How Google determines if a website is relevant and authoritative

What "website authority" means, and who defines it

The phrase has two meanings, and they are often mixed up.

In Google's own vocabulary, authority is a quality concept. The Search Quality Rater Guidelines ask raters to consider how far the content creator or the website "is known as a go-to source for the topic", and note that most topics have no single official source. Google's ranking systems guide describes reliable information systems that "surface more authoritative pages and demote low-quality content". Authority in this sense is not a number. It is a judgement built from evidence, and it is topical: the guidelines point out that a site can be a go-to source for one type of content and untrustworthy for another.

In SEO tools, authority is a vendor's score: Moz's Domain Authority (DA) and Page Authority (PA), Ahrefs' Domain Rating (DR) and URL Rating (UR), Semrush's Authority Score (AS). Each is computed by the vendor from its own link data, and each tries to predict ranking ability. None comes from Google.

So an "authority website", in the sense that matters for Google, is a site that people and other sites treat as a go-to source on its subject. A high DA or DR often goes with that reputation, because such sites attract links, but the score is a side effect, not the cause.

How Google decides whether a page is relevant

Relevance is judged per page and per query, after Google has worked out what the query means. Google's page on how Search ranks results describes the work like this:

• The meaning of the query. Language models interpret the few words typed, correct spelling and connect synonyms, so a page about adjusting laptop brightness can answer a search for changing it.
• Keywords on the page. The most basic relevance signal is that the page contains the same words as the query, for example in its headings or body text.
• Relevant content beyond the keyword. Google's example: for "dogs", a page that repeats the word hundreds of times is less useful than one with pictures, videos or a list of breeds.
• Aggregated interaction data. Google says it uses aggregated and anonymized interaction data to assess whether results are relevant to queries, and turns it into signals for its machine-learned systems.

Several named systems do this work. BERT helps Google understand how combinations of words express meaning and intent, RankBrain relates words to concepts, neural matching matches concepts in queries and pages, and passage ranking judges individual sections of a long page. The words in a domain name count as one small relevance factor, and the exact match domain system stops a domain such as best-places-to-eat-lunch from getting too much credit for its name.

How much each factor weighs depends on the query. Google's own example is freshness: it matters more for current news than for a dictionary definition. That is one reason no single number, from Google or anyone else, can describe how well a site "ranks" in general.

How Google judges quality, trust and authority

Once Google has a set of relevant pages, it tries to put the most helpful and reliable first. Its ranking page says its systems look for signals that help determine which content demonstrates expertise, authoritativeness and trustworthiness. It names one of those signals directly: whether other prominent websites link to or refer to the content, which it calls generally a good sign that the information is trustworthy. Aggregated feedback from its quality evaluation process then helps refine how the systems judge quality.

The signals Google has described in its own documentation:

Usability works more like a tiebreaker. Google says that when all other signals are relatively equal, content people find more accessible, such as mobile-friendly pages that load quickly, may perform better. It does not substitute for quality.

Where PageRank and links fit

PageRank is the link analysis algorithm Sergey Brin and Lawrence Page described in their 1998 paper, The Anatomy of a Large-Scale Hypertextual Web Search Engine. The patent states the idea plainly: a page's rank is calculated from the ranks of the pages that link to it, each divided by the number of links on that page, plus the chance that a "random surfer" jumps to any page instead of following a link. The result is that a page with one link from a very highly ranked page can outrank a page with many links from obscure ones.

That model dates from 1998, and Google says how PageRank works "has evolved a lot" since. It is still part of the core ranking systems, next to other link analysis systems that use links to understand what pages are about. Google's SEO Starter Guide puts it in proportion: PageRank uses links and is one of Google's fundamental algorithms, but there is much more to Search than links, and PageRank is one of many ranking signals.

Links made to game this are a policy issue, not a strategy. Google's spam policies define link spam as creating links primarily to manipulate rankings: buying or selling links that pass ranking credit, excessive link exchanges, automated link creation and similar schemes. Paid links are fine when they carry rel="sponsored" or rel="nofollow". The Penguin system, built against link spam in 2012, became part of the core ranking systems in 2016.

To see which links Google knows about, use the Links report in Search Console. It shows a sample of external and internal links, grouped by root domain, with your top linked pages, top linking sites and most common link text. It shows no score, because Google does not publish one.

What E-E-A-T is, and what quality raters actually do

E-E-A-T stands for Experience, Expertise, Authoritativeness and Trust. The rater guidelines define each: experience is first-hand or life experience of the topic, expertise is the knowledge or skill it needs, authoritativeness is being known as a go-to source, and trust is whether the page is accurate, honest, safe and reliable. Trust sits at the centre: an untrustworthy page has low E-E-A-T however expert it looks. Google added the first E, for experience, in a December 2022 update to the guidelines. The current version is dated September 11, 2025.

E-E-A-T is not a ranking factor in itself. Google's guide to helpful, reliable, people-first content says its systems use a mix of factors that can identify content with good E-E-A-T, and give even more weight to such content on "Your Money or Your Life" (YMYL) topics, which could significantly affect people's health, financial stability or safety, or the welfare of society. The SEO Starter Guide is blunter: it lists the belief that E-E-A-T is a ranking factor among the things not to focus on, and answers it with "No, it's not."

The people who apply E-E-A-T are search quality raters. Google's overview of the guidelines (November 2023) puts them at about 16,000, employed by external vendors across four regions and more than 80 languages. They rate results on two scales: Page Quality, how well a page achieves its purpose, and Needs Met, how useful a result is for a given search. They also take part in side-by-side experiments, where they compare results with and without a proposed change and say which they prefer. Google's testing page counts 719,326 search quality tests and 124,942 side-by-side experiments in 2023.

What raters do not do matters just as much. Google says they have no control over how pages rank and that rater data is not used directly in its ranking algorithms. The guidelines add that no single rating can directly change how a page or site appears, because humans could never rate every page on the web. Ratings are aggregated to measure whether Google's systems work, and serve as examples of good and bad results to improve them.

The guidelines are still worth reading as a self-audit, because they show how a careful outsider judges E-E-A-T. Raters look at three things: what the site and its authors say about themselves (an About page, author profiles), what independent sources say about them (reviews, references, news coverage), and what the page itself shows.

Does Google use domain authority?

Not in the sense the question usually means. Domain Authority is Moz's metric, and Moz says it is not a Google ranking factor and has no effect on search results. DR and AS are likewise proprietary metrics of Ahrefs and Semrush. If you meet the statement "domain authority is a ranking factor in Google's algorithms" in a quiz, the answer is false.

Google does, however, have site-level signals of its own, and says so. The ranking systems guide explains that its systems are designed to work on the page level, while site-wide signals and classifiers are also used and contribute to its understanding of pages. It adds the caveat that matters: good site-wide signals do not mean every page on a site will rank highly, and poor ones do not mean every page will rank poorly. The spam policies go further. As of September 2026, they state that Google generally presumes individual pages, including new pages, match the overall quality of the other pages on the domain.

Two spam policies introduced with the March 2024 core update show that site reputation is real enough to be abused:

• Expired domain abuse: buying an expired domain and filling it with low-value content, hoping to rank on the reputation of its previous owner.
• Site reputation abuse: publishing third-party pages on a host site mainly to benefit from ranking signals the host earned with its own content. For a dated timeline of both policies and later updates, see SEO changes from 2024 to 2026.

Internal documentation points the same way. The published reference for Google's Content Warehouse API client (the google_api_content_warehouse package on Hex, built from Google's googleapis repository) lists a field named siteAuthority among compressed per-document quality signals, without saying how it is computed or weighted. When about 2,500 leaked internal Search documents circulated in May 2024, Google confirmed to The Verge that they were authentic and warned against drawing conclusions from out-of-context, outdated or incomplete information.

The safe reading is the one Google states itself: site-level signals exist, you cannot see them, and no third-party score is a readout of them. The same applies to "Google page authority": Google ranks pages, but it has no public page score, and Moz's Page Authority is Moz's prediction, not Google's.

How Moz, Ahrefs and Semrush calculate authority scores

Each vendor builds its own index of links across the web and computes its score from that index. They differ in what goes into the score:

Moz Domain Authority predicts how likely a site is to rank compared with competitors. Moz trains its model against thousands of search results, so DA is an attempt to imitate ranking outcomes from link data. It is relative: when very large sites gain links, others can drop, which is why Moz says it is easier to go from 20 to 30 than from 70 to 80. Scores also move when Moz's index or model changes; Moz cites the 2019 rollout of Domain Authority 2.0 as an example. Page Authority applies the same method to a single page.

Ahrefs Domain Rating measures only the strength of a site's backlink profile compared with every other site in Ahrefs' database. Ahrefs explains that each linking domain splits its rating equally among the domains it links to, so a DR 10 site linking to three domains can pass more than a DR 80 site linking to a million. Nofollow links do not pass DR, and the exact formula is not disclosed. Because the scale is logarithmic and relative, your DR can fall when you lose no links at all, simply because others gained more.

Semrush Authority Score is a compound grade. Semrush's documentation lists link power (followed referring domains and their authority), organic traffic estimated from keyword positions, and six spam checks, such as more than 90% of referring domains sending only followed links, or many referring domains on the same IP network. Scores are refreshed every two weeks.

Moz trains its model on search results and Semrush estimates traffic from Google rankings, so both watch the output of search. None of the three can see Google's site-wide signals, its interaction data, its spam demotions or its manual actions. Semrush says as much: a Google penalty is unlikely to change a site's Authority Score.

How to use third-party authority scores sensibly

The scores are useful for comparison and triage. They are misleading as goals.

1. Compare like with like. Use one tool, on one date, against the sites that actually rank for the queries you want. A DA of 40 and a DR of 40 come from different indexes and formulas, and say nothing about each other.
2. Read trends, not points. A jump or drop after the vendor re-crawls or recalibrates says more about the tool than your site. Look at the direction over months.
3. Use them to sort, then look. A score can shortlist competitors or flag a suspicious link source. Ahrefs itself warns against judging a site on site-wide authority alone: check whether it has organic traffic, relevant content and a good chance of still being around in a few years.
4. Report outcomes, not proxies. Clicks, impressions and queries in Search Console's Performance report are Google's own data about your site. Semrush recommends pairing its score with conversions, rankings and traffic rather than relying on it alone.
5. Remember what the scores cannot see: content quality, E-E-A-T, site-wide demotions, manual actions and relevance to a specific query.

How to build the authority Google actually rewards

Google's guidance on helpful content reduces to three questions about every page: who made it, how, and why. The practical version:

1. Make the "who" obvious. Put bylines on articles, link them to author pages that describe the author's background, and keep an About page that says who runs the site. Google asks whether it is self-evident who created the content.
2. Show the "how". Cite primary sources, describe how you tested or researched what you claim, and disclose automation or AI use where a reader would reasonably ask how the content was produced.
3. Stay on your subject. Google's self-assessment asks whether the site has a primary purpose, and warns against entering a niche without real expertise to chase traffic. Authority is topical, so depth in one area beats thin coverage of many. Depth is not length: Google says it has no preferred word count.
4. Be worth citing. Google's self-assessment asks whether content offers original information, research or analysis, and whether you would expect a printed magazine, encyclopedia or book to reference it. That is the kind of page prominent sites link to and refer to, which is the signal Google names. Earn those references; do not manufacture them.
5. Look after the whole site. Site-wide signals, and Google's presumption that new pages match the rest of the domain, mean weak sections are not isolated. Correct pages that are wrong, and host third-party content only with real editorial oversight. Mass-deleting or re-dating old posts to look fresh is a different thing, and Google says it does not help.
6. Check your work in Search Console. Watch which queries each page gains, and read the Manual actions report if traffic drops sharply.

If you would rather have someone else audit this, our SEO service reports from Search Console, server logs and field data rather than vanity dashboards, adds author and date markup that supports E-E-A-T, and earns links with content worth citing instead of buying them. More articles on search are in the SEO topic.