From String to Entity. How I Made AI Recognize Who I Am

This post expands on my talk from I Love Marketing & Technology 2026, one of Poland’s largest marketing conferences, on building a personal brand in the era of generative search and language models.

You’ll learn, among other things:

• what an entity is in the context of a personal brand,
• how Google’s Knowledge Graph and LLMs identify people and brands,
• how to check your own resultScore in the Knowledge Graph API,
• what Trust Halo is and how the authority of an entity transfers to new projects,
• what Google’s patents (US12321706B2, US20180046717A1) and the Google Content Warehouse API leak mean for your personal brand.

You’ll also find supplementary materials for attendees here, including an 8-step checklist for building an entity, an interactive resultScore calculator, the source patents, and links for going deeper.

Your personal brand in AI. When the systems don’t know who you are…

Not long ago, in the eyes of AI, I was at best a more or less random string of characters. Which means I could have been one person, I could have been five, I could have been a million.

Entrepreneur.
SEO specialist.
Keto blogger.
Author.
Travel blogger.
And now AI, too…
and we’ll probably add something else next week 😉

Some will call it versatility. Others, fragmentation. A third group will deliver an internet-diagnosed case of ADHD. And still others will laugh that I have to turn everything into a business. And you know what? Every one of them is right.

The problem is that, to the algorithms, this is chaos. And chaos doesn’t become a brand.

Jeff Bezos once said:

“Your brand is what other people say about you when you’re not in the room.”

You know this one well, don’t you? And Bezos is undeniably a smart man – but what if I told you that today his line should read:

“Your brand is what the systems can reconstruct about you, without your help.”

The catch is this: if the system doesn’t know who you are, you don’t have a brand.

Building a brand isn’t overproducing content

How have you pictured brand-building online so far? You’re out there on the internet, making content in the broadest sense, right? A few years ago, that’s what I thought too. But I have bad news: that isn’t building a brand. That’s overproducing content. And to understand how to build an online identity instead of overproducing, you need to grasp one more thing.

The answer to the question of your identity is… the ENTITY of your personal brand

Do you know what connects:

• an IKEA bookcase,
• Coca-Cola,
• me,
• the Doge meme dog,
• and I❤️Marketing?

Each of them is an entity. A recognizable, distinct object to which the system assigns attributes, relationships, and context.

The IKEA bookcase is the Billy model. When you search for Billy, Google won’t first offer you Billy Idol – it’ll offer you that IKEA bookcase. Don’t believe me? Check it.

LLMs behave the same way: ask about Billy without setting any prior context, and they’ll explain that you probably mean the iconic bookcase.

An entity is something a system can unambiguously recognize. A person, a brand, a thing. Kabosu the dog. You’re an entity too. Or at least you should be. And how to be one is exactly what we’re going to talk about now.

Rule 1: The system doesn’t remember what you do. It remembers what you appear alongside.

The first question you should ask yourself is no longer “am I visible in Google.” It’s: does the system even know who I am?

Two years ago, my answer (translated from LLM into human) was: not really…

I’d been on the internet for years. More than fifteen of them. I wrote content, recorded courses, podcasts, webinars. I built companies. I wrote books.

And yet the algorithms saw my name in very mixed company: other SEO specialists, cookbook authors, bloggers, influencers, conferences, and assorted brands.

And unfortunately, the systems didn’t know whether this was one and the same person or a coincidental clash of names.

To them, “Ewelina Podrez-Siama” was a sequence of characters – in system language, a string. And a string is not an entity.

A bleak picture after building a career for over 15 years, isn’t it?

To fix it, I did a lot of things. But one was the breakthrough: Wikidata.

Evidence, not declarations

Wikidata is a database of facts that systems understand: who did what, and what they’re connected to. To put it loosely, it’s a kind of Wikipedia in the language of AI.

I added my profession there. Publishers. Business partners. Books with their ISBNs. Awards. In other words, external, independent evidence of my existence – not my own declarations.

From a string of characters with no connections, I presented the systems with… coherence and a legible graph of my relationships.

I didn’t change the content. I didn’t produce anything new. I only changed how the systems see me.

I could now go on about how my results in AI Overviews shifted within days. How successive post-training versions of LLMs started recommending me and citing my site. How my travel blog, launched in mid-February, racked up 4,000 clicks from search in a single month – a result of the trust transferring to the entity. I call this effect Trust Halo.

But let’s slow down. AI’s favor is a fickle thing. So instead of dwelling on isolated results, I’d rather focus on the certainties and the facts.

Wikidata isn’t a whim

Systems have used Wikidata for years. What’s more, Google patented a way to add knowledge to an AI model without retraining it (patent US12321706B2, granted in June 2025). The data source in the experiments was… Wikidata. More than 23 million knowledge triples describing 1.1 million entities.

If you take one thing away from this post:

Gather and organize external evidence of your existence, and try to get into Wikidata.

Wikidata isn’t Wikipedia. The barrier to entry is lower, and indexing by Google and LLMs is direct. Didn’t work the first time? Read the editors’ notes and keep trying until it does.

Read also: Wikidata step by step – how to get into the database Google trains language models on.

Rule 2: The system isn’t looking for keywords. It’s looking for an answer: who are you?

The system’s certainty about your identity is measurable. Picture a map. Not a road map – a map of meaning. The points: people, companies, places, concepts. Between them, connections:

• …works at…,
• …is the author of…,
• …specializes in…

The more coherent, confirmed connections there are, the more certain the system is that it knows who it’s dealing with. And the stronger your point on the map of meaning.

This isn’t just theory – Google really does build such a map. It’s called the Knowledge Graph. And you can check how confident the system is that it knows who you are.

resultScore – your score on the map of meaning

My score before organizing the entity was 12. I wasn’t a point on the map of meaning. After the work with Wikidata, the first jump – to 43 – happened within 11 hours. The next, to 97.4, within a few weeks. Since then it fluctuates between those two good scores (which is natural in the “life cycle” of an entity).

You might be wondering now whether 97 is a lot. It’s enough for a knowledge panel to display for my entity in Google. And for me to become, in the systems’ eyes, a single person. Finally.

That said, with resultScore the sky is the limit. Taylor Swift, for example, has a score of over 24,000.

The question for you is this:

Is your personal brand a point on that map of meaning? Are you the Taylor Swift of your industry, or more of a blank page?

The answer is your resultScore in Google’s Knowledge Graph. Below you’ll find a guide on how to check it, and a calculator to interpret the result.

And what about LLMs?

LLMs work differently, but the consequence is similar: if they have a coherent picture of you, they’ll reconstruct it. If they don’t, they’ll start guessing, hallucinating, or – worse – mistaking you for someone else.

So use the same definition of yourself everywhere – give the systems a legible pattern. Don’t call yourself a “marketing expert” one place, a “digital strategist” another, and an “AI ninja” a third. Update your messaging consistently, across every channel. And where you’re not sure an update will be possible later, avoid vague phrasing – even if it means writing that you’ve worked in SEO since 2009 rather than “for 17 years.” Don’t repeat my mistakes 😉

And once you’ve organized your messaging, remember that…

Rule 3: The system doesn’t believe what you say about yourself. It believes what others can confirm about you.

Let’s say you have a website. On it, you’ve written that you:

• have X years of experience,
• have worked with the biggest brands,
• are an expert in your field.

Well done! So has half the internet.

Paradoxically, that generic messaging isn’t actually the biggest problem. Because what you put on your own website is the least credible signal for the algorithm. Precisely because, in the end, anyone can write whatever they please about themselves.

In the language of search, this is described outright as a conflict of interest – something Search Quality Raters, the people Google employs to assess websites and whose verdicts then help train the algorithms, are highly sensitive to.

So you can write that you’re an expert, that you have a young, dynamic team and free-snack Fridays. The internet will still reply: “we’ll see about that.”

What is credible to the systems?

• media mentions,
• citations,
• Wikidata with external references,
• links from authoritative domains,
• ISBNs in a publishing database,
• what others say about you.

Connections to strong, recognizable brands build your credibility too, because every fact in Wikidata is not only evidence but also a potential knowledge triple in the system:

subject — predicate — object

Which, in my case, means:

Ewelina Podrez-Siama → is the author of → the book Marka osobista w czasach AI i generatywnego wyszukiwania

And a language model can then recall it. Without context. Without my social media. Without my website. Without my involvement.

So:

Gather every mention of yourself and check whether they build a coherent picture.

If they don’t, start working on them.

What does building a brand really look like?

First you’re on the internet making content – you know that part already. Then the systems start to identify your presence. It’s that building that turns you into an entity. You appear in the knowledge graph. Roughly at that stage, you become a brand. And from that moment, the systems expand your identity.

Steps 1–2 almost everyone does. Steps 3–6 hardly anyone does deliberately. And that’s exactly where the algorithm decides whether you’re someone specific (an entity) or a more or less random string of characters.

Thank you!

Thanks, to start, to everyone who stayed after the talk, sent messages, caught me in the hallways – it meant a great deal to me.

And thanks to the whole organizing team for the invitation to that demanding stage, where I got to share my perspective on the world of algorithms.

For those who weren’t at the talk, I invite you to the supplementary materials below.

Supplementary materials