Elevator Talk: Does Domain Authority Even Matter for AI Visibility?

I agree with everyone's take, especially Alyssa's point that studies like these shouldn't be taken as concrete evidence.

The way this was structured, I'd view it as: don't over-prioritize DA/DR-style metrics when evaluating AI visibility opportunities, and focus more on content depth, topical fit, clear answers, and original data — which I believe we've already shifted toward in both AI and SEO.

The main issue I had with this study was that it kept validating its results off of citation frequency rather than citation quality, which I think would have shown more impact. I'd be more curious to see a study on that.

I love the point about quality!

Retweet, Alondra.

I don't think DA/DR has been the end-all, be-all metric for your ability to rank in a SERP for a long time. The fact that there's a mix of high- and low-DR domains being cited by LLMs shows that domain authority as a concept is matching traditional search. The more important ranking factor is obviously freshness of content,  but having a strong backlink profile is going to get you crawled more often and noticed more quickly.

What I'd be interested to see is whether those low-DA/DR domains had any sort of broad reach like the bigger sites do. My guess would be that they do not.

Uh oh. We've stumbled across one of my biggest pet peeves. All sorts of studies are coming out with some version of "we did this, this happened, and therefore we KNOW this." But guys, these are observational studies. You cannot determine causation from them. Is it a correlation? Definitely. But not causation.

A glaring issue with this study is that they didn't categorize prompts. Right now LLMs are pretty RAG-based, which really prioritizes semantic relevance. For an everyday kind of query like "best sci-fi movies," an AI will happily cite a low-DA niche blog if the content fits the prompt well enough and be on its merry way. But YMYL topics like healthcare, finance, or legal, are a completely different story. Models are supposed to apply much heavier guardrails there, and that's not something I saw them considering here.

I also think they're getting confused about the purpose of DA. DA isn't a factor Google or LLMs use. It's not like they go check Ahrefs for a DA/DR score every time they rank or cite a page. But they are absolutely looking at the same web landscape (i.e. entity associations, brand mentions) that creates that score. DA is just a score created by SEOs to act as our best guess at how all of that creates authority. Of course it's not going to determine whether a page gets ranked or cited.

One thing that stood out to me: over and over they said the correlation is "basically zero." Then they removed top-cited sources with high DA, and the correlation got closer to zero. Then they looked at only blogs, and the results varied quite a bit from platform to platform. And then it was "basically zero" again. By the end of the article, I felt like the result of the entire study was: "we don't think this matters, but who's to say, since there isn't really much correlation with anything related to DA and citation frequency." 😂

That said — this does support what Blake and I have said for the last couple years about backlinks not being that important and DA not being a good measure of organic or AI success. More and more I'm seeing SERPs where you've got sites with a DA of 80–99 and sites with a DA of 10–50 performing side by side, sometimes with the lowest-DA sites ranking the highest. More and more it's becoming about content: relevance to the user's query and unique takes on the topic at hand.

From what I've observed, semantic and relevant content that is easy for the AI models to retrieve is what gets cited the most. I've seen super high-authority content on large websites not be cited at all but when I dig down, I'm seeing that it isn't semantically relevant for the content the queries are prioritizing.

I think we're going to see LLMs move away from just pure RAG and semantic matching for responses,  which they kinda already are. I don't know if anyone has been following this court case, but the EU is holding Google liable for misinformation in AI Overviews. I don't know if it'll stick, but it's likely going to mean more guardrails in responses. I think authority is going to be a big factor in that, and I'm guessing LLMs will have their own internal way of determining authority. It'll probably not be purely an SEO metric like DA — but until we know more (if ever), I think it's still our best guess.