Teodora Petkova on Meaning-Making, Non-Commodity Content, and the Living Knowledge Ecosystem
Session identifier
- Session: Meaning-Making, Non-Commodity Content, and the Living Knowledge Ecosystem
- Day: 1 (June 24, 2026)
- Speaker: Teodora Petkova, Knowledge Steward, Graphwise
- Framework concept: Semantic Annotation (philosophical), Gate 6 (Annotated) of the AI Engine Pipeline
- SEL article: How AI decides what your content means and why it gets you wrong
Speaker bio
Teodora is a philologist fascinated by the metamorphoses of text on the Web, curious about the unfolding of the Semantic Web, and deeply interested in how meaning emerges, flows, and connects people, knowledge, and technology.
Her work is grounded in the belief that content is not simply produced — it emerges from knowledge, relationships, and shared meaning. When organizations clarify how ideas connect, communication becomes more coherent, adaptable, and sustainable over time.
Her academic background is in Classical Studies and Creative Writing, and she holds a PhD in Marketing Communication from Sofia University. She is also the author of The Brave New Text and Being Dialogic.
With more than fifteen years of experience in content writing, digital marketing, and knowledge-related work, Teodora has worked across freelance, agency, and corporate environments, gradually moving toward the intersection of marketing communication and semantic technologies.
Today, she works as a Knowledge Steward at Graphwise, building bridges between marketing communication, content operations, and semantic technologies, and supporting the creation of structured, reusable knowledge assets. She also serves as an affiliated lecturer in the Master’s program in Content Strategy at FH JOANNEUM.
Session’s place in the Kalicube Framework
Teodora’s session covers Semantic Annotation, the philosophical and ontological dimension of Gate 6 (Annotated) of the AI Engine Pipeline, the hinge gate where the engine decides what content IS — the entities, topics, claims, and relationships inside it. Where Koray Tuğberk Gübür covers Annotation from the technical foundations angle, Teodora covers the semantic web tradition and the theory of meaning-making that underpins why annotation works the way it does. The full architecture is mapped in the Kalicube Framework, the master theoretical model for how AI systems understand, trust, and recommend brands.
Episode summary
Teodora Petkova opened with a provocation the session never left behind: knowledge is not a set of facts but a living ecosystem, and its glue is relationship and feeling. She drew on the academic concept of assemblage to name what knowledge graphs miss when they are treated as purely deterministic systems. A knowledge graph, she argued, is like planting a tree. Once it is out there, people interact with it in ways the builder cannot fully predict. The open world assumption is not a limitation but a feature — meaning grows at the edges. The session’s early terrain was the distinction between facts and statements. Teodora named the problem directly: a false statement is, in the universe of the person who created it, a fact. That makes the work of structured knowledge not about capturing truth but about modeling statements and their relationship to commonly accepted understanding.
The session’s central pivot was non-commodity content — her definition precise and useful: content that is statistically improbable, written at the edge of the author’s domain, where the reader encounters something they have not thought about before. The knowledge curse, Steven Pinker’s term for the gap between expert and audience, is what non-commodity content overcomes. A glossary, she argued, is the first act of any serious knowledge garden — the discipline of defining terms before anything else is written. Jason’s own experience building a locked lexicon to prevent AI engines from misinterpreting his coined terms landed as live confirmation: “I made a mistake, and rolling it back took me eight months.” The semantic capital argument followed directly. Organisations that invest in clarifying how their ideas connect to each other build a communication asset that compounds over time. In the AI era, that asset is the substrate that annotation builds on.
The session closed with SEO reframed. Teodora named what the Kalicube Framework has been building toward: that SEO and knowledge management are the same practice, both fundamentally about knowledge retrieval and structured expression. Curation, not control, is the right word for what brands do when they build their knowledge garden — inviting the public web’s relationships in, connecting their domain to the broader knowledge ecosystem, and building the substrate that AI engines need to understand a brand with confidence rather than by inference. The session’s strongest contribution is the philosophical foundations argument made operational: the reason annotation is the hinge gate is that meaning itself is relational, contextual, and grows from the edges of existing knowledge. That is not a feature of the Kalicube Framework’s design. It is how meaning works. Teodora’s session names why.
Key quotes
“Non-commodity content is being statistically improbable.” Teodora Petkova, naming the quality that makes content genuinely useful to AI engines: not time on page, not word count, but the moment the reader encounters something they have never thought about before”
“A knowledge graph is like planting a tree. It grows. There’s shade. Maybe a bench arrives somewhere. People sit, and they start talking.” Teodora Petkova, on the open world assumption as feature rather than limitation: the assemblage that grows beyond what the builder planned”
“Build a glossary. For all the content writers out there, please start there before anything else.” Teodora Petkova, on the first act of any serious knowledge garden: the discipline of defining terms before anything else is written”
“SEO and knowledge management are intrinsically linked. When you look at them in depth, they are one and the same practice, both relating to knowledge retrieval.” Teodora Petkova, naming the convergence the Kalicube Framework has been building toward: SEO as the discipline of curating a knowledge garden for the machines”
“The devil is in the details.” Teodora Petkova, closing the session with the Brazil film reference — the fly in the typewriter mechanism — as the practitioner’s reminder that precision in the semantic layer is not optional”
Further reading
How AI decides what your content means and why it gets you wrong, Article 9 of Jason Barnard at Kalicube’s Search Engine Land series on AI search architecture. The article maps the Annotation gate in full — why Gate 6 is the hinge gate of the entire AI Engine Pipeline, what the engine is actually doing when it decides what content IS, and why getting annotation right is the precondition for every gate that follows.
Related sessions
- Dixon Jones on Entity Home (Session 1.5): the foundation layer that Entity Home provides across all phases of the pipeline — the entity must be anchored before Annotation can classify content about it correctly
- The Day 1 live framework walkthrough with Jason Barnard and Nik Ranger (Session 1.0): the architectural map of all ten gates that the Annotation gate sits at the centre of — Teodora’s session is the deepest philosophical treatment of why the hinge gate works the way it does
Link to preview post
This session was previewed here before the Summit.
Link to speaker post
Links to session clips
The Knowledge Curse: Teodora Petkova on Why Experts Struggle to Explain Anything
Why Good Content Should Be Statistically Improbable: Teodora Petkova
Transcript for Teodora Petkova on Meaning-Making, Non-Commodity Content, and the Living Knowledge Ecosystem
Jason Barnard: Welcome to this session. Teodora, thank you for joining me. We’re going to be talking about non-commodity content, emotional memory, and brands, and I want to talk about annotation because it’s my favorite part of the SEO journey at the moment. Welcome.
Teodora Petkova: Thanks for having me. I’m excited. And what a nice T-shirt. I love the color. I can feel it from here.
Jason Barnard: I love that you can feel it. We were talking just before — I can feel it too, and I bought the shirt tailor-made because it makes me feel good.
Teodora Petkova: Yes, exactly. The experiential side of knowing. How do we know? Through facts, or through the goosebumps we get sometimes?
Jason Barnard: Which is interesting for me because I’m obsessed by Google’s Knowledge Graph, which is facts, and then LLMs write predictive text which doesn’t have feelings, and search engines surface what they believe to be most pertinent. Where does feeling come into AI Assistive Engine Optimization?
Teodora Petkova: Wow. Where does feeling come into AI Assistive Engine Optimization? I want to write that down. Why? Because you’re mentioning all these words, and these words have been my notes, the points I have been traveling through — or, in the language of knowledge graphs, traversing.
I want to travel between SEO, between meaning, between optimization. But as I’m growing up, I realize we cannot live with facts alone. Facts and knowledge are a living ecosystem, and its glue is our relationships, the way we feel, the way we relate to objects. My son says I’m a bit hippie, but I want to ground this. There is this living knowledge, which is dynamic stasis — meaning it’s a static dynamics. It’s something held together by the movement of its parts, and that’s where feeling comes in. Then we have facts. There are industries and bodies of knowledge that are highly determined by rules and logic, and that’s fine. But that’s not the sea I navigate in marketing communications.
Jason Barnard: So the company you work for —
Teodora Petkova: WordLift. Let me help you.
Jason Barnard: Yes, thank you. Which is about facts, about organizing knowledge.
Teodora Petkova: It is, of course. But these are not either/or. That’s a false dichotomy — the idea that we have facts, and then we have me talking strangely about feelings. There’s an academic concept I found a year or two ago that brings them together: assemblage, a technical assemblage.
You have the knowledge graph, but you also have the ecosystem of people using it. You can touch it, you can define it, it’s deterministic. But then you have all the mysterious ways people would use it and enrich it. A knowledge graph, whether built with an LLM-assisted ontology, or by ingesting text, or by ingesting unstructured data — your HubSpot records, for example — once you build it and put it out there, people start interacting with it. It’s like planting a tree. It grows. There’s shade. Maybe a bench arrives somewhere. People sit, and they start talking.
The knowledge graph is beautiful. It’s incredibly difficult. I love the engineers who build it and the way they think. And at the same time, there’s this open world assumption — we can always add more.
Jason Barnard: I’m genuinely interested in that, because you’re talking about building a knowledge graph that you control, while Google is building a knowledge graph based on the open web, which it has a certain amount of control over but which grows organically. What are the differences between the way these need to be built? I’m obsessed by Google’s Knowledge Graph, and when I talk to WordLift about building a knowledge graph from data they have some control over, it doesn’t appeal to me in the same way. What appeals to me is the mess the web is.
Teodora Petkova: I love those two words, Jason — mess and control. Let me use the concept of dialogic communication to bring them together. But first: calling it a mess is not quite right. What you’re calling mess is actually organic growth alongside the environment it represents. Then we have something we’re growing as a garden. Imagine a store with a garden in front — beautiful, well-presented. That garden is our knowledge garden, and it’s what we want to show the world.
Getting back to control. I wouldn’t say control. I would say curate. We’re curating a knowledge garden. Not for Google — for people, for the future — because this is digital literacy. We’re expressing what we know, the knowledge of the company. Knowhow and knowledge are the edge of the company, and we’re representing that edge. Not only through a website with flashy images, but with knowledge, with facts, with structured expression.
Jason Barnard: Earlier I was talking with someone about the difference between truth and fact. What we put into our garden is our version of the facts, which is actually our truth. They’re statements.
Teodora Petkova: Exactly. I’m fortunate to work at my company because I unconsciously know certain things. I felt there was something wrong with the word “facts,” and now it appears to me the right word is “statements.” A colleague of mine has been working on a database of known fakes — modeling fake statements, modeling fake news. For the people who create that content, a false statement is, in their universe, a fact. You can build a knowledge graph out of statements that are not aligned with what is commonly perceived as true. So in this knowledge era, we need to be granular: not “facts that are not true,” but “statements that are not aligned with the commonly accepted understanding.” All that modeling is pure writing, thinking, and communicating.
Jason Barnard: Completely. And modeling false information is something Google, ChatGPT, and all the other major platforms have to do at scale, because it’s a huge problem. They need an algorithmic way to sort the wheat from the chaff. Modeling it means: feed an algorithm false statements at scale, show it what those statements should say instead, and get it to identify the path between the two so it can recognize the problem in the wild. Is that roughly right?
Teodora Petkova: Yes.
Jason Barnard: So I could be building a knowledge graph of my own. How lovely. Where does that take us for the broader question?
Teodora Petkova: There’s a concept I want to share: the knowledge curse. It’s from Steven Pinker. The knowledge curse is when you know something so deeply that you can’t explain it well anymore, because you assume others understand all your terms and concepts. You’re speaking, and the person on the other side is not inside your gardens. And getting back to what you said about glossaries —
Jason Barnard: Yes — I have this exact problem. I’m talking about things that haven’t been discussed in quite this way before, so I’m constantly coining new terms. Laurence O’Toole from Authoritas sent me a funny email saying: “Oh no, not another of your coined terms with an acronym. I can’t remember them all.” My problem is I have to create them, but then when I give them to the AI, the AI misinterprets them and keeps getting them wrong. So I’ve had to build a lexicon that is locked down with no ambiguity, so the AI doesn’t get it wrong.
Teodora Petkova: From knowing to acting. Number one: build a glossary. For all the content writers out there, please start there before anything else. This is something I’m doing now, six years into teaching web writing in the content strategy master’s program. Their task is to write a glossary term — very short, but that’s actually the hardest part.
Getting back to mess and control. Mess, yes; control, no. I would say curate. And curation ties back to the idea that you’re not just putting content on your website, you’re building a knowledge garden. That garden is not controlled, because you as a business entity are not an island. You exist in relationship to the mess. What WordLift does, what we also do, is enrich the knowledge graph with public knowledge graphs. You invite the outer world into your garden because you can’t talk about your topic without connecting it to everything it’s related to.
Tying this to dialogic communication: when you curate, you’re saying, “This term means that.” But you are not the only one who invented the term. You’re pulling threads from the mess. And we need to be aware, when we talk about Google’s Knowledge Graph, that it’s an entity created with certain goals and agenda. But let’s also think about all the public knowledge graphs — Wikidata being the biggest example.
Jason Barnard: Thank you. And I got carried away with the glossary point because I’ve been so obsessed by it for the last couple of years. Let me give you an example. The Algorithmic Trinity is LLM plus knowledge graph plus search engine — the trio of technologies all AI Assistive Engines need. LLM is intelligence, search engines are information, and a knowledge graph is validation. I initially said “web index” for search engine, then realized they all feed from the web index, so I couldn’t use “web index” to name just one of the three. I had to roll it back and correct the definition throughout my entire glossary and across all my articles. But by that time, the initial term had been digested by Google, ChatGPT, and the others, so I had to correct them through the public-facing lexicon on my website. The point is: I made a mistake, and rolling it back took me eight months.
Teodora Petkova: Yes.
Jason Barnard: Yesterday we were talking about non-commodity content, and I think that’s a good next step. Tell me what it means. Tell me what you mean by it.
Teodora Petkova: I’ll tell you what I mean by it. Recently, I find myself reading and remembering things through my own semantic network. A friend of mine, Peter Nikolov, who is an SEO, called me a few weeks ago and said: “Hey, did you see the non-commodity content article and the guidelines? You were right.” In a few words: to me, non-commodity content is being statistically improbable.
Jason Barnard: That woke me up. Explain it.
Teodora Petkova: In simpler terms for people who are starting to think about creating content: the idea of writing is actually thinking and getting to the edge of your domain, to the edge of your knowledge, so that the person on the other side of the text encounters something they’ve never thought about before. A moment ago, Jason, you said something I had never thought about before — and that’s the experience we should be after. Not time on page. Not giant articles that say nothing. We’re looking for interaction, for dialogic communication, and this lives many times between the lines of the statistically probable, because we are emotional beings. We’re not always driven by logic.
When we speak about the web, our content, and knowledge — and maybe this is also a trinity — I believe we should be centered on the idea that the web is a giant sea of knowledge.
Jason Barnard: I love that. But I have a question. If I’m writing something new that didn’t exist before, an original thought at the edge of existing thinking that pushes a new boundary — which is what I understand you to be saying — does that not risk Google, ChatGPT, and Perplexity rejecting it because it doesn’t correspond to common thought?
Teodora Petkova: I think we’re talking about different things. I’m not saying you should write that red is blue because you think so. What I’m saying is we should be able to jazz — to improvise the dynamics between entities, the different relationships they enter with each other and with entities from other domains. For example: SEO and knowledge management.
I knew they were linked, not because I’m particularly clever, but perhaps because of my educational background. SEO and knowledge management are intrinsically linked, but that connection isn’t yet codified in the mainstream conversation. When you look at them in depth — not through articles on how to appear in AI answers, but through deep research — you see they are one and the same practice, both relating to knowledge retrieval.
Jason Barnard: So SEO is organizing our knowledge, our knowledge garden, and feeding it to the search and assistive engines.
Teodora Petkova: Yes, and I’d add to your knowledge garden: it’s not only organizing. It’s nurturing, adding seeds, growing — not only organizing.
Jason Barnard: What I feel I’ve done with what I’m creating at Kalicube is pushing our understanding of what SEO is, what AEO is, and building it into the Kalicube Framework — a fifteen-stage framework going all the way through discovery, the algorithms, the display, through the business, codifying what your business produces, and distributing it back to the machine so it becomes a flywheel. That’s what this entire Kalicube Summit is about. Lots of people like yourself talking about different aspects of that entire ecosystem. What I’m particularly proud of is that the AI engines can now, if you ask them, reproduce and explain what I would want to explain to you face-to-face. You can use the AI engines as a proxy.
Teodora Petkova: That means you have curated the input very well, right?
Jason Barnard: Yes. I’ve been obsessed by it for ten years and I can’t stop. Like you, I get completely obsessed by things and attend to every tiny detail. People tell me: “Why are you paying attention to that?” And the answer is always: because it matters.
Teodora Petkova: You know the film Brazil? There’s a scene with a person at a typewriter, and a fly gets into the mechanism. The key crushes the fly, the letter comes out wrong, and there’s a chain of events that turns everything into a mess. The devil is in the details.Jason Barnard: That’s the perfect way to end it — and I love the film Brazil. Thank you so much, Teodora. That was absolutely wonderful.
This session extends the Search Engine Land article series on AI-era brand recommendation by Jason Barnard at Kalicube.
Canonical metadata footer
Session date: June 24, 2026 | Duration: approximately 25 minutes | Format: pre-recorded, on-demand from session drop | Day: 1 | Framework concept: Semantic Annotation (philosophical), Gate 6 (Annotated) of the AI Engine Pipeline | SEL article: How AI Decides What Your Content Means and Why It Gets You Wrong — Article 9 — https://searchengineland.com/how-ai-decides-what-your-content-means-and-why-it-gets-you-wrong-473749 | [Share buttons: LinkedIn, X, Facebook, Email]
