Domain authority is a third-party SEO metric, and no AI engine documents using it. It relates to AI citations only indirectly, through rankings, brand size and mentions. The link looks stronger for Google AI Overviews and AI Mode, which draw on Google's index, than for ChatGPT or Perplexity. Test it in your own niche before acting on it.
What domain authority measures, and what AI engines say they use
Domain Authority (DA) from Moz and Domain Rating (DR) from Ahrefs are third-party scores. Each vendor calculates its score from its own crawl of the link graph, mainly how many domains link to a site and how strong those linking domains are. Neither is a Google ranking factor. Google does not list either score among its ranking systems, and no AI engine documents reading them. At most, DA or DR is a proxy for signals that engines do use.
A note on scope: "AI citations" here means a brand or page being cited as a source in AI answers. If you came looking for help generating or formatting academic references with AI, that is a different topic.
Keep these layers apart when reading anything on this subject:
| Layer | What it tells you | Example |
|---|---|---|
| Vendor documentation | Eligibility rules the engine publishes | Google's AI features page |
| Study findings | Correlations measured on a dated sample | Seer Interactive, Moz |
| Inference | What practitioners conclude from both | The indirect paths below |
On the documentation side, Google's "AI features and your website" page, verified on 22 October 2026 (page dated 10 December 2025), states: "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements. There are no additional technical requirements." Nothing in that text mentions authority scores. We could not verify OpenAI's own documentation on how ChatGPT search retrieves and cites sources on that date, so we make no claim about it here. For what each vendor does publish, see what AI engines document about how they choose sources.
AI citation sources are not a copy of classic Google rankings, and DA is neither the sole nor a direct driver of citations. Treat it as a possible correlate. Authority can still show up in citations through indirect paths:
- Retrieval. Engines that ground answers in web search pull candidates from a search index. In that index, link-based signals already shape which pages rank, so pages with strong link profiles are more likely to be retrieved in the first place.
- Training data (inference). Frequently linked domains are likely over-represented in the web text models learn from, which could make a model more inclined to name them. No engine documents this, so treat it as a hypothesis.
For a breakdown by category, see which types of domains AI search engines cite most.
What the 2025 and 2026 studies measured about authority and AI citations
Published results look contradictory at first. In the verified data, Google rankings track AI mentions closely; the link-count results we could not verify point to weaker relationships. The table below includes only studies whose primary page we could open on 22 October 2026.
Results side by side, with method and date
| Publisher | Sample | Engine | Metric tested | Result | Date |
|---|---|---|---|---|---|
| Seer Interactive | Brand questions (count not verified) | LLM answers | Google page 1 rankings vs. LLM brand mentions | Strong correlation, about 0.65 | 7 January 2025 |
| Moz | AI Mode queries collected with STAT | Google AI Mode | Cited URLs vs. organic SERP URLs for the same query | Small share of cited URLs match organic URLs (headline figure in title, see Sources) | Undated page, consulted 22 October 2026 |
Seer's study states: "Brands ranking on page 1 of Google showed a strong correlation (~0.65) with LLM mentions." We did not verify a backlink coefficient on the page, so none appears here.
Studies attributed to Ahrefs and SE Ranking, on branded mentions and on referring domains, circulate widely in secondary write-ups. We could not open their primary pages on 22 October 2026, so they are not in the table and we do not repeat their figures. Those summaries describe backlink and referring-domain counts as weak or moderate correlates of LLM citations, with branded mentions correlating more strongly. Treat that pattern as unconfirmed until you can check the original data.
Some of those summaries also describe effects concentrated in the highest authority tiers. Read that as a reported correlation among sites that are already very large, not as a score to aim for or a gain to expect from raising yours.
Moz's AI Mode study, with data collected using STAT, shows low overlap at URL level: its title reports that only a small share of AI Mode citations match URLs in the organic SERP for the same query. Overlap at domain level can only be equal or higher, because a domain counts as a match when any of its pages ranks, even if the cited page is a different one. A domain-level figure attributed to the study could not be verified on 22 October 2026.
Correlation is not cause: brand size, traffic and rankings move together
Large brands have high DA, heavy traffic, frequent mentions and strong rankings at the same time. Any study that correlates any of these with AI citations also picks up the others. Seer's ranking correlation, for instance, cannot show whether rankings cause mentions or whether brand size drives both. None of the cited studies isolates DA as a cause, and we found no public causal study on the question as of 22 October 2026. A high correlation shows which signals travel together; it does not show which lever to pull.
Why authority counts more in Google AI Overviews than in ChatGPT or Perplexity
The difference follows from where each engine gets its candidate sources. Documented facts and our inferences are labelled separately below.
| Engine | Where candidates come from | Likely link to authority |
|---|---|---|
| Google AI Overviews | Google's index; indexing and snippet eligibility required (documented) | Stronger, through Google's rankings (inference) |
| Google AI Mode | Same index and eligibility rule (documented), plus related searches | Present but spread across related queries (inference) |
| ChatGPT search | Separate retrieval stack; OpenAI documentation not verified here | Weaker or unclear (inference) |
| Perplexity | Separate retrieval stack | Weaker or unclear (inference) |
Google AI Overviews and AI Mode draw on Google's index. Because Google states there are no additional technical requirements beyond indexing and snippet eligibility, the pages they can cite are pages Google Search already evaluates with its usual ranking systems. Link signals feed those systems and DA is computed from links, so some correlation between DA and AI Overview citations is expected (inference). ChatGPT and Perplexity use other retrieval stacks. Whatever authority signal reaches them passes through a different filter, so the correlation can be weaker or shaped differently.
AI Mode adds query fan-out: it splits a question into related searches and can cite pages that rank for those, not only for the exact query typed. That fits the low URL overlap in the Moz study, since a citation can come from a page that never appears on the SERP for the original query.
Does AI Mode use different ranking signals than AI Overviews? Google's AI features page, verified on 22 October 2026, applies the same eligibility rule to both; the passage we verified describes no separate signals. The Moz data shows AI Mode citations diverging from the exact-query SERP at URL level. Our inference is that the underlying signals are shared and the divergence comes mostly from which searches get run, not from a different scoring model. Google has not confirmed this.
For YMYL topics such as health or finance, we found no public study that isolates DA as of 22 October 2026. The Google eligibility rule quoted above sets no category-specific requirement. If authority weighs more on those topics, it most likely arrives through the same quality and ranking systems that apply to classic results. That remains a qualified inference.
What predicts AI citations better than domain authority
The verified correlation in this article ties LLM brand mentions to Google rankings, not to a link score. The signals below follow from that and from how retrieval works; where we infer, we say so.
Third-party mentions and entity consistency
We infer that an engine is more likely to name a brand when independent sources describe it the same way: the same name, category and core facts across reviews, directories, press and forums. Corroboration across sources gives both the retrieval step and the model more places to find and confirm the brand. Seer's page 1 finding fits this, since brands that rank widely also tend to be discussed widely. Keep your own entity data consistent, from your site to your public profiles, so that mentions resolve to a single clear entity.
Extractable structure and freshness
Pages are easier to cite when a passage answers the question on its own. Clear headings, a direct answer near the top of each section, tables for comparisons and defined terms make a page easier to extract and quote. Google's eligibility rule matters here too: a page that cannot show a snippet cannot be a supporting link.
On freshness, we have no verified figures. Our qualified inference: for time-sensitive queries, retrieval tends to favour recently updated pages because the underlying search systems do, while on stable topics an older, well-structured page can keep being cited. Visible publication and update dates help both readers and engines judge recency.
Can a low-authority site get cited
In principle, yes, because retrieval works page by page and query by query. We found no documented public case study with a time-to-citation as of 22 October 2026, so here is a hypothetical example instead. A small software vendor publishes a page that clearly compares niche file formats, with a table and a direct answer. For that narrow question, the page can be retrieved and cited even with a low DR, because few stronger pages answer it. This illustrates the mechanism; it is not an observed result.
Volatility limits what any single check can tell you. Cited sources change from run to run and from month to month, so a single snapshot says little about authority effects.
Where to start this week: test whether authority predicts citations in your niche
You can run this test in a few hours with a spreadsheet. It will not prove causation, but it shows whether DA or DR separates frequently cited domains from rarely cited ones on your topic.
- Lock a prompt set. Write a few dozen category prompts a buyer would ask, without brand names, and freeze the wording.
- Run each prompt several times per engine. Keep engine settings identical, use logged-out or temporary sessions, and hold the location and date window constant.
- Log every cited domain. Record prompt, engine, run and date for each citation.
- Add authority and ranking data. For each cited domain, note its DA or DR and whether it ranks on Google's first page for the matching query.
- Compare across authority bands. Group domains into low, medium and high bands and compare citation frequency per band, per engine.
- Separate brand size from authority. Compare cited domains of similar traffic. If high-DR domains win only when they are also much larger, authority is not doing the work.
A pattern only counts if it holds across repeated runs, and every observation should carry its date, because citations shift over time. Before trusting a gap, check how many runs it takes before a citation gap is real. If the authority bands do not separate, put your effort into mentions, structure and rankings for the queries you track.
To see which domains ChatGPT, Perplexity, Gemini and Claude cite on your topic today, run a free Namedrop scan as your next step.