May 15, 2026
Jon Earnshaw had two hours until his BrightonSEO talk when Martijn Hoving and Diantha van Surksum sat him down backstage. He had spent months on this one. His stated mission, year after year, is the same: to make the audience think about the world differently. This year, he had his clearest evidence yet.
In Jon’s live demo at BrightonSEO, a site called RunRepeat was pulling more AI citations for running shoe conversations than Nike, Adidas, and On combined. RunRepeat cuts shoes in half in a lab and publishes the data. It doesn’t even sell running shoes.
Jon clicked through one of their citations mid-demo and nearly bought a pair. “I don’t run,” he said. “But I nearly bought running shoes.”
That outcome has three causes. Jon attributes all three to a phrase by Liz Reid. Google patented the scoring mechanism four years ago.
Why can't you optimize a page for a prompt?
The SEO industry's first instinct when a new ranking surface arrives is to find the equivalent of the keyword. For AI search, that instinct produced prompt tracking: pick the queries relevant to your brand, run them daily, and record whether your domain appears.
Jon Earnshaw has watched this pattern for twenty years, and he's direct about it:
The structural reason: an AI search prompt returns results from multiple pages, videos, forum threads, and other sources, assembled into a response. There is no single page to optimize for it. The ranking surface is the conversation, not the document.
Jon cites a Liz Reid data point: users entering AI Mode are now asking at an average of around sixty words. A year earlier, the average was twenty-three. The queries aren't shorter and more keyword-like in AI search. They're longer, more exploratory, and more specific than anything traditional SEO keyword research ever modeled.
What is a conversational space, and how do you build one?
The unit of optimization Jon has been building toward for two decades is the conversational space: a mapped set of prompts that covers all the discussions a relevant audience is actually having on a given topic.
He demonstrated it with an example. A luxury fashion brand thinking about Wimbledon wouldn't ask "what keywords should we target?" They'd ask what conversations people are having. Wimbledon dress code. Wimbledon glamour. What brands are people wearing in which seats? Those are real conversations, each of which maps to a cluster of prompts.
The research method Jon describes for a barbecue brand is more direct still: he stood in a barbecue store for an hour and listened. The conversations were all about accessories. He built the conversational space from what customers said to one another, not from what the brand assumed they were searching for.
The output is a set of prompts that actually map the discussion. Coverage in those prompts, tracked daily, becomes the KPI.
When you're visible in a conversational space, you have presence. When you're absent despite strong organic rankings, Jon calls it a visibility paradox: authority that hasn't transferred into AI citation. That gap points directly at the content.
Ready to start scaling your business?
Let's get started
What does depth, uniqueness, and originality actually require?
When the citation doorway isn't opening despite brand strength, Jon is definitive about the cause:
Jon attributes the phrase to Liz Reid. She used it to describe the content threshold Google requires before opening a citation: a reason good enough to pull someone out of an ongoing AI conversation and send them to an external page.
RunRepeat clears that threshold in every running conversation because the content is physically impossible to replicate without the same equipment and methodology.
They test shoes in a lab. They cut them in half. The data they publish on midsole compression, outsole durability, and stack height is primary research that no brand description page or affiliate review can reproduce.
Nike and Adidas produce content that describes their products. RunRepeat produces content that tests products they don't even sell. The AI system, faced with a query about running shoes, has an obvious choice: send the user to a brand page or to a lab report.
What did Google already tell us about this in 2022?
The three-word framework Liz Reid used has a technical counterpart that has been in the public record since 2022.
Google's information gain patent (US20200349181A1, filed 2018, granted June 2022) describes a scoring mechanism for how much new information a piece of content adds relative to the existing corpus.
The patent evaluates uniqueness at the sentence and phrase level using vector embeddings. Content that adds nothing new scores low. Content that adds substantive information the user hasn't already encountered scores high.
That is the technical definition of depth, uniqueness, and originality.
The SEO community discussed the patent briefly when it was granted in 2022, then moved on. It didn't fit the keyword optimization playbook cleanly. Now it fits exactly. AI search systems that must decide whether to cite a source or keep the conversation going are making a version of the information gain calculation with every response.
Is this content worth leaving for? Does it add enough to make it better for the user to go there?
RunRepeat passes that test. A page that rephrases the product description in slightly different words does not.
The visibility paradox Jon describes (strong organic visibility, absent AI citations) is the information gain test revealing itself. Classic ranking rewards relevance to a query. Information gain rewards contribution to a corpus. A page can be highly relevant and contribute nothing new. That page will rank organically and miss citations.
What happens when business agents start shopping for your customers?
Jon's longer-horizon argument moves the stakes further. The Macy's case from March 2026 is his anchor.
Macy's launched Ask Macy's, a conversational shopping agent powered by Google Gemini, across all digital channels. Revenue per visit among users who engaged with the agent was 4.75 times higher than among those who didn't. The agent went out, had a conversation with the customer, and closed transactions at a scale no static page had reached.
The implication Jon draws from this is structural:
When a personal agent shops on a customer's behalf, it doesn't browse brand pages. It interrogates business agents. Two agents have a machine-to-machine conversation. The human is downstream of that decision. What gets interpreted in that exchange is not the brand's homepage. It's the entity: the product attributes, the consistency of information across surfaces, the specificity of the data the agent can retrieve.
Entity consistency (the same description, the same attributes, the same structured data across every surface) stops being a hygiene task and becomes the product. An agent that encounters contradictory or generic information about a brand will route to the brand that has its attributes tightened.
Jon's advice for teams not yet in the Google Merchant Center Business Agent program: build a GPT, train it, link it to your site, and test it with colleagues. Ask it whether your website has enough content to do a conversation justice. The answer tells you everything about your information-gain problem.
Ready to start scaling your business?
Let's get started
What is fast content, and why does it favor smaller brands?
The content type that clears the depth, uniqueness, and originality bar fastest is what Jon calls fast content: first-hand, authentic, social, and trusted. A short video of a real customer talking honestly about a product. An interview with someone who actually used the thing.
The structural reason fast content wins is that large brands are slow. They have approval processes, brand guidelines, and legal review. A smaller brand with a genuine customer relationship can produce a first-hand account the same day an event happens. That account adds information no competitor has. It's primary. It passes the information gain test. It contributes something the corpus doesn't already contain.
For smaller businesses operating in categories dominated by large brands, this is a structural advantage. Large brands are not failing to cite AI because they lack authority.
They're failing because their content is generic. The smaller brand with a lab, a real customer community, or the willingness to stand in a barbecue store for an hour has information the large brand doesn't.
The hot topics failure Jon Earnshaw still carries
Jon's biggest SEO mistake happened at a media company. He'd been inspired by work with CNN on their Donald Trump content, where thousands of conflicting pages were cannibalizing each other. He fixed that. Emboldened, he took the hot topics concept to a UK TV broadcaster and convinced them to invest in large topic hub pages that would own the breaking news conversation.
It was an abysmal failure. The traffic never came. The investment was wasted.
Twenty years later, the lesson he draws from it is not that the strategy was wrong but that the execution and timing were missed. Topic authority (owning the conversational space around a subject) is precisely what he is talking about now. The media company built the pages before the moment when it would have worked.
The conversation, it turned out, was always the right unit. The technology has finally caught up.
Resources
- Pi Datametrics: Jon Earnshaw's search intelligence platform, built around conversational space tracking and citation acquisition
- Google information gain patent: US20200349181A1, "Contextual Estimation of Link Information Gain," granted June 2022
- RunRepeat: the running shoe site that cuts shoes in half. The clearest live example of depth, uniqueness, and originality in practice
- Fortune: Macy's AI shopping assistant drives 4.75x spend
- Search Engine Journal: Google's information gain patent explained
Listen to the full conversation
The full SEO Cast interview with Jon Earnshaw runs in English and covers his full BrightonSEO talk preview: conversational spaces, citation doorways, business agents, and what the Macy's data means for every brand now building out its AI presence. The RunRepeat example lands about two-thirds of the way through and is worth the whole episode. Listen on Spotify.
Head of SEO
As Head of SEO at Seeders, Hans helps companies scale smarter by combining search behaviour, paid performance, and conversion insights into systems that drive real growth. Recognised as a Top 20 Content Marketer by Small Business Trends, he brings a lean, experimental approach focused on building what actually works. Rather than relying on reports or one-off strategies, Hans focuses on creating marketing systems and tools that turn visibility into measurable business impact. His goal is simple: help clients move from scattered efforts to structured growth.
