is not yet formalized, what is the current status of paid partnerships with publishers as an AI influence strategy?
Shell has run paid partnerships that were subsequently cited in LLM responses. Some worked. Some did not. The outcome appears to depend on the publisher (whether it allows crawler access), the content format (structured, extractable content performs better than rich graphic layouts), and the specific LLM in question.
Golin’s Jonny Bentwood framed the paid vs. earned question as a volume argument: one Forbes advertorial is one crawlable page; two hundred earned placements are two hundred pages. In current conditions, volume favors earned.
But the volume framing misses something.
A well-placed paid partnership in the right publication – one that AI currently favors for citations in a given category – carries a compounded return. It functions simultaneously as traditional earned media (reaching the publication’s human readership), as AI citation content (if the publisher is crawlable and the format is right), and as a branded association with the publication’s authority in the LLM’s assessment of source credibility.
That triple function is not well understood yet. It is not priced for yet. And it will not remain unpriced for long.
Here is the structural dynamic that the next 18 months will resolve.
The major publishers that currently block AI crawlers are doing so because they have not yet been adequately compensated. When that compensation comes – through licensing deals, through revenue-sharing arrangements, through whatever commercial model emerges – their content returns to the citation pool. The Times, The Telegraph, potentially the Financial Times at greater depth – these return as AI visibility assets.
When that happens, the value of premium editorial relationships with those publications compounds again. A placement that currently delivers only human readership will also deliver AI citation coverage.
More immediately, the publishers who are currently open to AI crawling have an underappreciated opportunity to develop paid partnership products that explicitly sell AI visibility alongside traditional readership reach. “Which,” alongside tech, trade, and specialist publications that currently dominate AI citation rankings – these are sitting on an asset whose value has not yet been articulated in commercial terms.
Brand communicators and agencies that understand this dynamic have a narrow window to build publisher relationships and partnership structures before those relationships are explicitly priced as AI visibility products.
The field is not waiting for permission. In the Q&A at the Generative Pulse Summit, an audience member asked the panel directly whether paid partnerships between corporates, agencies, and established publications can function as AI citation strategy. The honest answer from the panel was: yes, conditionally, and we’re still learning the conditions.
That is what the beginning of a market sounds like.
The AI search landscape is not settled. Every practitioner on the Generative Pulse Summit stage said some version of the same thing: what worked six months ago may not work today, and what works today will be updated in another six months. Model updates shift citation hierarchies. Crawler policies change. LLM platforms develop their own commercial interests that affect which content they surface.
What is stable is the underlying dynamic: AI systems need authoritative, crawlable, well-formatted content to answer the questions billions of people are asking them. Publishers that provide that content, in formats AI can use, from domains AI trusts, are the infrastructure of the answer-engine era.
The question for brand communicators is not whether to engage with that infrastructure. It is whether to build those relationships now, on terms that are still being set, or later, when the market has priced them in.
+++
This piece draws on reporting from the Muck Rack Generative Pulse Summit UK, London, May 12, 2026.
+++
Mike Klein is Editor-in-Chief of Strategic.
Contact editor@strategic.global or visit strategic.global/join to discuss whether your organization and its content are a fit for a Strategic corporate publishing partnership.
© 2026 Stratpair Ltd., trading as Strategic. Registered in Ireland: 747736
