The Moat Test: how eight Australian websites win competitive search results with almost nothing else

Written by Lucas Durante, Founder & CEO, and reviewed by Sara Smith, Head of SEO, Odin Digital.

Headline finding

Every one of the eight Australian sites we pulled apart wins on exactly one asset a competitor cannot copy this quarter, and the pages that rank are just the delivery mechanism for it. One plumbing site ranks 675 suburb pages whose body copy is 100 percent identical because it has 2,500 Google reviews. One life insurer ranks the most valuable page in its category with zero structured data. One law firm ranks 116 pages it has not touched since May 2024.

Key findings

• All eight sites have one uncopyable asset each, and it is a different asset in almost every case: review volume, a supplier-written directory, a brand plus calculators, an incumbent brand, one credentialled person, named practitioners, niche narrowness, and an exact-match domain.
• Proximity Plumbing ranks 675 blocked-drain suburb pages that share 100 percent of their body copy. Four pages we compared are 6,816 words each and differ by two words. Only the title, H1, canonical and geo-coordinates change. The site has more than 2,500 Google reviews at 4.9.
• TAL's life insurance page carries no JSON-LD at all. No FAQ, Product, Service or breadcrumb markup. It is 4,304 words on a brand Google has trusted for decades, and it ranks in one of the most valuable financial services categories in the country.
• Lord Commercial Lawyers has 116 of its 229 URLs last modified in May 2024 and still outranks national firms on Melbourne commercial law queries. Freshness is not what is holding that site up.
• Whelan Lawyers ranks on Wix with a 1.1 MB service page and three H1s, because 23 of its 170 pages are about franchising law and nothing else on the site dilutes that.
• Easy Weddings' category pages overlap only 41 percent between Sydney and Melbourne, not because someone wrote them differently, but because 2,979 vendors write their own storefronts and the category pages are built from them.
• Two of the eight break Google's structured data guidelines to keep review stars in search results. It works for them because their fundamentals would survive a penalty. A site with nothing else would not.
• Four of the six moat types are available to almost any business, and the two that are not, an incumbent brand and an exact-match domain, are the two nobody should plan around.

Why this matters in Australia

Most advice Australian businesses receive about search comes from two places: American publishers describing American markets, and agencies describing what they sell. Both tend to say the same thing. Publish more, publish fresher, make every page unique, fix the technical checklist, add structured data.

We wanted to know whether the sites that actually win hard Australian search results do any of that. So we took eight sites in eight brutally competitive local markets, sites that outrank better-funded competitors, and crawled them. Not their rankings, which were the starting condition, but their architecture: every URL in their sitemaps, the words on their pages, the similarity between pages, the structured data in their HTML, and what they claim about themselves.

The answer is that the standard advice describes the coverage layer, and the coverage layer is not what wins. This paper is for anyone deciding where the next twelve months of marketing budget goes, and for the agencies who have to justify it. It is also the diagnostic we now run before we take on a client, and the reason the rest of this research series exists.

Methodology

Subjects. Eight Australian websites supplied as known winners in competitive search categories: Proximity Plumbing (Sydney plumbing), Easy Weddings (national weddings marketplace), Compare the Market (insurance comparison), TAL (life insurance), A Lovely Planet (travel content), Lord Commercial Lawyers (Melbourne commercial law), Whelan Lawyers (Melbourne commercial and franchise law), and Commercial Cleaning Sydney (Sydney office cleaning). They were chosen to span local services, marketplaces, financial services, content publishing and professional services, and because in each case a smaller or less-funded site is outranking larger competitors on head terms.

Crawl. All eight were crawled on 2 September 2026 with a desktop user agent and no JavaScript execution. URL counts come from each site's own sitemaps, recursively expanded through sitemap indexes. Roughly 18,000 URLs were mapped across the set.

Page similarity. Jaccard overlap over full-page word sets, including navigation and footer text. Comparisons between pairs of pages on the same site are meaningful. Absolute values are less so, because shared navigation inflates every score.

Structured data. JSON-LD parsed from raw HTML on every crawled page. Microdata was not parsed.

Claims and reviews. Review counts, awards and guarantees are read from on-page copy and schema. Google review totals are the businesses' own published claims and were not independently verified against Google.

Limitations. A Lovely Planet returned HTTP 403 to every non-browser request, including robots.txt and every sitemap path, so its architecture could not be crawled. Background on its authorship comes from published interviews. Ranking performance itself was not independently measured. These sites were supplied as known winners, and this study reverse-engineers how they win, not whether.

What this is not. Eight sites is a qualitative study, not a statistical one. The finding is a pattern that held across all eight, with no exceptions, in eight different categories. It is the hypothesis for the quantitative studies that follow in this series, not a proof. The State of Australian Small Business Websites study, published alongside this paper, measures 1,480 sites against the coverage-layer checklist and finds that most fail it; this paper is the argument for why passing it would not have been enough.

Finding 1: one moat each, and everything else is coverage

The hypothesis going in was that these sites would show craft: better content, tighter architecture, sharper differentiation. Some do. But the pattern that holds across all eight is blunter.

Each site has one asset a competitor genuinely cannot copy this quarter. Everything else on the site, including work that would fail a standard audit badly, exists only to give that asset a URL to land on for each query.

URL counts from each site's own sitemaps, 2 September 2026. TAL's sitemap is paginated beyond a reasonable crawl.

The practical reading: identify the one moat you can build in your market, build it obsessively, then use programmatic coverage to cash it in. Doing the coverage without the moat is what produces the 100-page site that ranks for nothing, and that is most of the market.

Finding 2: differentiation is a metadata problem once the entity is strong

Proximity Plumbing is the single most instructive site in the set. We compared four of its blocked-drains suburb pages: Bondi, Vaucluse, Alexandria and Parramatta. Each is 6,816 words. The vocabulary sets differ by two tokens. Jaccard similarity between them is 100 percent.

What does vary is the metadata layer, and only that:

/blocked-drains/bondi/      title: Blocked Drains Bondi | Proximity Plumbing | 5 Rated
                            h1: Blocked Drains Bondi        lat: -33.8930748
/blocked-drains/vaucluse/   title: Blocked Drain Vaucluse | Proximity Plumbing | 5 Rated
                            h1: Blocked Drains Vaucluse     lat: -33.855761
body copy: identical, 6,816 words, 2 tokens different

Every guide on plumbing SEO says explicitly not to do this. "Genuinely localised content, not thin templated pages where only the suburb name changes." It works anyway, and it works because the moat is off-site and overwhelming: more than 2,500 Google reviews at 4.9, a Waverley Local Business Award, a lifetime labour warranty, a $0 call-out, a 30-minute response guarantee. The site has 675 blocked-drain suburb pages and a further 629 pipe-relining suburb pages built the same way, 2,411 URLs in total.

The underrated part is 247 individual project pages naming real jobs at real addresses: pipe relining in Vaucluse, strata hot water in Bellevue Hill, a roof leak in Surry Hills. That is first-hand experience in the sense Google's quality guidelines mean it, and it is the only genuinely unique content on the domain. It is also the asset almost every trade, clinic and agency has sitting unused in a job history.

Distribution is a flat hub. The areas-we-service page is 988 words carrying 250 internal links. Each suburb page carries 108 internal links but only about ten to sibling suburbs. Cheap, flat, effective.

Read this as a licence to scale, not a licence to skip the moat. On a weak entity, 675 identical pages are a thin-content liability and the kind of thing a quality update removes in a day. On a strong one, they are coverage.

Finding 3: the most defensible content is written by other people

Easy Weddings is a marketplace, not a publisher. Its architecture is roughly 35 categories crossed with 45 locations: wedding venues, photography, celebrants, cakes, each against every market. That produces 1,558 category-by-location pages, 1,240 suburb pages and 4,623 articles. Underneath sit 2,979 vendor storefronts that the vendors write and maintain themselves.

The category-by-location pages are genuinely differentiated. Sydney venues and Melbourne venues overlap only 41 percent. Venues and photography in the same city overlap 31 percent. Nobody templated that. The vendors did, one listing at a time, and each listing carries LocalBusiness markup with address and opening hours. A single category page is a structured directory of that market. The Sydney venues page runs to 10,910 words and 232 internal links.

The lesson generalises further than weddings. If a business has suppliers, partners, franchisees, members, stockists, or alumni, a structured directory of them is a moat that scales without a content team, and it produces exactly the differentiation programmatic pages usually lack. The people listed also link to their own listing, which is the link profile taking care of itself.

Finding 4: a calculator is the one thing an AI answer cannot summarise

Compare the Market is the textbook version of the model most agencies teach, executed by someone with a television budget. Deep vertical clusters, 271 pages on health insurance alone. A 567-post news operation that keeps the domain in the news cycle. The health insurance hub runs to 8,820 words, 162 internal links and 20 FAQ questions in schema.

The part most agencies skip is a real calculator per vertical: Lifetime Health Cover loading, Medicare Levy Surcharge, car insurance, life insurance. A person arrives with a question, finishes the job on the page, and never needs the insurer's own site. That is how an affiliate comparison site outranks the products it compares.

It is also why the cluster is defensible against an AI summary, which is the question every content-led site now has to answer. An AI engine can read a 3,000-word guide to the Medicare Levy Surcharge and give a user the answer without a click. It cannot run the calculation for that user's income, family status and cover date without sending them to the tool. The calculator is the part of the page that survives the AI Overview.

Finding 5: for an incumbent, the technical layer is decoration

TAL is the control in the experiment. Its life insurance page, one of the most commercially valuable URLs in Australian financial services, ships no structured data at all. No FAQ markup, no Product, no Service, no breadcrumbs. Sitewide we found one schema block, an Organization on the homepage. The page is 4,304 words and 122 KB of HTML.

This is worth internalising precisely because it is annoying. For an entity with genuine brand authority in a category where Google is cautious, the technical layer is almost decorative. Schema is an accelerant for entities Google is unsure about, not a ranking factor for entities it already knows. Never benchmark a challenger's technical checklist against an incumbent's. TAL is not ranking because of what is on the page, and copying its on-page setup would be copying the wrong half of the machine.

The corollary matters for every business that is not TAL: the technical checklist is worth more to you than it is to them, because you are the entity Google is unsure about. That is why the State of Australian Small Business Websites study measures it. It is necessary. It is nowhere near sufficient.

Finding 6: in content, the entity is a person

A Lovely Planet sits behind a Cloudflare bot challenge. It returned 403 to every non-browser request we made, including robots.txt and every sitemap path. That almost certainly locks out GPTBot, ClaudeBot and PerplexityBot too. In a period when every agency is selling AI-search visibility, a site that ranks in one of the most competitive content niches on earth is refusing the whole conversation.

The moat is a person. Hayley Lewis started the blog in 2014 and writes for the Telegraph, Metro and the Evening Standard, and produces content for Lonely Planet. That is experience and expertise as Google defines them, corroborated on third-party domains she does not control. The bylines are the link profile. For a founder-led business this is the cheapest authority route available, and it transfers in a way a company brand does not: the person can change agencies, start a second business, or sell the first, and the entity goes with them.

Finding 7: in professional services, depth and names beat freshness

Lord Commercial Lawyers is a boutique beating national practices on Melbourne commercial law with a 229-page site. The structure is orthodox and tight: 18 practice-area pages, standalone long-form guides that are not blog posts (buying commercial premises, non-disclosure agreements, guarantees and indemnities, probate in Victoria), six named solicitors with individual pages, and a hand-written llms.txt that reads like a firm brochure rather than a generated dump.

The uncomfortable part is that 116 of the 229 URLs were last modified in May 2024. In slow-moving categories, a correct answer written once by a named solicitor decays far more slowly than the industry's publishing cadence implies. The guides are deliberately separated from the blog. Evergreen assets get their own top-level URLs, not a dated feed.

Whelan Lawyers makes the same point from the other direction. It is a Wix site with a 1.1 MB service page and three H1s, ranking in Melbourne commercial law. Every technical instinct says this should not work. It works because of narrowness. Rather than competing for "commercial lawyer Melbourne" head-on, the site is built around franchising law, 23 dedicated pages, with construction, corporate, commercial and property leasing as secondary clusters and 93 insight articles feeding them. One subject treated like an obsession.

Whelan's llms.txt is auto-generated by Wix and exposes a live MCP endpoint at /_api/mcp with named tools for AI agents to query the site directly. A small firm got AI-agent infrastructure for free from its platform while agencies charge to build it by hand. The technical floor is rising without anyone hiring an SEO, which is one more reason the floor is not where the competition is.

Finding 8: go one level below the city

Commercial Cleaning Sydney is the smallest site in the set, 122 URLs on WordPress, and the one with the most obvious single cause. The domain is the query. Exact-match domains were supposedly neutered a decade ago. On a commercial-intent local query with a modest competitive field, this one is still carrying the page.

What the site adds on top is cheap and smart: micro-local content that names buildings and precincts rather than cities. Hygiene standards for medical facilities in Sydney CBD. High-rise window care. Strata maintenance. Cleaning standards in Barangaroo. Each is a real query with a real buyer and near-zero competition, and each article carries Article and Person schema with a named author. The homepage runs to 2,309 words.

Proximity targets 675 suburbs, not eight capitals. Easy Weddings has 1,240 suburb pages. The precinct layer is where competition thins out and intent stays commercial, and most agencies stop at the capital city. Our own site had 178 location pages when this study was run, and every one of them was a city.

Six patterns that hold across all eight

1. One moat, then coverage. Not one of these sites wins on page craft. Each has a single uncopyable asset and uses programmatic pages purely to put that asset in front of every query variant. The failure mode most sites have is the reverse: coverage with no moat behind it.
2. Differentiation is a metadata problem, not a copy problem. Proximity's suburb pages share 100 percent of their body copy but every one has a unique title, H1, canonical and geo. That is apparently sufficient once the entity is strong. On a weak entity it is a thin-content liability.
3. Go one level below the city. 675 suburbs, 1,240 suburb pages, Barangaroo and the CBD strata market. The precinct layer is where competition thins out and intent stays commercial.
4. Named humans, everywhere it matters. Six solicitor pages. One person with national mastheads. Person author schema on a cleaning blog. Even the smallest sites do this. Company-as-author is the outlier, and it is the weaker choice.
5. Real jobs beat written content. Proximity's 247 project pages, actual addresses and actual work, are the only unique content on that domain and are worth more than the 675 templated pages combined. Every trade, clinic and agency has this asset sitting unused in a job history.
6. Tools outrank articles. A calculator per vertical. A marketplace that is functionally a search tool. Both outrank the products and vendors they index. An AI answer can summarise your guide. It cannot run your calculator.

What the eight change about the standard model

Two things they do that you should not copy

Product schema for review stars. Both Proximity Plumbing and Lord Law mark up a service as a Product with an aggregateRating. That is how a site still gets star ratings in search results after Google stopped showing them for self-serving Organization and LocalBusiness markup in 2019. Lord Law's commercial law practice area is a Product with a Brand and a 4.8 rating. Proximity's suburb pages are a Product with 581 reviews, the same 581 on every one of the 675 pages regardless of suburb. It works, and it is against the guidelines. It is one update or one manual action from disappearing, and it takes the rankings with it.

Keyword-stuffed structured data. Proximity's PostalAddress.streetAddress reads "Plumber in Point Piper" rather than a street address. The Place on every suburb page is named "Blocked drains NSW". Small, deliberate, and the kind of thing that reads as manipulation if anyone looks.

Note the asymmetry. Proximity can absorb a penalty because 2,500 reviews and a real local business survive it. A site whose only asset is the markup cannot. A business with a genuine moat can afford a policy violation. A business without one is betting the domain on it. The State of Australian Small Business Websites study found self-declared rating markup on 8.9 percent of Australian business homepages, and on more than a quarter of pest control sites. Almost none of them have Proximity's cushion.

The moat diagnostic

Before keyword research, before a technical audit, answer one question: what is this business's moat, and does it already exist?

Two of the eight sites win on assets that are not on this list: an incumbent brand and an exact-match domain. Neither can be planned for. If you have one, use it. If you do not, do not spend a year trying to build one.

The sequencing matters. Moats take months. Coverage takes weeks. The instinct is to do the fast, visible work first. Every site in this study says the opposite: the coverage only pays once there is something behind it.

How to run the moat test on your own site

1. Count your reviews on Google. Then find the top three competitors for your main query and count theirs. If you are not in the top three on review count, that is the gap, and it is the whole gap.
2. List every named person on your site. If the answer is none, or a company name, you have no entity. Pick one person and start.
3. List every page that describes a real, specific job, project or case. Not case studies waiting for a client's approval. Jobs. If the number is under ten, that is the cheapest unique content you will ever produce.
4. Ask what your customer calculates before buying. A quote, a dosage, a payback period, a tax outcome, a comparison. If there is an answer, that is the tool.
5. Ask who else could write pages for you. Suppliers, partners, alumni, stockists, franchisees. If there is a list, that is the directory.
6. Ask what you know from operations that nobody outside can see. Cost per lead, response times, conversion rates, seasonal demand. If there is data, that is the report.
7. Only then look at the coverage layer: which suburbs, which services, which precincts, and whether the technical basics are in place. The State of Australian Small Business Websites study is the checklist for that layer.

What we did with it

The audit that preceded this study found our own site technically excellent: 389 pages, clean architecture, de-templated city pages, real schema. This study explained why that had not converted. We had the coverage and not the moat: 25 Trustpilot reviews, no indexable case studies, one tool, and a founder who was not yet a searchable entity. Against the diagnostic, we had access to four of the six moats and were building none of them at pace.

So the order changed. Reviews became the first commercial priority ahead of any search work. Project pages replaced case studies. The founder became the entity, with bylines and a Wikidata item. One tool per vertical, starting with an advertising-compliance checker for healthcare. And this research desk, because "proprietary data from operations" was the row we could act on fastest. The rest of this series is that decision playing out in public.

Dataset and citation

• Dataset: moat-test-2026-09.csv, one row per site: URLs, page counts, moat category, similarity measurements, schema counts, last-modified distribution.
• Suggested citation: Durante, L. (2026). The Moat Test: how eight Australian websites win competitive search results with almost nothing else. Odin Digital Research. odindigital.com.au/research/the-moat-test
• Charts on this page are embeddable with a credit link. Media enquiries: [email protected]
• Sites named in this paper are named because the findings about them are complimentary and the data is public. No site was contacted before publication.

Back to Odin Digital Research