We checked 247 Sydney vet clinic websites. Most of them are slowly losing clients.

Method: an automated crawl of 247 clinic websites in Greater Sydney, plus Google PageSpeed lab data for each homepage. This is a study of the websites themselves: we did not measure bookings, phone calls, Google rankings or what AI assistants actually recommend.

Why we did this, and who it is for

Every agency says "your website has problems". We wanted to know which problems, how often, and whether the polished clinics do better than the rest, so we pointed a crawler at 303 vet-related listings and kept the 247 that turned out to be working Sydney vet clinic websites.

Vets are a good test case. Nearly every clinic depends on local search, and a pet owner looking for a new vet is often worried, usually on a phone, and wants two answers fast: is this clinic near me, and can I call them now?

At least 45 of the 247 clinics belong to a group: 13 are flagged as chains in our listing data, and 32 more turned out to be VetPartners clinics, 14 of them on one shared website platform (see section 5). The rest appear to be independent, so at most clinics the website is the owner's or practice manager's call, not a head office's.

A word on timing. Jobs and Skills Australia's Occupation Shortage List¹ has rated both vets and vet nurses as in shortage in NSW every year from 2021 to 2025. With staff that short, some clinics are booked out for weeks. If yours is one of them, nothing here is urgent, although your website can still answer routine questions (where you are, when you are open, where to go after hours) before they reach the front desk. The data matters most to clinics that need new clients:

• new and independent practices;
• mobile and home-visit vets (10 clinics in our sample list a service area instead of a shopfront address);
• emergency and after-hours clinics, which often see people who have never used them before.

We are not naming or ranking individual clinics. Terms with a dotted underline link to a plain-language glossary at the end; on a computer, hover over one to see its definition.

1. Speed: 8 seconds to appear on a test phone

Google's main loading measure is Largest Contentful Paint (LCP): the moment the biggest visible element, usually the hero photo or headline, appears on screen. Up to 2.5 seconds is "good"; over 4 seconds is "poor".

Only 9 sites out of 247 pass the "good" bar, and 80, roughly one in three, take more than 12 seconds to show their main content on the test phone.

Real visitors see something faster. On the 40 sites where Google also publishes real-visitor data (sites with enough Chrome visitors for Google to report on, so not a typical sample), three in four phone visits saw the main content within about 2.9 seconds at the typical site, against 8.1 seconds in the lab on the same sites. Treat the lab number as a fair comparison under the same tough conditions, not a stopwatch for your clients.

The usual cause is simple: the page asks the phone to download too much. The typical homepage weighs 3.0 MB. On three sites in four the phone also downloads a large amount of code the page does not use while it loads, and two in three still load an old code library that modern sites no longer need. That usually comes from a ready-made theme with a dozen add-ons on top, and it can usually be fixed without a redesign (fix 04 below).

2. The code Google reads: few say they are a vet

Structured data (schema markup) is a few lines of code, invisible to visitors, that tell search engines in a fixed format what the business is, where it is, when it is open and how to call it. For a clinic the correct type is VeterinaryCare. It costs nothing and is one of the few signals a small business fully controls.

Compare that with the clinics' Google listings. In our listing data, 241 of the 247 clinics (98%) have claimed their Google Business Profile, taking control of how Google describes them there. On their own homepages, only 31 (13%) tell Google, in the format it reads directly, that they are a vet clinic. Almost half of those 31 are clinics of one corporate group, VetPartners, whose shared template includes it; among the other 215 clinics, 17 (8%) do.

Two caveats. Google can and does work out that a site is a vet clinic from the page text and the Business Profile, and Google states plainly² that structured data does not guarantee any particular result. We also did not test what ChatGPT or Perplexity say about these clinics, so we cannot claim structured data changes their answers. It is simply the one place the clinic describes itself in a fixed format that machines can read without guessing, and 87% have not used it to say they are a vet.

3. AI search: what clinic websites tell AI tools

More people now put questions to AI assistants that they used to type into Google, although there is no good data yet on how many pet owners choose a vet that way. We checked three AI-related signals for each site, plus whether Google publishes real-visitor speed data for it. They do not measure whether ChatGPT or Perplexity actually recommends the clinic; that takes a different test, described at the end of this section.

How this connects to AI answers. When someone asks an assistant to recommend a vet nearby, it searches the web and reads what it finds. On most clinic sites, what it finds is thin:

• 87% do not say they are a vet clinic in the code Google reads;
• 57% get the main heading wrong;
• on 16 sites we found no phone number in the homepage text (possibly because a script draws it, Google runs scripts before reading a page; ChatGPT's crawlers do not^3,4).

None of this shows that an assistant skips a clinic; it means the site gives the assistant less to work with than it could.

Fourteen sites, thirteen of them with an identical list, name the training crawlers of OpenAI, Anthropic, Common Crawl and ByteDance (GPTBot, ClaudeBot, CCBot and Bytespider) and ask them to stay away, which asks those companies not to use the site to train future AI models. None of the 14 blocks the separate crawlers that ChatGPT, Claude and Perplexity use to search the web (OAI-SearchBot, Claude-SearchBot, PerplexityBot), so the effect on whether an assistant can find the clinic today is probably small; one of them also blocks ChatGPT-User, which ChatGPT uses to open a page when someone asks it to. Most probably inherited the block from a template rather than choosing it. Two other sites use an allow-list that shuts out every crawler it does not name, which keeps out the training crawlers and also PerplexityBot and Claude-User.

Read the CrUX number carefully. Google publishes real-user speed data only for sites with enough Chrome visitors to clear its threshold, and more than four in five Sydney vet websites are below it. That is not a traffic count: a busy clinic with loyal clients can still sit under the line. Treat it as a weak hint about scale, not a verdict.

We did not ask ChatGPT about each of the 247 clinics in this study. SEENMETRIC's Standard Audit does it for one clinic: up to 5 questions a pet owner might ask, put to ChatGPT, with the answers quoted in full, so you can see whether the assistant names your clinic and what it says. It is a snapshot of what the assistants said on the day, not a guarantee of what the next person will see.

4. The basics: headings, phone links, maps and images

The heading problem is mostly a theme problem. Page builders let a designer mark the clinic name, the tagline and the promo banner all as the main heading, or none of them. The page looks fine, so nobody notices for years.

The phone link is a one-line fix, yet on 44 homepages we found no tap-to-call link. On 16 of them we found no phone number in the homepage text at all, possibly because a script draws it or it is an image, which our crawler does not read. Images fare no better: on half the sites, at least every second image has no alt text, which is both an accessibility failure and a lost signal for image search.

5. Does a better clinic have a better website?

Google review counts are a rough sign of how established a clinic is. If established clinics invest in their sites, the ones with the most reviews should score clearly better. They do not: the map is the one clear gap, and on the PageSpeed score they are, if anything, a little behind.

Clinics with 200 or more reviews have the lowest median PageSpeed score (54, against 57 and 61), though that gap is not statistically significant, and they are no better at having exactly one main heading. Their one clear lead is the map: 35% show a Google map on the homepage, against 7% of clinics with under 50 reviews. On the other checks the differences between the groups are not statistically significant either.

We left one group out of that chart on purpose. 32 clinics in the sample link to VetPartners, a corporate group, or load files from its shared website platform, and 19 of them have 200 or more reviews. Fourteen of the 32 share one VetPartners website build, and those 14 are exactly the ones that say they are a vet; the other 18 run separate WordPress sites. 31 of the 32 have structured data, but our crawler found no Google map on any of their homepages. Counted in, they make clinics with many reviews look better at structured data than they are. Left out, 88% of clinics with 200+ reviews still do not say they are a vet in that code.

Our reading: website quality seems to depend more on the platform and whoever built the site than on how many reviews the clinic has.

The website still matters. A clinic with 200 reviews was usually built up over years through word of mouth, location and reputation. The site does its real work on people who have not heard of the clinic yet: the family that has just moved to the suburb, the first-time puppy owner, the pet owner whose old vet has closed or changed hands. This data cannot show how many of them pass over a slow or unclear site, only that a strong reputation today is not evidence the website is doing that job. For a newer or smaller clinic, that is good news: the bar is low, and the well-reviewed competitor down the road has not cleared it either.

6. The platform shapes the site

We tagged each site by the platform it runs on. WordPress powers 135 of the 247 and carries most of the worst performance.

The practical takeaway: on WordPress, a speed problem is most likely plugins and images, and usually fixable without a rebuild. On a hosted builder, speed is mostly out of your hands but the on-page basics are entirely in them.

What to check first if you want new clients

Do these five checks on your own phone, the way a new client would.

1. Run the test we ran. Open pagespeed.web.dev, enter your web address and read "Largest Contentful Paint" on the Mobile tab. Over 4 seconds puts you with the 83% of clinics in Google's "poor" band. Timing it yourself will usually look faster: your phone is quicker than the test phone and has probably stored your site from earlier visits.
2. Tap your phone number, ideally on an Android phone. If the dialler does not open, a new client has to copy the number by hand (fix 01 below). iPhones often turn a plain number into a link on their own, so a working tap on an iPhone proves little.
3. From the homepage, find your address, opening hours and what to do after hours. Each should be one tap away. If you close in the evening, does the site say where to go instead?
4. Search Google for "vet" plus your suburb. Scroll past the map listings to the ordinary results and find your website. Is the blue headline cut off? Do the headline and the line under it say what you do and where? Google usually builds both from your page title and description, which you can change, though it sometimes rewrites them.
5. Ask ChatGPT, or whichever assistant you use, "Which vet would you recommend in [your suburb]?" Does it mention you, and what does it say? Treat one answer as a sample of one: assistants vary their answers between people and between days.

What to fix first

These are in the order we would tackle them: first what a new client sees, then what only search engines and AI tools read. Times are typical for a web developer and vary by platform; on Wix and Squarespace you can do 01, 02, 03 and 06 yourself in the site editor. We did not measure the effect on calls or bookings, so each note says what the fix changes, not what it will do for bookings, and ends with a quick way to confirm it worked.

• 01 · 10 min
  Make the phone number tappableA new client on a phone can call you in one tap instead of copying the number. We found none on the homepage of 18% of sites; the fix is a tel: link. Check: tap the number on an Android phone; the dialler should open with the number filled in.
• 02 · 30 min
  Put your hours, address and after-hours plan on the homepageA worried pet owner wants to know whether you are open now, where you are and where to go if you are closed. Show all three near the top of the homepage, and name the emergency clinic to call out of hours. We did not measure how many sites do this. Check: ask someone who has never used your site to find all three on their phone within a minute.
• 03 · 1 hour
  Fix the page title, description and main headingThe title is usually the blue line people see in Google, so it should name the clinic and suburb in about 60 characters; the description is the line under it. Each page also needs one main heading. 57% of homepages get the heading wrong. Check: a week or two later, search Google for your clinic's name and read the blue line and the line under it; Google can take that long to pick up changes.
• 04 · 1 day
  Make the homepage load faster on a phoneCompress images, load scripts later and remove plugins you do not use; on WordPress the last often helps more than any caching tool. Aim for "Largest Contentful Paint" under 4 seconds on the mobile test at pagespeed.web.dev, and under 2.5 for "good". Check: run the test two or three times after the change; a single run can vary by a few points.
• 05 · 30 min
  Describe the clinic to Google in codeGives Google your clinic type, address, phone and hours in the format it reads directly, matching your Business Profile. Some AI tools may read it too; we did not test that. 87% of homepages do not say they are a vet in this code. Technically, this is structured data typed as both VeterinaryCare and LocalBusiness. Check: paste your homepage address into validator.schema.org; it should list a VeterinaryCare item with your address and phone.
• 06 · 1 hour
  Describe every image, and state its size in the codeScreen readers and Google read the description (alt text). Stating each image's width and height lets the browser keep space for it, so text does not jump. On about half the homepages, half or more of the images have no description, and on 42% half or more have no stated size. Check: in the pagespeed.web.dev report, the warnings about images without [alt] attributes or without explicit width and height should be gone.
• 07 · 15 min
  Check what your robots.txt tells AI crawlersFourteen sites ask AI training crawlers to stay away in robots.txt, most with the same copied list. That affects future AI training, not ChatGPT's search results. Two more shut out every crawler they do not list, which also keeps Perplexity out. If yours is one of these, decide whether you meant it. Check: open your web address followed by /robots.txt in a browser and read the lines that start with User-agent.

How we measured, and what to be careful with

• Sample. 247 veterinary clinic websites in Greater Sydney (the ABS definition, which includes the Central Coast; 8 clinics are there or on the Hawkesbury fringe), taken from Google Business Profile listings. We started with 303 listings. 25 sites could not be checked: 11 did not respond and 14 refused our crawler or returned an error page (mostly HTTP 403), so a few better-protected sites may be missing. We removed 27 more that were not a Sydney vet clinic's own website: directory pages, a university, a shelter and a charity, a pet store and a pet food shop, suppliers, a laboratory and other businesses, a corporate head-office site, services for other vets, online-only services, one practice in Perth and one in Helensburgh, a duplicate and our own test entry (the full list is in our research files). We also left out four listings with no working clinic website: two showed only a web host's "account suspended" or "website unavailable" page, one was an empty placeholder page, and one said the practice had closed. Unless a chart says otherwise, percentages are shares of these 247 sites.
• Crawl. For each site we fetched the homepage with a desktop and a mobile user agent, robots.txt, the sitemap, llms.txt, and up to 15 internal pages. Everything was parsed from the HTML as served; we did not execute JavaScript, so anything a script draws (a phone number, a map) can be missed. Inner-page checks cover the 241 sites where at least one inner page loaded (the broken-link check covers the 245 where we tried at least one inner link). Two homepages have under 50 words of text (one is a single page with little more than an address and phone number), so they fail most homepage checks; that can move any share by at most one point. A contact-page check proved unreliable and is not used.
• Performance. Google PageSpeed Insights API, mobile strategy, one run per homepage. These are lab numbers on a throttled connection, not real-user timings, and a single run has noise of a few points. Bands and medians across 247 sites are robust; any single site's number is not. On the 40 sites with Google real-visitor data, at the typical site three in four real phone visits saw the main content within 2.9 s, against 8.1 s in the lab, and the lab ranking only loosely matched the real-world one.
• Listing data. Review counts and the claimed-profile and service-area flags come from the Google Business Profile listing data we built the sample from, not from our crawl. "Claimed" means someone has taken control of the listing; it says nothing about how complete or current it is. The chain flag comes from that list and misses groups that keep a clinic's local name: it flags none of the 32 VetPartners clinics, which we found from links and shared files on their homepages. Other groups may be missed the same way.
• Known gaps. The map check only detects Google Maps iframes. Structured data is read from the homepage only, in JSON-LD, the usual format. 33 homepages also carry microdata, an older format; in the 240 homepages we saved, none of it names a business or a vet, so it could only add four sites to the first step of the funnel. Platform labels come from our listing data, checked against a fresh fetch of each homepage on 27 September; where they disagreed we used what the homepage showed, and 24 sites could not be identified. robots.txt files were re-read the same day with a parser that follows the standard grouping rules. "Describes a business" counts LocalBusiness, Organization and their subtypes. We did not check backlinks, rankings, Google Business Profile completeness, or what any AI assistant says about any clinic.
• No individual clinic is identified. All figures are aggregates.

Thresholds for LCP and CLS follow Google's published Core Web Vitals guidance (LCP good up to 2.5 s, poor over 4 s; CLS good up to 0.1). More than 200 KB of unused code, titles over 60 characters and inner pages under 300 words are our own cut-offs. We also checked the server's security settings (security headers); most sites miss some of them, but they do not change how a new client finds or contacts a clinic, so we leave them out. Structured data types from schema.org. Data collected 26 September 2026; platform labels and robots.txt re-checked, and figures recalculated, on 27 September 2026 after removing entries that were not working clinic websites. Where a site did not respond to the platform re-check, we tried again and then used a copy of its homepage saved the same day; for four sites with no copy we kept the listing label where our 26 September crawl agreed with it.

Sources

1. Jobs and Skills Australia, Occupation Shortage List: veterinarians and veterinary nurses in NSW, 2021 to 2025.
2. Google Search Central, Local business (LocalBusiness) structured data: adding structured data does not guarantee that any search feature will appear.
3. Vercel, The rise of the AI crawler, 17 December 2024: in a study of crawler traffic, none of the major AI crawlers, including OpenAI's and Anthropic's, ran JavaScript; Google's crawler does.
4. Edgecomet, How GPTBot and ChatGPT-User handle JavaScript, January 2026: ChatGPT-User, the bot that opens pages for ChatGPT's answers, fetched the HTML and ran no scripts; GPTBot, which collects training data, downloaded script files but rarely ran them.

Glossary

The technical terms used in this report, in plain language. Where a term has a threshold, it is Google's unless the method section lists it as ours; the 60-character title length is a common estimate of what Google shows.

Main heading (H1)

The main headline of a page, marked in the page code as "heading level 1". Visitors see it as the biggest text; Google reads it as the one-line summary of what the page is about. Each page should have exactly one. A page can look fine and still have none, or three.

Page title

The text that appears in the browser tab and as the blue clickable headline in Google search results. It is set in the page code, not on the visible page. Google shows roughly the first 60 characters, depending on screen width, and sometimes rewrites the title.

Meta description

The one or two sentences Google often shows under the blue link in search results. It does not affect ranking directly but decides whether someone clicks. Without one, Google writes its own snippet from the page text, and it sometimes does that even when a description exists.

Canonical tag

A line in the page code that says "this is the official address of this page". It stops Google treating www and non-www, or http and https, versions as separate duplicate pages competing with each other.

Alt text

A short written description attached to an image in the code ("Golden retriever on the examination table"). Screen readers read it aloud for blind users, and Google uses it to understand and rank the image. Without it an image is invisible to both.

tel: link

A phone number wrapped in a link that starts with "tel:". On a phone, tapping it opens the dialler with the number filled in. A number written as plain text has to be copied or retyped on many phones; iPhones often detect plain numbers on their own.

Structured data (schema markup)

A small block of code, invisible to visitors, that describes the page in a fixed format search engines, and possibly some AI tools, can read without guessing: type of business, name, address, phone, opening hours, location. "LocalBusiness" and "Organization" are the generic types; more specific types exist for most industries. It is usually written in a format called JSON-LD.

VeterinaryCare

The specific structured data type for a veterinary clinic, defined at schema.org. It tells Google, and possibly some AI tools, not just "this is a business" but "this is a vet". At schema.org it sits under MedicalOrganization, not LocalBusiness, so on its own it cannot carry opening hours or map coordinates. The usual approach is to type the clinic as both VeterinaryCare and LocalBusiness in the same block.

LCP, Largest Contentful Paint

How long after a visitor opens the page the largest visible element (usually the hero photo or main headline) actually appears. Google's own metric for perceived loading speed. Good: 2.5 seconds or less. Poor: over 4 seconds.

Core Web Vitals

Google's three official user-experience metrics: LCP (loading speed), INP (responsiveness to taps) and CLS (how much the layout jumps around while loading). They are a confirmed, if modest, ranking factor.

PageSpeed Insights

Google's free tool that loads a page on a simulated mid-range phone over a slow connection and grades it from 0 to 100, with a breakdown of what slows it down. The score is a lab measurement; it is consistent between sites but not the same as what a visitor on fast Wi-Fi experiences.

CrUX, Chrome User Experience Report

Google's public database of real speed measurements collected from Chrome users. A site appears in it only once enough Chrome users have visited to meet Google's threshold. Being absent means Google does not have enough qualifying measurements for the site; it is not a traffic count and says nothing about where visitors come from.

Unused JavaScript

Script code the browser downloads but that did not run while the page loaded in Google's test; some of it may only run later, when a visitor taps something. Typically comes from plugins, themes and tracking tools that load everywhere "just in case". It costs download time on every visit.

jQuery

A JavaScript library from 2006 that made building interactive pages easier when browsers were inconsistent. Browsers no longer need it, but many WordPress themes and plugins still load it. Its presence is a reliable sign of an older, heavier setup.

robots.txt

A plain text file at the root of every website where the owner tells automated crawlers (Google, Bing, AI companies) which parts of the site they may read. Blocking a crawler here asks it to keep the site out of whatever that crawler feeds; AI companies run separate crawlers for model training and for search.

llms.txt

A newer, optional plain-text file that summarises a website for AI assistants: what the business is, its key pages, how to cite it. Not used by Google Search; adopted by some AI tools.

Google Business Profile

The free listing that appears in Google Maps and in the box on the right of search results, with the clinic's address, hours, photos, star rating and reviews. Formerly called Google My Business. It is separate from the website but the two reinforce each other.