The State of Australian Small Business Websites 2026

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

Headline finding

Two in three Australian small business websites take more than a second to start responding, one in five has no structured data at all, one in eleven blocks the crawlers that AI engines use to read the web, and one in eight makes a "best", "leading" or "number one" claim about itself on its homepage. Only 30.7 percent pass all five of the basic checks an AI engine needs to read, place and trust a business.

We crawled 1,820 Australian business websites across 39 industries and eight capital cities on 3 September 2026. 1,480 could be fully measured. The rest blocked us, timed out, or could not complete a secure connection.

Key findings

• 22.5 percent of sites have no JSON-LD structured data, and 61.5 percent have no local business markup of any kind, which is the markup that tells Google and AI engines what the business is and where it operates. Of the sites that do have structured data, 40.0 percent have only what their platform ships by default. (n = 1,480)
• Only 30.7 percent of sites are AI-readable on all five basic counts: reachable by a non-JavaScript crawler, not blocking AI bots, carrying structured data, carrying local business markup, returning a real 404, and publishing a sitemap. Plumbers are the best industry at 67.3 percent. Real estate agencies are the worst at 12.6 percent.
• 66.4 percent take more than one second to return the first byte of the homepage, and 28.2 percent take more than three. The median is 1.5 seconds. Measured from Melbourne with a single request per site, so treat the absolute figure as indicative and the platform comparisons as robust.
• 9.3 percent of sites blocked our crawler outright with a 403, a 429 or an empty challenge page. A standard non-JavaScript client, which is what most AI crawlers are, cannot read one in eleven Australian business homepages. Legal and trades sites block most, at 12.8 and 12.7 percent.
• 13.2 percent make a superlative claim on the homepage: "Australia's best", "leading", "number one", "award-winning". Mortgage brokers lead at 46.4 percent. Landscapers, plumbers and electricians are all above 26 percent. Dermatologists and cardiologists are at zero.
• 8.9 percent carry self-declared star-rating markup on their own homepage. Google stopped honouring self-serving ratings in 2019 and its guidelines call this misuse. 87 of the 131 sites doing it have no visible third-party review widget, meaning the rating is typed into the code rather than pulled from a review platform. Pest control companies are the worst at 27.3 percent.
• 6.8 percent return a 200 status for a URL that does not exist, the soft-404 defect that lets a site's index fill with pages that are not real. Just over half of those redirect the missing URL to the homepage. Pest control at 14.5 percent and dentists at 10.2 percent are the worst industries.
• Only 5.0 percent block any AI crawler in robots.txt. GPTBot is blocked by 3.9 percent, ClaudeBot by 3.4 percent, Google-Extended by 3.4 percent. The debate about blocking AI is happening almost entirely at large publishers. Australian small businesses are, by default, letting every AI engine in. 1.1 percent reference an llms.txt file.
• 33.5 percent have no analytics we could detect. Professional services firms are the worst at 54.9 percent without Google Analytics or Tag Manager.
• WordPress runs 60.5 percent of the sample. Squarespace is second at 5.6 percent and Wix third at 4.3 percent. Wix homepages are the heaviest, with a median of 676 KB of HTML against 190 KB on WordPress. Squarespace and Duda are the slowest to respond, with medians above three seconds.
• Sites that rank in search are more optimised and more boastful than sites that do not. Businesses we found through search results had structured data 83.0 percent of the time and made superlative claims 16.9 percent of the time. Businesses we found through OpenStreetMap, which does not care whether a site ranks, had structured data 71.8 percent of the time and made superlative claims 10.0 percent of the time.
• Only 31.4 percent get all five on-page basics right: exactly one H1, a title under 60 characters, a meta description, a viewport tag and an HTTP-to-HTTPS redirect. 18.0 percent of homepages have no H1 and 18.4 percent have more than one.

Why this matters in Australia

The website statistics Australian business owners read are American. They describe American hosting, American platforms and American consumers. Nobody has measured the Australian small business website at scale, by industry and by city, with a published method that anyone can repeat.

It matters more now than it did two years ago, for one reason. The readers of a business's website are no longer only people. ChatGPT, Gemini, Perplexity and Google's AI Overviews now recommend businesses by name, and they form those recommendations from what they can read. A site that returns a challenge page to a crawler, has no structured data, and takes two seconds to respond is invisible to that layer of the internet whether or not it ranks in classic search.

Three groups will use this paper. The business owner deciding whether their website is fine. The agency or web designer who has to tell them it is not, and who inherits the platform defaults measured here. And anyone in the industry who wants an Australian baseline to compare a client against. It is the first edition of an annual series, and the method is published so the second edition can be compared with the first.

Methodology

Sampling frame. We needed businesses, not websites, so we built the sample from two independent sources and merged them.

1. Search results. 2,691 Bing queries of the form "{industry term} {city or suburb}" across 39 industry terms, eight capital cities and 40 city and suburb terms, run on 2 and 3 September 2026. Bing was used because Google blocks automated querying. This source is biased toward businesses that already rank, and we report it separately for that reason.
2. OpenStreetMap. Every business point in the eight capital-city metropolitan areas carrying a website tag and one of 28 OpenStreetMap category tags matching our industries (dentist, physiotherapist, lawyer, accountant, plumber, electrician, and so on), retrieved from the Overpass API on 3 September 2026. This source does not care whether a business ranks, and it contributed 773 of the 1,480 usable sites.

Inclusion. A domain entered the sample only if it was on an Australian top-level domain (.com.au, .net.au, .au, .id.au), was not on a 176-entry exclusion list of directories, aggregators, media outlets, government, education, franchisor and national-brand domains, and did not appear for four or more unrelated industries, which marks it as a directory we missed. Each domain was assigned to the industry and city it appeared for most often. To stop the large cities dominating, we capped the sample at 30 domains per industry-by-city cell. Domains that appeared in three or more capital cities were flagged as national (15.4 percent of the usable sample) and kept, because a franchisee's site is still a small business site. Where the flag changes a finding we say so.

The final sample is 1,820 domains: 736 healthcare, 395 trades, 234 property and finance, 190 professional services, 149 legal, 82 technology, 34 other. By city: Melbourne 403, Sydney 378, Brisbane 373, Perth 214, Canberra 157, Adelaide 156, Hobart 100, Darwin 39.

Measurement. Each domain was fetched once on 3 September 2026 from a Melbourne IP address with a desktop Chrome user agent and no JavaScript execution. We requested the HTTP homepage to observe the redirect, the HTTPS homepage, robots.txt, the sitemap declared in robots.txt or /sitemap.xml, and one deliberately non-existent URL to test 404 handling. From the homepage HTML we parsed the title, meta description, H1 count, viewport tag, canonical, all JSON-LD blocks, tracking and chat scripts, platform fingerprints, phone numbers, ABN mentions, booking and quote calls to action, dollar figures, and superlative claims matched against a fixed pattern list ("Australia's best", "leading", "number one", "#1", "award-winning", "most trusted", "premier"). Time to first byte is the time from request to response headers on the HTTPS homepage request, single sample, uncached.

What "usable" means. 1,480 of 1,820 domains returned a 200 status with a real HTML body on their own domain. Rates in this paper are percentages of those 1,480 unless a table says otherwise. The remaining 340 are a finding in themselves and are reported in Finding 3.

Limitations. Single request per site, so time to first byte is indicative. No JavaScript execution, so a site that injects structured data client-side is counted as having none, which is also how most AI crawlers see it. The superlative-claim pattern is conservative and will miss claims phrased differently. Platform detection depends on fingerprints and 16.8 percent of sites could not be identified; most of those are custom builds. The two sampling sources have different biases and we publish both cuts. Core Web Vitals were not measured in this edition; they require a Google API key and will be added in 2027.

Finding 1: the schema gap

Most sites that have structured data have the structured data their platform ships by default. The most common types are WebSite (997 sites), SearchAction (711), WebPage (700), ImageObject (654) and Organization (641). Those are WordPress and Yoast defaults, and 40 percent of sites with any structured data have nothing else. The median site with structured data carries nine distinct types, almost all boilerplate.

Only 215 sites carry LocalBusiness or a subtype like Dentist, Plumber or LegalService, which is the markup that actually describes a local business: what it is, where it is, when it opens, what area it serves. 214 sites carry Person markup, which is how a named practitioner or founder becomes an entity search engines can connect to their work. 117 carry Service markup and 120 carry FAQ markup.

The platform cut makes the point. Squarespace sites have JSON-LD 100 percent of the time and Wix 96.8 percent, because the platform writes it. WordPress sites have it 87.6 percent of the time. Sites we could not identify, mostly custom builds, have it 44.6 percent of the time. Webflow sites, which are usually designer-built, have it 38.5 percent of the time. The builders that produce the most structured data produce the least useful structured data, and the custom builds that could carry anything carry the least.

By sector, property and finance is the worst at 41.0 percent with no structured data. Real estate agencies are the single worst industry at 48.4 percent, followed by property managers at 50.0 percent and cardiologists at 52.4 percent. Trades are the best sector at 14.9 percent, and pest control, roofing and plastic surgery are all under 10 percent.

Local business markup is where the national flag matters. Sites flagged as national carry LocalBusiness markup 30.3 percent of the time against 40.0 percent for local sites. A franchise template describes the brand, not the branch.

Finding 2: the AI-readable third

We built a composite from five things every AI crawler needs to read, place and trust a business: the site does not block AI crawlers in robots.txt, it carries structured data, it carries LocalBusiness markup, it returns a real 404 for a missing page, and it publishes a sitemap. A site passes only if it passes all five.

By city, Adelaide leads at 38.7 percent and Perth at 37.0 percent. Melbourne is the worst capital at 22.3 percent, dragged down by a 31.3 percent no-structured-data rate that is the highest of any city.

The pattern is not about money. Plumbers and dentists are not richer than orthopaedic surgeons and wealth managers. They are in industries where local search has been competitive for a decade, where the agencies serving them have industrialised the basics, and where the businesses themselves buy website packages that ship LocalBusiness markup. The referral-driven professions never had to compete in a search result and their websites show it.

Finding 3: sites we could not read

Nearly one in ten Australian business sites returns a block page to a request that does not execute JavaScript. That is mostly bot protection from the hosting provider, and it is the same response GPTBot, ClaudeBot and PerplexityBot receive. Legal sites block at 12.8 percent and trades at 12.7 percent; technology firms, who presumably understand what the setting does, block at 4.9 percent. The rate is the same whether the site came from search results (10.7 percent) or the map (8.9 percent). These businesses are not blocking AI on purpose. Their host is doing it for them, and most of them do not know.

Of the usable sites, 10.0 percent do not redirect HTTP to HTTPS, so a visitor who types the address without the prefix lands on the insecure version. The rate is 6.4 percent on WordPress and 22.9 percent on custom builds. Only 27.3 percent send an HSTS header.

Finding 4: speed

The absolute numbers carry the caveat above: one uncached request from one Melbourne connection. The distribution is the finding. A quarter of sites answer in under half a second and a quarter take more than three seconds, and the difference is almost entirely the platform.

Sites on the two most popular hosted builders are two to three times slower to respond than a typical WordPress site. Wix also ships the heaviest pages by far, a median of 676 KB of HTML on the homepage, three and a half times the WordPress median. Custom builds ship the lightest, a median of 70 KB, and are the least likely to carry structured data. Light and empty is not the same as fast and good.

Canberra and Darwin have the slowest median response, above 2,000 ms, and Hobart the fastest at 1,175 ms. Estate planning lawyers are the fastest industry at 585 ms, and dentists, psychiatrists and physiotherapists are all above 2,100 ms, which is where the hosted healthcare website builders live. National-flagged sites respond in a median of 1,298 ms against 1,661 ms for local sites, which is the franchisor's hosting budget showing.

Finding 5: the AI layer

Blocking is rare and, when it happens, it is a block-everything list rather than a considered choice. The sites blocking GPTBot almost all block Bytespider, CCBot, Amazonbot and Meta too, which is the signature of a copied robots.txt template. Almost nobody blocks the crawlers that actually deliver search referrals: six sites block PerplexityBot and two block OAI-SearchBot, the crawler ChatGPT uses to cite live pages. National-flagged sites block at 7.5 percent against 4.6 percent for local sites. Custom builds block at 8.0 percent, Squarespace at zero.

Seventeen sites, 1.1 percent, reference an llms.txt file. The AI-readiness conversation has not reached the Australian small business website.

Put Findings 2, 3 and 5 together and the position is: 5 percent of businesses have chosen to keep AI engines out, 9 percent are keeping them out by accident, and the remaining 86 percent are open but two-thirds of them are unstructured.

Finding 6: what sites claim about themselves

Trades make the most claims: 22.4 percent of trade sites carry a superlative and 17.4 percent carry self-declared rating markup. Healthcare, which is regulated on exactly this point, is at 6.3 percent and 6.6 percent. Dermatologists and cardiologists make no superlative claims at all in our sample. Property and finance is the boastful white-collar sector at 25.1 percent, driven by mortgage brokers at 46.4 percent and real estate agencies at 22.1 percent.

By platform, Webflow sites are the most likely to carry both a superlative (19.2 percent) and a self-declared rating (15.4 percent), which fits a designer-led build where the copy was written to impress. Squarespace sites are the least likely on both counts, 4.8 and 3.6 percent.

The self-declared rating finding deserves emphasis. 131 sites assert an aggregate star rating in their own homepage markup. 87 of those have no visible review widget on the page, so the number is not being pulled from Google or a review platform; it is typed into the code. Google's documentation says self-serving reviews are not eligible for rich results and the practice has been treated as spam since 2019. We found it on 8.9 percent of Australian business homepages, and on more than a quarter of pest control sites.

We removed a fabricated rating from our own site the week before this crawl. It had been placed by a previous build and nobody had checked it. We suspect most of the 87 are the same story: an SEO plugin or a template, not a decision. That does not make it less of a liability.

Finding 7: the tracking and conversion stack

A third of sites (33.5 percent) have no Google Analytics or Tag Manager we could detect. Only 1.8 percent have a Meta pixel without Google analytics, so it is not that they measure elsewhere; they do not measure. Professional services firms, accountants and architects mostly, are the worst at 54.9 percent unmeasured.

Conversion mechanics split cleanly by how the industry sells. Pest control (92.7 percent), physiotherapists (90.2 percent), cosmetic clinics (88.9 percent) and HVAC companies (86.4 percent) nearly all have a booking or quote action on the homepage. Real estate agencies (15.8 percent), cardiologists (19.0 percent), architects (21.4 percent) and IT companies (25.4 percent) mostly do not, because the enquiry happens through a portal, a referral, or a conversation. 70.9 percent of sites have a clickable phone link and 54.9 percent have a form on the homepage. 14.2 percent of homepages have fewer than 300 words of visible text.

Finding 8: the basics

The mobile viewport problem is solved: 98.6 percent of sites have the tag. Everything else is not. The heading problem is symmetrical, 18 percent with no H1 and 18 percent with several, and both are mostly theme defaults, a logo wrapped in an H1 or a slider that makes every slide one. The hosted builders that write the best structured data write the worst headings: Squarespace passes all five basics 15.7 percent of the time.

Sites with a sitemap declare a median of 86 URLs. 17.5 percent have fewer than 20 pages and 16.5 percent have more than 500. Sitemap freshness is better than the copyright footer suggests: 84.8 percent of sitemaps with a last-modified date were touched in 2026, but of the 54.9 percent of sites that show a copyright year, 22.5 percent show 2024 or earlier. The site is being updated; the footer is not.

Finding 9: search-visible sites versus everyone else

The two sampling frames describe two different populations. Businesses that appear in search results have been worked on: more structured data, more conversion prompts, and more claims about being the best. Businesses drawn from the map, which is closer to the true population of Australian small businesses, are plainer, slower to have been optimised, and less likely to boast. Read the search-derived figures as what the competitive layer looks like and the map-derived figures as what the average business looks like. Most published website studies only ever look at the first group.

Finding 10: national brands versus local businesses

Sites flagged as national, mostly franchise branches and multi-city practices, are faster and slightly better instrumented, and otherwise not better. They carry less local business markup than a single-location business, they block AI crawlers more often, and they make more superlative claims. A franchise template is a website for the brand that happens to have a suburb in the title.

Industry table

Percent of usable sites in each industry with 15 or more. AI-readable is the five-part composite from Finding 2.

[Each row links to the matching /industries/ page on publication.]

City table

[Each row links to the matching web-design-{city} and seo-{city} pages on publication.]

Melbourne's structured data gap is real and unexplained by platform mix: its WordPress share is average. Its sample is also the most healthcare-heavy of the three big cities, and Melbourne healthcare sites skew toward psychology and allied health, where structured data is weakest. Canberra makes the most superlative claims of any capital at 19.4 percent and blocks AI crawlers the most at 8.1 percent. Adelaide is the best-instrumented and the most AI-readable, and also the most likely to carry a self-declared star rating.

What to do about it

For a business owner, in the order that matters:

1. Check that a crawler can read you. Fetch your homepage with curl -A "GPTBot" or ask your host whether bot protection challenges non-browser requests. If it does, whitelist the major AI and search crawlers. This is a five-minute fix that 9 percent of businesses need and almost none know about.
2. Add local business structured data that says what you are. Not the WebSite and SearchAction blocks your plugin ships, but LocalBusiness or its subtype with your address, phone, opening hours and service area, and Person markup for the people whose names are on the door. 61.5 percent of sites are missing the first and 85.5 percent the second.
3. Remove any star rating you typed into your own code. If a rating is not pulled live from Google or a review platform, it is a liability, not an asset.
4. Return a real 404. Test it by visiting a made-up URL on your own domain. If you get your homepage with a 200 status, Google will eventually index nonsense.
5. Delete the superlative. "Melbourne's leading" earns nothing with Google, invites a complaint under Australian Consumer Law if you cannot substantiate it, and in healthcare is a breach of advertising rules.
6. Install analytics. A third of businesses cannot say where a single enquiry came from. Google Analytics is free.
7. Fix the H1 and the title. One H1 that says what you do and where. A title under 60 characters. Only 31 percent of sites get the five basics right and most of the failures are theme defaults nobody looked at.
8. If you are on Squarespace, Duda or Wix, know the cost. Two to three times the response time of a WordPress site, and on Wix, three times the page weight. They also ship the best default structured data and the worst headings. It may still be the right choice, but it is a choice.

For agencies and web designers: the platform defaults you inherit are doing your schema, and they are doing it badly. The single highest-leverage change across this whole dataset is real LocalBusiness and Person markup on the sites you already manage. The second is checking whether your hosting stack is blocking AI crawlers on your clients' behalf.

Frequently asked questions

Is this a study of "small businesses"? It is a study of businesses in 39 local-service industries found in search results and on the map in Australia's eight capitals. National brands and franchises are flagged and reported separately, not excluded. It does not cover e-commerce, hospitality or retail, which will be separate studies.

Why Bing and not Google? Google blocks automated querying, and we did not want to use a paid proxy service whose sampling we could not describe. The map-derived half of the sample corrects for Bing's ranking bias, and the two halves are reported separately.

Why no Core Web Vitals? Google's field data requires an API key and a quota we did not have for this edition. Time to first byte is a server measure and is not a substitute. The 2027 edition will add them.

Can I get the data for my industry or city? Yes. The dataset is published with the paper and includes every measurement for every domain. Contact us for a cut.

Will you tell me whether my site was in the sample? Yes, on request, along with your row.

Dataset and citation

• Dataset: au-smb-websites-2026-09.csv, one row per sampled domain with every measurement, plus the sample frame, the exclusion list, and the sampler, crawler and analysis code. Domains are included so the study can be replicated. No domain is singled out in this paper.
• Suggested citation: Durante, L. (2026). The State of Australian Small Business Websites 2026. Odin Digital Research. odindigital.com.au/research/state-of-australian-small-business-websites
• Charts on this page are embeddable with a credit link. Media enquiries: [email protected]

Back to Odin Digital Research