Topical Authority and Domain Authority: One Is a Strategy, One Is a Scoreboard

These two terms get spoken in the same breath so often that business owners understandably assume they are two flavours of the same thing. They are not even the same category of thing. One is a number a software company calculates about your website. The other is a property your website either has or lacks, visible in how completely it answers the questions inside a subject. Confusing them is expensive, because the two lead to opposite spending decisions: chasing the score pushes budget toward link acquisition, while earning the property pushes budget toward research and writing. This page separates them properly, shows how each is measured and how each gets gamed, and closes with a direct answer on which one a small Canadian business should pursue.

Domain authority is a vendor's prediction, not a search signal

Domain authority (DA) is a metric created by Moz, a Seattle SEO software company, in the mid-2000s. It scores a domain from 1 to 100 based mostly on the quantity and quality of links pointing at it, on a logarithmic scale where climbing from 20 to 30 is far easier than climbing from 70 to 80. Ahrefs sells a near-equivalent called Domain Rating, Semrush sells Authority Score, and the three numbers routinely disagree about the same website because each vendor crawls a different slice of the web and weights it differently.

The honest framing, which Moz itself uses, is that DA is a prediction: an estimate of how likely a domain is to rank relative to competitors, reverse-engineered from the outside. It is a weather forecast about your website, produced by a company that has never seen Google's actual weather instruments. Forecasts can be useful. They are not the weather. If you want the full teardown of how the vendor scores are built and where each one distorts, our Ahrefs vs Moz comparison walks through the crawler and index differences line by line.

Google's own answer, on the record more than once

This is not a matter of interpretation, because Google employees have addressed the question directly and repeatedly. In 2019, asked on Twitter whether Google used a site-wide metric like DA, Search Advocate John Mueller answered flatly that Google does not use domain authority. In a 2022 Reddit thread he went further, saying Google does not use it at all for search crawling, indexing, or ranking, while acknowledging the number can still be handy for quick outside-in comparisons. Search Engine Journal's ranking-factor review reaches the same verdict, and Moz's own documentation concedes DA is not a Google ranking factor and has no direct effect on results.

Mueller has allowed that Google holds internal signals that map loosely onto site-wide quality, and Google's 2023 antitrust testimony surfaced references to site-level quality scoring. But none of those internal signals is Moz's number, none is purchasable, and none moves when your DA moves. When an agency reports "we raised your DA from 22 to 31" as a deliverable, it has reported a change in Moz's opinion of your links, which Google does not read.

Topical authority is emergent, and that is the whole point

Topical authority has no dashboard number, no vendor, and no 1-to-100 scale, which is exactly why it is the more real of the two. It is an emergent property: it appears when a site has covered a subject densely enough, with enough genuine expertise and enough internal connection between pages, that search systems begin treating the site as an entity that knows the subject rather than a site that mentioned it. A plumbing site with one page about sump pumps has a page. A plumbing site with eighteen interlinked pages covering sump pump sizing, battery backups, discharge lines, winter failures, permits, and replacement costs has a position in the topic graph.

Nothing about this is mystical. Modern retrieval systems, Google's included, map queries and documents into semantic space and resolve entities across pages. A cluster of interlinked, consistent, expert documents about one subject produces exactly the corpus shape those systems trust: every page reinforces the vocabulary, entities, and internal links of its siblings, so confidence in any one answer is propped up by the presence of the rest. That is Vector 9 in our methodology, Cluster: the deliberate construction of depth so the brand becomes the entity for a topic rather than a lone URL. The full build sequence, from question inventory to interlinking, is documented in our guide to building topical authority.

How each one is measured

DA is measured for you, instantly and for free, which explains most of its popularity. Paste a URL into Moz and a number appears. Topical authority has to be measured indirectly, and the honest proxies are behavioural: how many distinct queries within one subject the site ranks for, whether new pages in the cluster rank faster than the site's historical average, whether the site earns featured snippets and AI Overview citations for subject questions it has covered, and whether AI assistants name the brand when asked about the subject. Those are harder numbers to pull than a DA score. They are also the only ones tied to revenue.

How each one gets gamed, and who pays for it

Because DA is computed from links, it can be inflated with links, and an entire grey industry exists to do it: private blog networks, expired-domain redirects, paid guest posts on sites built to sell them. The trap is that gaming DA fools Moz without fooling Google; worse, the same manipulated links that raise the score can trigger Google's link spam policies and SpamBrain's link-graph analysis. We have audited backlink profiles where the money spent inflating a vendor score would have funded a year of real content.

Topical authority gets gamed too, mostly through programmatic page generation: hundreds of thin variations on one template, published to simulate coverage. That approach now collides with Google's scaled content abuse policy, formalized in March 2024, and with quality classifiers that read the corpus rather than the page. Simulated depth has the structure of coverage with none of the information, and both search and AI systems have become competent at telling the difference. The version that works cannot be faked cheaply, because the thing being measured is accumulated, interconnected substance.

Why AI engines reward depth: the long-tail evidence

The strongest reason to prioritize topical depth in 2026 comes from research into how language models learn in the first place. A 2023 study by Kandpal and colleagues at UNC, Google Research, and UC Berkeley (arXiv:2211.08411) tested whether a model's ability to answer a factual question depends on how often the relevant facts appeared in its training data. The dependence turned out to be steep: BLOOM-176B's accuracy jumped from roughly 25 percent to above 55 percent as the count of relevant pre-training documents rose from tens to tens of thousands. As the authors put it, "language models struggle to capture the long-tail of information on the web."

For a small business, that finding cuts two ways. Your brand, your services, and your regional specifics live deep in that long tail; a model trained on the open web has effectively never heard of you, so it cannot recommend you from memory. But the same study found "retrieval-augmentation can reduce the dependence on relevant pre-training information," and retrieval is precisely how ChatGPT Search, Perplexity, Gemini, and AI Overviews now operate: they fetch live documents and cite what they fetch. A dense, interlinked cluster about your subject is retrieval-bait by construction. Every additional well-structured page in the cluster is another document the retriever can land on and another co-occurrence of your brand with your subject. We unpack the paper's full implications for niche brands in our long-tail knowledge gap explainer, and the citation-tactics side of the same coin in our breakdown of the Princeton GEO paper.

Notice what plays no role anywhere in that mechanism: a vendor's link score. No retriever consults Moz.

What the 2026 citation numbers say

The observational data lines up with the mechanism. A 2026 AI visibility audit reported that sites with full topical coverage were about 6.5 times more likely to be cited by AI tools than sites with fragmented coverage of the same subjects. The same body of research finds a rank-1 organic result carries roughly a 58 percent chance of AI citation against 14 percent for position ten, and Ahrefs' analysis of 17 million citations found AI-cited content averages about 26 percent fresher than what ranks in classic results. Coverage depth, ranking strength on the specific query, and freshness: all three are properties of a maintained topic cluster, and none of the three is a property of a domain-level link score.

In Matt's audits of Ontario service businesses, a recurring pattern makes the gap concrete: owners can quote their DA from a free checker, but almost none have ever asked an AI assistant the questions their own customers ask, and when they do, the businesses being named are the ones whose sites answer those questions in depth, not the ones with the prettiest link profiles. No prospect has ever asked an assistant for "a DA 40 company." They ask who fixes the problem, and the engines answer from coverage.

Making topic ownership machine-readable

Topical authority is earned in the writing, but it should also be declared in the markup, because entity resolution is easier when every page in a cluster states what it is about and what it belongs to. The about and isPartOf properties on Article schema do that work. Here is a minimal pattern for one page inside a sump pump cluster:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Sump Pump Battery Backup Sizing",
  "about": {
    "@type": "Thing",
    "name": "Sump pump systems",
    "sameAs": "https://en.wikipedia.org/wiki/Sump_pump"
  },
  "isPartOf": {
    "@type": "WebSite",
    "@id": "https://example.com/#website"
  },
  "author": {
    "@type": "Person",
    "name": "Jane Doe",
    "jobTitle": "Master Plumber",
    "knowsAbout": ["sump pumps", "basement waterproofing"]
  }
}

Repeated across a cluster with a consistent author entity and a shared WebSite node, this gives crawlers and retrievers the same message the internal links already carry: these pages belong together, one accountable expert stands behind them, and the subject is declared rather than implied. Markup does not create authority, but it removes ambiguity from authority you have earned. This mapping-and-declaring work sits inside Vector 6 of our process, Structure.

The verdict for a small business budget

Chase topical authority. Treat domain authority as a thermometer you glance at quarterly, never as a target you pay to move. The reasoning is arithmetic as much as strategy: a national competitor with a 15-year-old link profile has a DA lead you will not close on any budget a small business holds, but their coverage of your specific niche in your specific region is usually shallow, and coverage is a contest you can win by simply doing the work they will not. Depth is also the rare investment that pays on both surfaces simultaneously, classic rankings and AI citations, because both systems select from the same corpus you are enriching.

We have watched this play out on our own flagship engagement. Mattress Miracle, an independent Brantford retailer, grew from roughly 1,000 to 82,400 monthly organic visits (SEMrush, April 2026) on the strength of exhaustive topic coverage, out-ranking national mattress brands whose vendor scores dwarf its own. Results of that size depend on industry, competition, and existing digital presence, and no honest agency promises them; the transferable lesson is narrower and more useful. The site did not win by acquiring a bigger number. It won by leaving no question in its subject unanswered.

One caveat keeps this honest: links still matter to Google, even though Moz's estimate of them is not a ranking input. Earned links from relevant industry sources remain a genuine signal, and a topic cluster that becomes the reference on its subject tends to attract exactly those links without a single outreach email. Build the depth, and the score you were told to chase quietly improves as a side effect. That order of operations, depth first and links as consequence, is the one the evidence supports.

Where the score still earns its keep

None of this makes DA useless; it makes DA a triage tool rather than an objective. Three legitimate uses survive. First, sizing up a SERP before committing to a keyword: if every ranking domain scores 70-plus and yours scores 20, the head term is probably a two-year project and the long-tail variants are the smarter entry. Second, filtering prospects when you are the one earning links, since a directional estimate of another site's link equity beats no estimate. Third, spotting trouble, because a sudden DA collapse often flags a lost batch of links worth investigating. What the score cannot do in any of those cases is tell you whether a site actually gets chosen for anything, which is the only question a marketing budget exists to answer. More of our primary-source breakdowns live in the research library.