Google Decides the Dominant Interpretation of Your Name Three Times, Not Once
And the three verdicts disagree far more often than anybody bothers to check.
A client showed me a Knowledge Panel he was proud of, and he had every right to be, because it was complete, accurate, and sitting exactly where he wanted it. Then he asked ChatGPT who he was, and it described a man in another country with the same name. Nothing was broken. Both systems were working perfectly, they’d simply reached different verdicts about which person the name belongs to, and he’d never thought to check whether they agreed.
Google’s own vocabulary already splits a name into dominant, common and minor
Google supplies the vocabulary here, which is exactly why it carries weight. In the search quality rater guidelines, Google tells its raters that a query can carry more than one interpretation, and it ranks them: the dominant interpretation is what most people mean when they type those words, a common interpretation is what a meaningful minority mean, and a minor interpretation is everything else. Raters are told to judge results against that hierarchy, which makes that hierarchy an instruction about what Google intends to serve rather than a note on what it happened to find.
Apply that to a personal name and the stakes get very concrete very quickly. If two people share a name, one of them is the dominant interpretation and the other is minor, and there’s no version of the world where both win. Somebody is the Scott Duffy, somebody else is a Scott Duffy.
The same contest runs in the Knowledge Graph and inside the model, on separate evidence
Here’s the part I haven’t seen anybody write down. That contest runs three times over, because Google runs three systems, and I named them the Algorithmic Trinity back in 2024 for exactly this reason. The search engine decides which interpretation gets the results page. The Knowledge Graph decides which entity gets the panel. The large language model decides which person the name pulls up when it answers off the top of its head.
Three decisions, three bodies of evidence, three verdicts that are free to disagree and routinely do.
Each of the three systems resolves a name by a completely different mechanic
The Knowledge Graph picks one, the search engine distributes, the model weights
The Knowledge Graph is close to binary, because a panel triggers for a single entity, so dominance there is a discrete win: your ID gets the panel and the other person gets “See Results About”, or the reverse. The search engine is a distribution, so dominance is a proportion of the page rather than a switch, and you can hold eight results out of ten and still be sharing your own name in public. The model is neither, because there’s no ID and no ranking inside it, only association strength, so dominance there means the name pulls your attributes harder than anybody else’s, and it degrades quietly rather than flipping.
Three mechanics means three different kinds of failure: you can lose the panel outright, you can bleed the page one result at a time, or you can watch the model hedge, which looks like nothing at all until the day it costs you a deal.
The three verdicts run on different clocks, which is why they drift apart
The search engine re-decides continuously, in real time, on every query. The Knowledge Graph re-decides on update cycles. The model only re-decides when it retrains, or when a grounding pass overrides what it already believes.
For me this is the whole of it. Three systems reading overlapping evidence on three different clocks will fall out of step, always, and the drift is invisible from the inside because you can’t see when your content was last recruited into a training run or a graph update. Fix your website today and the search results can move this week, the panel may move this quarter, and the model may not move until it is rebuilt. Anyone who tells you they have “fixed your AI presence” in a fortnight is describing a clock that doesn’t exist.
A flawless Knowledge Panel beside an AI that names a stranger is the standard failure, not a freak one
Most brands have never audited this, so they measure one system, usually whichever one flatters them most, and assume the other two agree. They don’t. The common shape is a business that has done real work on its website and its search results, holds the page comfortably, and is still a minor interpretation inside the model, because the model was trained before the work landed and nothing has forced it to reconsider.
Your buyer doesn’t experience three systems. They ask a machine who you are, they glance at the page when they are not sure, and they check the panel when it matters, and any one of those three moments can be the one that decides it. Two out of three is a coin flip on the moment you don’t control.
Scott Duffy lost the dominant interpretation once and took it back without doing anything new
Scott Duffy came to Kalicube in 2021 as the minor interpretation of his own name, because the other Scott Duffy teaches Microsoft Azure to a million and a half students out of Toronto and had the panel, the page and the machine answer between them. We moved all three, and then held them.
A while later a Google update landed while the trainer’s student numbers were climbing hard, and for a short spell he was back on top. Scott did very little additional online while we took the position back, because the recapture came from corroboration rather than activity: more sources agreeing, more consistently, for longer. Dominance is a position you hold, and holding it is the work.
You audit this by asking all three systems the same question and comparing the answers
Ask the model who you are, cold, in a clean session. Search your own name and count how much of the page is actually yours. Open the panel and read it as a stranger would. Then put the three answers side by side, and the gaps between them are your work order, in priority order, starting with whichever system your buyers reach first.
That comparison is the diagnostic that sits underneath everything we do at Kalicube, and it’s the reason The Kalicube Process™ starts at the bottom of the funnel: understanding is what all three systems are arguing about, and until they agree who you are, nothing built on top of that argument is safe.
Dominance in one system is not dominance, it’s a third of your own name.
Status: Original concept, first publication. Published 17 August 2026 on jasonbarnard.com.
Publication note: Dominant interpretation, common interpretation and minor interpretation are Google’s terms, defined in Google’s search quality rater guidelines for search results. The extension of that hierarchy to all three components of the Algorithmic Trinity, the observation that the three verdicts are reached independently on different evidence and different clocks, and the resulting audit method, are mine, published here first on 17 August 2026. The Algorithmic Trinity was coined by Jason Barnard in 2024.