Forget Domain Authority, Think Brand Authority: What I Said in 2017 and Why It Took Nine Years to Land
Nine years ago I stood in a room in Metz and told a hall full of French SEOs that the number they checked every morning was measuring something Google had never built, and the room was courteous about it, and almost nobody changed what they did on the Monday. I’ve made the same argument every year since. It took a technology that strips the domain out of the answer entirely before it finally landed.
In November 2017 I told a room full of French SEOs in Metz that Google was measuring their brands and ignoring their domains
The talk had a title I still like: Oubliez Autorité Domaine, pensez Autorité Marque. Forget domain authority, think brand authority.
It was the third SEO Camp Day Lorraine, at TCRM Blida in Metz on 24 November 2017, Olivier Andrieu was on the bill, Erlé Alberton had crawl budget and Nicolas Creanico had links, and I stood up and said the metric most of that room used every day was pointed at something Google has never had.
The agency write-up afterwards described the talk as a full year of research into brand authority inside Google’s algorithm, arguing that domain authority barely exists within it. That was Digital Position, in the room, writing it up themselves, and the same piece sent readers to Kalicube®.pro for the tools I’d already built to measure brand authority. The archived homepage from January 2018 still shows them: brand authority defined as the confidence Google has in its understanding of a brand, and an algorithm tracking twenty-five thousand of them.
Ten weeks later I made the same argument in English, on Semrush in February 2018, and signed off by promising a piece on why you should be obsessed with brand authority and forget domain authority. It has taken me rather longer than I implied.
I lead with all of that because the difference between a claim and a corroborated claim is the whole subject of my working life.
Domain authority was never a Google metric, and everybody in the industry knew it
This still surprises people who weren’t paying attention at the time.
A tool vendor invented domain authority. It was a useful proxy, built by clever people, designed to approximate something Google was doing that nobody could see directly, and Google never confirmed it, never published it, and said repeatedly that they held no such score. The industry took it up anyway, because a number you can watch beats an uncertainty you can’t.
That was the trade everyone made, and then over about a decade we forgot we’d made it. Agencies reported on it, clients asked about it, link buyers priced against it, and an entire economy grew on top of a proxy that the company it was proxying had disowned.
My argument in Metz went at the object rather than the arithmetic: the number was calculated perfectly well, and it was pointed at the wrong thing.
Google trusts entities rather than domains, so a new site with a clear brand beats an old site with a vague one
Here’s the thing I’d spent 2017 working out.
Google asks whether it understands what a thing is, and whether it’s confident enough in that understanding to put the thing in front of a user, and both of those are questions about an entity: a person, a company, a product, a brand. A domain is a container the entity happens to use, so the container gets measured while the contents get evaluated.
Once you see it that way a lot of confusing behaviour resolves. A brand new domain with a well-understood entity behind it outperforms an old domain attached to something Google can’t pin down, and two sites with identical link profiles behave differently because one carries a brand the algorithm understands and the other is a stranger. Entity confidence explains that. Domain authority never could.
I’d been circling it for three years by then, without the vocabulary to say it cleanly at the start.
I built the argument in public across three years and three conferences, which is why you can check it
In January 2015 I gave my first ever SEO talk, at the same Metz event, on web reputation as the foundation of search rankings. Journal du Net covered the day and reported that I’d demonstrated how web reputation impacts search rankings, and what I was arguing, clumsily, is that the sum of what the web says about you is what Google works from.
In October 2016 in Nancy I widened it to brand experience across every digital touchpoint, and SEO Camp’s own report of the day has me focused on the e-reputation of brands.
In 2017 I put it together, first in Lyon, arguing that Google is a child and that we ought to be educating it rather than gaming it, then in Metz with the conclusion those three years had been building towards.
Reputation, then experience, then education, then authority. I’d love to tell you I planned that sequence, and the truth is I was following a thread and reporting what I found each year.
I can lay it out now because other people wrote it down at the time, and that’s the mechanism rather than the luck. It’s what I’ve spent the years since building a company around, and the full record of who published what, and when, sits on its own page.
Nine years passed before the argument became obvious, because AI removed the domain from the equation entirely
If you want to know why brand authority is suddenly everywhere in 2026, the answer is structural rather than intellectual.
When a language model answers a question there’s no domain in the response, no blue link, no URL for anyone to evaluate, no site whose strength you could assess. There’s an entity, and there’s whether the model is confident enough about that entity to name it.
Domain authority didn’t lose the argument. It lost its subject matter.
Everything an AI system does when it decides whether to recommend you comes back to the question I was asking in Metz: does it understand what you are, and is it confident enough in that understanding to say so out loud? A model holds an entity and a level of confidence about that entity, assembled from everything it has ever read about you.
Brand authority is independent sources agreeing about you, which is the one thing you cannot buy
So what is brand authority, operationally, once you stop looking for it in a tool?
It’s the degree to which independent sources agree about what you are, which is a question of how many independent people say the same consistent thing about you in places the algorithms already trust, rather than a count of who mentions you at all.
Three properties follow, and they matter more than any number.
Independence comes first: your own site is a claim, your partners are corroboration, and editorial, academic and analyst coverage is proof, which you can cause but can’t author. Consistency does the quiet work underneath it, because conflicting information suppresses confidence instead of averaging out, and a machine reading two founding dates for one company gets less sure about everything attached to that entity rather than picking the likelier date, which is exactly why a cheap model drops contradictions instead of resolving them. And legibility decides whether any of it counts, because corroboration a machine can’t parse is corroboration that never arrives, and the three machines reading you each want it in a different form. I had seven years of my own corroboration sitting in a book that nothing could crawl, which counted for exactly nothing until I put it on the web.
None of that is buyable, which is precisely why the industry preferred a metric that was.
Measuring domain authority for a decade cost the industry more than measuring nothing would have
There’s a smug version of this argument, and it’s the boring one.
Domain authority cost the industry precisely because it was actionable: it gave people something to optimise, so they optimised it, and the optimising built a trade in link buying, domain flipping and moving a number that the company it described had said it never kept.
Every hour spent there was an hour spent away from the thing actually being evaluated. A decade of effort, pointed at a proxy.
If you want one practical takeaway it’s the one I gave that room in Metz and have given every room since. Stop asking how strong your domain is, start asking whether the machines understand what you are, and whether enough credible independent sources agree with you that they can be confident about it.
That was the answer when Google was deciding which blue links to show. It’s the same answer now that an AI is deciding whether to say your name at all.
- Talk title anchored to the Webagogie programme listing, which carries the title.
- The Digital Position write-up anchors the year-of-research and domain-authority claim, because that is what the page says on its face. Independent, dated 24 November 2017. This is the anchor citation for the whole piece.
- Journal du Net anchors only the 2015 demonstration, which is what it reports.
- SEO Camp’s 2016 report anchors brand e-reputation, nothing wider.
- The two SlideShare decks are your own content on a third-party platform. They prove dating and existence, not independent endorsement, so they carry deck titles and no interpretive weight.
Gap to close: the Wayback snapshot of kalicube.pro showing brand authority on the homepage. Worth having, though the Digital Position citation already proves the same fact from an independent source, which is stronger.
Ready to hand to Writing OS for final voice and structure. Run voice-check before the copy-editor.