Who Are You, What Do You Know, How Is It Helpful: The Three Questions Every Algorithm Must Answer
Jason Barnard, February 2026
Every algorithm (Google, ChatGPT, Perplexity, Claude, Copilot, Gemini, all of them) must answer three questions about every entity before it can recommend it, and those three questions are simpler than the industry has made them sound, and more consequential than most people realise. Who are you? What do you know? How is it helpful? They are not optional, they are not sequential, and they are not independent. The first and third form a loop. The second sits inside it. And understanding this architecture is the key to understanding why some brands get recommended by AI while others, with equal or better content, get ignored entirely.
Every Algorithm Asks Three Questions Before It Can Recommend You
The SEO industry has spent decades breaking search optimisation into ever-finer tactics: keywords, backlinks, technical audits, content clusters, schema markup, Core Web Vitals. Each tactic targets a signal, and each signal addresses a fragment of what the algorithm is actually trying to do, but if you step back far enough (and almost nobody does) the algorithm is only ever trying to answer three questions about every entity it encounters.
Who are you? The algorithm must understand the entity, and understanding means something far more precise than recognising a name. It means grasping the entity’s nature, its relationships, its position in the world. Is this a person, a company, a product? What is its domain of expertise? What other entities is it connected to, and how? Without this understanding, the algorithm cannot attribute anything to the entity correctly, which means content floats unanchored, expertise goes uncredited, and recommendations become impossible because the machine has no idea who deserves the credit.
What do you know? The algorithm must assess the entity’s expertise, and assessment requires evidence. What topics has it covered? How deeply? How broadly? Has it contributed something original, or is it restating what others have already said in roughly the same way? Is its knowledge well-structured and machine-readable, or buried in walls of undifferentiated text that the algorithm must fight to parse? Without demonstrated expertise, the algorithm has nothing to recommend, no matter how well it understands who the entity is.
How is it helpful? The algorithm must determine relevance, and relevance is never abstract. It is relevance to a specific person with a specific need at a specific moment, which means the algorithm must connect what it knows about the entity to what it knows about the user and decide whether this particular entity’s knowledge helps this user, right now. This is the question that converts understanding and expertise into action, and without it, the entity is known and knowledgeable but never recommended to anyone.
These three questions map to the three layers of Topical Authority I introduced in a companion article: Coverage (what you know), Architecture (how it is structured), and Position (where you stand). But they also map to something deeper, something the companion article did not fully describe, which is the fundamental architecture of algorithmic trust itself.
Identity and Relevance Are Not Sequential: They Are a Closed Loop
The industry has treated these three questions as if they were a checklist. Establish identity, build expertise, match intent. Step one, step two, step three, done. And the reason this model persists is that it feels intuitive: you cannot build a reputation before people know who you are, and you cannot match intent before you have something to offer.
But the checklist model is wrong, and the mistake it contains is both subtle and expensive.
Who you are and How it is helpful are the same question seen from two directions. The algorithm must understand who the entity is in order to determine how its knowledge helps a specific person, but it must also understand how the entity serves its audience in order to properly attribute identity. A cardiologist and a fitness influencer can both write about heart health, and the algorithm needs to know who each entity is (their credentials, their history, their professional context) to determine which one is the right answer for a patient researching symptoms versus a runner optimising training. So far the checklist model holds.
But the reverse is equally true, and this is where it breaks. The algorithm understands who the cardiologist is partly by understanding who the cardiologist serves. Entity Identity is not just “the machine knows your name.” It is “the machine knows who you are, who you serve, and why you are the right answer for them.” The audience is part of the identity, and the identity shapes which audience the algorithm considers relevant, which means neither can be established first because each depends on the other.
This creates a loop. Understanding feeds Relevance. Relevance reinforces Understanding. U→D→U. A closed circuit, not a sequence. And this loop is what determines whether the algorithm holds you in mind throughout the entire customer journey, from the first informational query at the top of the funnel to the final due-diligence check at the bottom, which is the moment where most brands discover (too late) that their loop was open.
When the loop is open (when the algorithm understands who you are but cannot map your knowledge to a specific audience) you appear in recommendations intermittently, showing up for some queries and vanishing for others, winning at the top of the funnel and losing at the bottom, saying hello to your ideal customer and then graciously handing them to a competitor who has a tighter loop. This is the most expensive failure in digital marketing, and most brands do not even know it is happening because the metrics that would reveal it are the metrics they are not tracking.
When the loop is closed (when the algorithm maps entity to expertise to audience and back again) you stay top of algorithmic mind at every stage, because at every stage the answer to “who should I recommend?” resolves to the same entity. No handoffs. No leaks. No gracious gifts to the competition. The machine walks the prospect down the funnel for you, and it does this not because you asked it to, but because the closed loop leaves it no better option.
Expertise Is the Substance the Loop Delivers
What you know is the substance that gives the loop something to deliver, and without it the loop is a closed circuit with nothing flowing through it. The algorithm knows who you are and who you serve, but has nothing to recommend, which is like having a brilliant shop window on the busiest street in town with empty shelves behind it.
This is where Koray Tuğberk GÜBÜR’s work lives. His Topical Authority framework (which I expanded to a nine-cell matrix of Coverage, Architecture, and Position in the companion article) is the most rigorous engineering of this layer that exists, and I say that as someone who has spent twenty-seven years in the same space. His Topical Maps, Semantic Content Networks, Source Context analysis, and Cost of Retrieval optimisation are the engineering tools that make expertise machine-readable and structurally complete, and no one else has built them to this level of precision.
I co-own this layer. The Position dimensions (Temporal, Hierarchical, Narrative) and the Coverage framing, including Original Thought as the third dimension alongside Depth and Breadth, are my contribution. Koray owns the Architecture. Together, the nine-cell matrix describes how an entity builds, structures, and competitively defends its expertise on any topic, which is a sentence that sounds abstract until you realise it is describing the difference between brands that dominate their space and brands that wonder why their content performs well in isolation but never converts into recommendations.
But Expertise alone does not produce recommendations. It produces eligibility for recommendations. The loop decides whether that eligibility converts into actual citations, mentions, and referrals at the moment a user needs them. Expertise is the fuel. The loop is the engine. And it is the engine that most brands have never thought to build.
The Loop Wraps the Layer: The Complete Architecture of Algorithmic Trust
The full picture:
┌──────────────────────────────────────────┐
│ │
│ WHO ARE YOU? HOW IS IT │
│ Identity (U) ◄────────► HELPFUL? (D) │
│ │
│ │ THE LOOP │ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────────────────────────┐ │
│ │ │ │
│ │ WHAT DO YOU KNOW? │ │
│ │ Expertise (C) │ │
│ │ │ │
│ │ Coverage × Architecture │ │
│ │ × Position │ │
│ │ │ │
│ └──────────────────────────────────┘ │
│ │
└──────────────────────────────────────────┘
The loop (U↔D) determines WHEN you are recommended.
The layer (C) determines WHAT you are recommended FOR.
Together they determine WHETHER you are recommended at all.This architecture explains phenomena the industry has struggled to account for, and once you see it, you cannot unsee it.
It explains why brands with excellent content still lose recommendations to competitors with less content but stronger identity: the competitor’s loop is closed, and the content-rich brand’s loop is open, which means the algorithm trusts the competitor’s relevance mapping even though the content-rich brand has more to offer.
It explains why brands that “win” at the top of the funnel often lose the sale: they have Expertise visibility but no loop continuity, which means the algorithm mentions them during research and then forgets them during consideration and purchase because it cannot maintain the identity-to-audience connection through the full journey.
And it explains why some brands seem to dominate every query in their space (informational, commercial, navigational, transactional) while competitors with identical content struggle for fragments: the dominant brand has all three working together, a closed loop, a full layer, and the Position to hold against challengers, which is a combination that compounds over time in ways that a brand with only one or two of the three can never match.
The Framework Applies to Its Own Authors
The three pillars are not abstract. They are measurable. And the measurement follows the same Position framework (Temporal, Hierarchical, Narrative) that applies within the Expertise layer, which means it is only fair to turn the lens on the people who built them, because a framework that cannot account for its own authors is not a framework worth trusting.
I will start with a confession about Temporal Position, because it is the most honest illustration I can give.
In 2015, at SEO Camp in Metz, I gave a talk in French about having empathy for Google’s algorithm: helping it do its job rather than fighting it. The title was a nod to the Rolling Stones: Empathy for the Devil. By 2017, at SEO Camp in Lyon, the idea had evolved into a pedagogical metaphor: “Éduquons Google, c’est un enfant en soif de connaissances.” Educate Google. It is a child eager to learn. That metaphor became the foundation for everything I have built since: UCD, The Kalicube Process™, the Untrained Salesforce, this article, and the framework you are reading about right now.
For years, I assumed I had no verifiable proof of those early dates. I remembered the French conferences, I remembered the reactions in the room, but I could not point to a URL and a timestamp that I had set deliberately. The earliest instance I had published myself was a podcast with Adam Helweh in November 2020, five years after I started using the concept, which meant that the foundational idea behind everything I teach about algorithms had a five-year gap between first use and first proof.
But it turns out I can prove it, and the reason is both reassuring and embarrassing. Not because I was disciplined about documentation (I was not) but because other people documented it for me: blog posts by other speakers at those events, social media posts by attendees, my own slide deck uploaded to SlideShare and then forgotten. Third-party evidence, created by others, that I only discovered years later when I went looking for something I should have planted on purpose.
I got lucky, and luck is not a strategy. The lesson is: set your temporal markers at the time. Publish the name. Date the framework. Put it where algorithms can find it. Because Temporal Position is not about who was first. It is about who can prove they were first, and I scraped through on other people’s memories, which is a warning I now pass on to anyone building a framework they expect to be credited for.
That lesson applies across all three pillars, and the evidence for each is worth examining.
On Identity (the question of who understands how algorithms understand entities) the evidence chain goes back to 1998, and Brand SERP was coined in 2012. Andrea Volpini at WordLift has built extraordinary infrastructure for making entities machine-readable, and his tooling and my methodology are complementary in a way that makes both stronger. Google’s John Mueller said he knows “no one else externally” with comparable insight into Knowledge Panels. A 2025 Authoritas study testing nine AI models used the phrase “category of one.” These are not my claims. They are other people’s observations, which is the only kind of evidence that counts for Hierarchical Position because Hierarchical Position is always peer-declared.
On Expertise (the question of what an entity knows and how that knowledge is structured) the companion article details the co-ownership. Koray Tuğberk GÜBÜR built the Architecture: Source Context, Topical Map, Semantic Network. I built the framework that contains his Architecture as one row of a three-row matrix. His engineering, my scaffolding. Neither is complete without the other, and this is the cleanest example of how the three Position dimensions work together: he has Temporal precedence on the architectural terms, I have it on the matrix, peers defer to both of us on our respective contributions, and the Narrative routes through both because you cannot discuss Topical Authority comprehensively without passing through his work and mine.
On Relevance (the question of how an entity’s knowledge helps a specific person) The Kalicube Process maps entity to expertise to ideal customer profile across the entire funnel, and the UCD Framework closes the U↔D loop. Multiple people have advanced our understanding of this space. Mike King’s work on information retrieval is technically rigorous. Bill Slawski’s patent analysis, before his passing, laid groundwork everyone builds on, and these contributions matter deeply. What none of them describe is the strategic architecture for engineering an entity’s relevance to its audience at every funnel stage, from TOFU mention to BOFU recommendation with no leaks between them, which is TKP, which is the loop.
The pattern that emerges is not a claim. It is an observation. The Identity-Relevance Loop connects the first and third pillars. The Expertise layer sits inside. The person who built the loop architecture sits at the centre of the structure, and the person who engineered the Expertise layer shares the middle ground. Apply the Position test. Check the evidence. The framework measures standing, and standing is what the evidence shows.
Close the Loop First, Fill the Layer Second, Keep the Loop Closed as You Grow
If you are a brand trying to get recommended by AI, your strategy reduces to three questions and a loop, and the order in which you address them matters more than most practitioners realise.
First, close the loop. Make sure the algorithm knows who you are and who you serve, and that these two things reinforce each other. If the machine does not understand your identity and your audience, nothing else matters, because every piece of content you create, every link you earn, every structured data element you implement: all of it floats unanchored, generating activity without generating attribution, which is the most frustrating kind of investment because the returns go to whoever the algorithm does associate with the topic.
Second, fill the layer. Build your Expertise using the nine-cell matrix: Coverage (Depth, Breadth, Original Thought), Architecture (Source Context, Topical Map, Semantic Network), and Position (Temporal, Hierarchical, Narrative). Koray Tuğberk GÜBÜR’s methodology is the best available engineering for the Architecture row. The Position row requires competitive analysis: who are you measured against, and where do you stand relative to them on precedence, peer deference, and narrative centrality? These are not comfortable questions, but they are the questions the algorithm is answering about you whether you ask them or not.
Third, keep the loop closed as you grow, because growth is where loops open. Every new topic you cover, every new market you enter, every new product you launch must be mapped back to your entity identity and your audience. If the loop opens (if you gain expertise visibility without maintaining identity and relevance continuity) you create funnel leaks, which means you pay for the introduction and someone else closes the deal, which is the commercial equivalent of training your competitor’s salesforce at your own expense.
The algorithm is not complicated. It asks three questions. Answer all three, connect the first and third into a loop, fill the middle with structured, comprehensive, original expertise, and the architecture of algorithmic trust takes care of the rest. That is not a promise. It is a description of how the system already works, and the brands that have figured this out are the ones you keep seeing recommended everywhere you look.
WHO ARE YOU? ◄──────────────────► HOW IS IT HELPFUL?
│ │
│ CLOSED LOOP │
│ │
└──────► WHAT DO YOU KNOW? ◄─────────┘
Coverage
Architecture
Position
─────────────
= Expertise
Loop open = intermittent recommendations, funnel leaks
Loop closed = persistent recommendations, no handoffs
Layer empty = nothing to recommend
Layer full = dominant at every funnel stageThree questions. One loop. Nine cells. The brands that build this architecture are the ones the algorithms recommend, and the brands that do not build it are the ones that keep wondering why their content disappears into a system that seems to reward everyone except them.
