Your Domain Authority Score Doesn’t Decide Who AI Cites. Here’s What Does.

Domain authority does not decide which sites AI engines cite. In a Featured audit dataset of 22,881 AI citations, 34.5% pointed to domains scoring under DA 40, and 76% of the cited domains showed up exactly once. If you’ve been telling yourself that AI citations are reserved for big brands with big link profiles, the numbers say otherwise.

That doesn’t mean authority is irrelevant, and this article won’t pretend it is. What it does mean is that a lot of mid-size companies are using a weak link profile as a reason not to try, and the data doesn’t back that excuse. Below we walk through what two recent studies measured, where each one has limits, what seems to predict a citation instead, and a short plan you can start this month. It sits under our SEO work because the fix is mostly an SEO fix with a different target.

Does domain authority decide who AI cites?

Not on its own. Surfer analyzed about 5 million unique citation URLs across 20,000 prompts over three months, covering Google’s AI Mode, AI Overviews, ChatGPT and Perplexity. It tested three authority measures: PageRank, Harmonic Centrality, and Surfer’s own Domain Score. The Spearman correlations with how often a domain got cited came out at −0.065, +0.010 and −0.051. Surfer’s own summary is that authority “correlates negatively with AI citation frequency, but barely,” close enough to zero to be noise.

One detail is worth reading twice. When Surfer removed the top 5% strongest domains, the slight negative correlations mostly disappeared. Their guess is that engines may be spreading citations around on purpose, to avoid leaning on a handful of giant publishers. That’s their hypothesis, and the study can’t confirm the motive. But the pattern fits what Featured found in a completely different dataset.

Can a small or low-authority site really get cited by AI?

Yes, and it happens more often than most teams assume. Featured ran 405 GEO audits between June 2 and June 21, 2026, covering 378 brands across 344 industries and 1,983 buying questions. Those audits produced 22,881 citations from 11,499 distinct domains, and Featured then looked up each domain’s Moz Domain Authority on August 21, 2026.

Two more numbers from the same report explain why a small domain has a shot. About 76% of the cited domains were cited exactly once, and the 1,000 most-cited domains together accounted for 45.1% of citations. So more than half of all citations went to a long tail of sites that each got a small handful.

Now the honest part. The median DA across distinct cited domains was 37, but the median across individual citations was 56. Higher-authority sites still collect more citations per site. The claim here isn’t that authority is worthless. It’s that a low score doesn’t lock you out.

If authority isn’t the problem, why does your brand never show up?

Because most of what AI engines cite isn’t the brand’s own site. Featured classified 10,176 of its citations by source type. Editorial content made up 64.5%, competitor domains 22.6%, reference sources 8.6%, and community sources 2.6%. Brand-owned content was 1.7%.

Put those next to the headline finding and the picture gets uncomfortable. Featured reports that 78.5% of the brands it audited never appeared in AI answers about their category, and 92.6% never showed up as a citation at all. When they did get cited, competitor pages were cited about 13 times as often as the brand’s own content. For most of these companies the gap wasn’t a DA gap. It was that the pages engines pulled came from somewhere else.

What predicts an AI citation if it isn’t a link score?

The two best-documented answers are original information and being known by name. Surfer’s read of its 5 million citations is that engines favor content that is anchored in established consensus while still contributing something new, meaning original research, proprietary data and real expert perspective. We cover that side in our piece on original research, and it lines up with how we’ve seen citable pages behave.

The second answer is brand recognition. The 5W Research report from May 2026, which pulls together several large citation datasets, states that brand search volume is the strongest known predictor of citation likelihood, at a 0.334 correlation, and stronger than backlinks. That figure is 5W’s compilation and we haven’t re-derived it, so treat it as a directional signal. It also matches something we’ve argued before: engines have to decide who to believe, and familiarity is part of that. Our post on mentions versus citations breaks down how the two scoreboards differ.

Here’s our own read, and it is interpretation, not a measured result. A link score tells a search engine how a page was linked to. An answer engine has a different job: find a passage that answers this exact question and that other sources won’t contradict. A small site with a clear, specific answer can do that job as well as a big one.

Does that mean you should stop building links?

No. Both studies have limits that argue against tossing your link program. Featured measured DA after the fact, on Perplexity results, from audits of brands that chose to hire an auditor. Surfer’s authority scores are PageRank, Harmonic Centrality and its own Domain Score, and DA from Moz is a different metric. Neither one proves that links have no effect. They show that a strong link profile isn’t sufficient, and that a weak one isn’t disqualifying.

Links still do their normal work in classic search, and Google’s AI features draw on that same index. What changes is the priority. If you have a limited budget, the data says the marginal dollar is better spent on being the most specific, best-supported answer to a question your buyers ask than on chasing another point of DA. Our piece on topical authority covers how depth on a subject builds the kind of authority these engines seem to care about.

What should you do about it this quarter?

Run a short test before you change strategy. Here’s the order we’d use.

1. Write down 20 buying questions. Use the questions your sales team hears, not your keyword list.
2. Ask each one in ChatGPT, Perplexity and Google’s AI Mode, and log every cited domain. Note the source type, and look up each domain’s authority score in whatever tool you already use.
3. Find the low-authority winners. Any small site that gets cited for your category proves the door is open. Read its page and note what it does that yours doesn’t: a specific number, a definition, a comparison table, a named author.
4. Check where editorial sources are doing the work. If most citations go to publications and review sites, your job includes earning coverage there, not only publishing on your own domain.
5. Fix or build one page per gap. Lead with a direct answer, add one piece of original data you can source, and put a real expert’s name on it.
6. Re-run the same 20 questions in 30 days. AI answers vary run to run, so repeat each question a few times and look at the pattern, not one screenshot.

That’s essentially a compact version of the work in our AI visibility audits, and it costs you a few hours rather than a link-building retainer.