For Australian clinics, cookies are not the problem. What your tracking sends, and what it is used for, is.
- Cookies are not what changed. Chrome kept third-party cookies, and Safari and Firefox already restrict them.
- The change that matters is regulatory. On 11 June 2026 the Privacy Commissioner found that Medmate and Monash IVF breached the Australian Privacy Principles through tracking pixels. Collecting sensitive information needed consent, and a cookie pop-up and a general privacy policy did not provide it.
- The OAIC’s advice goes further than “get consent”. As reported by iTnews, it asks organisations to collect the minimum through pixels and says sites that handle sensitive information should not run them at all.
- The practical response has three parts. Audit what your pixels send, keep measuring results separate from building ad audiences, and rebuild attribution around the calls and bookings you hold yourself. This applies to health service providers of any size.
Do this week
- List every tool on your website that sends data to another company. The audit section below shows how.
- Find out which of them fire before a visitor makes a consent choice, and get privacy advice on whether to pause them.
- Ask whoever runs your marketing for the full tag list and a plain account of what each one sends.
This is general information, not legal advice. Where we report a regulator’s finding we say so, and where something is our own analysis we say that instead. This guide is written for clinics and practices that market to patients, not for pharma teams marketing to healthcare professionals. Take advice from a privacy lawyer before you change your tracking.
Are third-party cookies going away? What changed in 2025 and 2026
No. Google has kept third-party cookies in Chrome and, on 17 October 2025, retired ten of its Privacy Sandbox technologies, including the Topics and Protected Audience APIs that were meant to replace them. Safari has blocked third-party cookies by default since 2020, and Firefox partitions them. For a clinic, the bigger change is regulatory.
What is cookieless tracking?
Cookieless tracking is measuring marketing results without relying on third-party cookies, using first-party data, server-side events or aggregated reporting. It describes a method, not a privacy standard: a cookieless setup can still send health information to an advertising platform.
Which clinic channels depend on cookies?
Most of the channels that bring in new patients, starting with Google search ads, sit at the low end of this scale. The table is our assessment, not a published ranking.
| Channel | Dependence on third-party cookies | Why | What to check |
|---|---|---|---|
| Google search ads | Low | Ads match what someone searches for, not a profile built from cookies | Whether your conversion actions count calls and bookings, not just clicks |
| Local search and Google Business Profile | Low | Google reports actions such as calls and direction requests itself, with no tag on your site | What your profile reports for your account |
| Organic search | Low | Visits are measured by your site analytics, not cross-site tracking | What your analytics tool collects and shares |
| Meta prospecting ads | Medium | Targeting leans on Meta’s own account data, but conversion signals usually come from your pixel | Whether Meta has classified your data source as health and wellness |
| Retargeting | High | It depends on recognising a returning visitor and building an audience from their site behaviour | Whether you have valid consent and platform eligibility |
| Display and programmatic | High | Cross-site identifiers are central to how these audiences are bought | Whether you need it at all |
How the main methods compare
| Method | How it works | What it discloses | Fit for a clinic |
|---|---|---|---|
| Third-party cookies | A cookie set by an ad or analytics company recognises a visitor across different sites | Browsing activity linked to the visitor by the cookie owner | Shrinking in Safari and Firefox, still present in Chrome, and not the main risk |
| Browser pixel or tag | Code on your page sends page views and events from the visitor’s browser to a platform | Page address, page title, event names and any identifiers the tag is set to send | The setup the OAIC scrutinised, so it needs the closest review |
| Server-side events | Your own server sends events to the platform instead of the browser | Whatever you choose to send, which can be the same data a pixel carries | More reliable delivery, but the same disclosure question |
| First-party records | Calls, enquiries and bookings are kept in your own systems with the source recorded | Nothing goes to an ad platform unless you choose to send it | The foundation of a clinic’s measurement |
The three layers of clinic tracking: recognise, transmit, use
What is a tracking pixel?
A tracking pixel is a small piece of code on a web page that sends information about a visitor’s activity to another company, usually an advertising or analytics platform.
Tracking does three separate things, and different rules govern each. Most confusion comes from treating them as one.
| Layer | What happens | Examples | Who sets the rules |
|---|---|---|---|
| 1. Recognise | Working out that two visits belong to the same person | Cookies, click IDs, device IDs, hashed email or phone | Browsers and the visitor’s settings |
| 2. Transmit | Sending information about what a visitor did to another company | Meta and TikTok pixels, Google tags, call-tracking scripts, server-side events | The Privacy Act, plus each platform’s data terms |
| 3. Use | Doing something with that information | Measuring results, optimising bids, building audiences, retargeting | Platform advertising policies, plus direct marketing rules in the Privacy Act |

Say a visitor reads a page called /anxiety-counselling on a psychology practice’s site. A cookie or other identifier recognises them (layer 1). A pixel sends the page address, which names the service, to an ad platform (layer 2). The platform shows them an ad about it later (layer 3).
Cookie news only touches the first step. Medmate shows why. Its Meta pixel had Advanced Matching enabled, which sends hashed contact details so the platform can match a visitor to a profile. Its TikTok pixel sent full page URLs that sometimes disclosed the condition or medication a visitor was researching. Neither needed a third-party cookie.
This three-layer split is our framing, not the regulator’s. We use it because it keeps three questions apart: can we recognise people, what are we disclosing, and what are we doing with it.
OAIC tracking pixel decisions: what Medmate and Monash IVF mean for clinics
Medmate and Monash IVF in brief
The Privacy Commissioner’s determinations against Medmate Australia Pty Ltd ([2026] AICmr 41) and Monash IVF Pty Ltd ([2026] AICmr 40) are dated 11 June 2026. They followed an OAIC review of 50 health provider websites, and the investigations began in December 2024. Hamilton Locke’s analysis cites the determinations paragraph by paragraph.
| Detail | Medmate Australia | Monash IVF |
|---|---|---|
| Type of provider | Online telehealth and prescription platform | Fertility services and treatments |
| Period of the conduct | April 2021 to 9 December 2024 | July 2012 to 9 December 2024 |
| Pixels | Meta pixel (with Advanced Matching enabled) and TikTok pixel | Up to seven at various times, including Meta pixels on specific fertility pages, plus Google Ads and Google Analytics |
| What they sent | Hashed contact details through Meta, and full page URLs through TikTok that sometimes revealed a condition or medication | Page-level data tied to pages on egg freezing, donation, health checks and IVF consultations, plus Custom Audience lists with names, email addresses, phone numbers and gender uploaded to Meta |
| How it was used | Retargeting on Meta and TikTok, including people who had bought particular medication or sought weight management services | Retargeting on Meta based on browsing behaviour, age, gender and prior page interactions |
| Consent and assessment | A cookie pop-up added in the final weeks of the investigation period, found insufficient. No privacy impact assessment before deploying the pixels. | Required consent not obtained. No privacy impact assessment before deploying the pixels. |
| Principles considered | APP 3.3 (collecting sensitive information), APP 5.1 (notice) and APP 7 (direct marketing) | APP 3.3 (collecting sensitive information), APP 5.1 (notice) and APP 7 (direct marketing) |
| What happened next | Stopped using all tracking pixels during the investigation and reviewed its privacy policy, consent mechanisms and agency contracts | Disabled unused pixels, audited its Meta pixel, began a privacy impact assessment and began work on consent functionality |
Press coverage reports that both providers received declarations that they must not repeat or continue the conduct. They were also ordered to destroy, where legally permitted, the sensitive information held in their pixel provider dashboards and to confirm that to the OAIC within 90 days. The same coverage puts the two sites’ combined traffic at well over a million visitors a year.
Five findings, translated for marketers
| What the OAIC found | What it means for your marketing | Principle involved |
|---|---|---|
| Visiting a health service website generates health information, which is sensitive information. Browsing pages about specific conditions, medications or treatments was enough. | Sensitivity attaches to the page a visitor reads, not only to forms they submit. A visitor who never fills in anything can still generate sensitive information. | APP 3.3 (collection of sensitive information) |
| Both providers argued they could not identify the people targeted. The Commissioner rejected that: where an organisation can single someone out and treat them differently, including by retargeting, the information is personal information even without a name. | An anonymous-looking ID or a hashed email does not settle the question. | Definition of personal information |
| The providers held the information even though it flowed to Meta’s and TikTok’s servers, because they had control over what was collected. | “It’s Meta’s data, not ours” is not a defence. The clinic that commissioned and configured the tool carries the responsibility. | APPs 3 and 5 |
| A cookie pop-up and a privacy policy were not enough. Consent must be informed, voluntary, current and specific, and given with capacity, and notice is expected at the point of entry. | Pixels need their own consent and their own notice. A cookie banner covers neither unless it says what the pixel does. | APP 3.3 and APP 5.1 |
| Using sensitive information to retarget people was direct marketing without consent. The most significant harm identified was the downstream use, not the collection. | Building audiences from health-site browsing carries the heaviest burden of the three layers. | APP 7 (direct marketing) |
The OAIC’s review of 50 health provider websites found that pixels were commonly installed by marketing agencies or web development teams as part of advertising campaigns, often because of a belief that pixel data is anonymous.
The 2026 OAIC community attitudes survey, as Hamilton Locke reports it, found that 91% of Australians consider targeted advertising based on sensitive information neither fair nor reasonable. For a clinic, that is a patient-trust problem before it is a regulatory one.
What the decisions did and did not say
- They did not ban pixels outright. The OAIC’s guidance says the Privacy Act does not prohibit the technology, and asks for data minimisation and regular review. Its reported advice for sites that handle sensitive information is stronger: do not run them at all. Many clinic sites fall into that group.
- They examined pixels. We have not seen the Commissioner address server-side conversion tools. Our reading is that the logic turns on what is disclosed rather than how it is delivered, so moving events to a server does not by itself answer the consent question.
- No fine appears in the coverage we reviewed. The outcome was declarations and orders to destroy data. That does not tell you what the next case would look like.
- Google Analytics was not the focus. It was among Monash IVF’s tools, but the coverage centres on the social media pixels. Commentators say the reasoning reaches any tool that sends data to a third party in a way that lets someone be singled out.
“We’re a small clinic”
The Privacy Act still applies to you. The OAIC explains that it covers any health service provider regardless of turnover, which includes medical practitioners, dental and allied health practices and chiropractors. State and territory health records laws, including in Victoria, New South Wales and the ACT, can add further obligations.
Why US advice about HIPAA does not carry over
Most guides on this topic are written for US healthcare and revolve around HIPAA and business associate agreements. None of that is Australian law. Here the Privacy Act, the Australian Privacy Principles and the OAIC set the rules.
How to audit the tracking pixels on your clinic website
You cannot assess what you cannot see. In one case an external web design company had left 50 tracking pixels on a provider’s site without its knowledge, and the OAIC’s review found agencies and web teams often installed pixels without considering privacy obligations.
1. Find every tool
Check these places:
- The tag manager container (for most clinics, Google Tag Manager): every tag, its triggers and the pages it fires on
- Scripts hard-coded in the site’s header, footer and theme files
- Plugins for analytics, chat, forms, pop-ups and social feeds, which sit in your website build and are easy to forget
- Booking widgets and patient forms hosted by another system
- Heatmap and session replay tools, which can record what visitors type into forms
- Call-tracking scripts, including the code that swaps phone numbers
- Embedded video and maps
- The ad accounts themselves: Meta Events Manager data sources, Google Ads conversion actions, and any TikTok, LinkedIn or Pinterest tags
The last item is the shortcut. If an ad account lists a pixel, something is sending data to it, even if nobody remembers installing it.
A 10-minute test
- Open your site in a private browser window and do not accept or decline any consent notice.
- Right-click, choose Inspect, and open the Network tab.
- Reload the page, then click through a service page, the booking page and a contact page.
- Look for requests going to domains owned by advertising and analytics companies such as Meta, TikTok and Google.
- Click a request and read what it carries: the page address, page title, event names and any identifiers.
- Write each one down against your inventory, with the page it fired on.
The free Meta Pixel Helper and Google Tag Assistant browser extensions list the Meta and Google tags on a page. They will not show every other vendor, so use them as a shortcut, not a substitute.
Keep one row per tool in your inventory: vendor, where it is installed, who installed it and when, pages it fires on, what it can see, purpose, whether it fires before consent, your decision, owner and review date.
2. Read the payload
| Page type | What it can reveal | What to check |
|---|---|---|
| Home page | Which clinic, and its specialty | Whether the clinic’s name or focus implies a sensitive service |
| Service or condition pages | The treatment or condition of interest | URL path, page title, query parameters and which tags fire |
| Booking and enquiry pages | Intent to be treated, often with contact details | Form field names and values, button labels, event names and identifiers |
| Confirmation pages | That a patient booked, sometimes which service | URL, event names and parameters carrying the service chosen |
| Blog and education pages | Interest in a condition | URL and page title |
Look for page addresses (including everything after the question mark), page titles, referrers, event names, form fields, button text, email and phone (including hashed versions), and device and IP data. Clean parameters are not enough on their own, because the path still names the service. A page at /treatments/ivf tells the platform what the visitor read.
3. Classify each tool
Ask three questions of each tool:
- Do we need this to measure something we actually use?
- Could what it sends reveal or imply that someone is seeking health care?
- Could the recipient single the person out?
Then decide:
- Remove if the answer to question 1 is no.
- Keep if it is yes to question 1 and no to both 2 and 3. Call tracking that sends nothing to an ad platform often lands here.
- Modify if it is yes to question 1, and 2 or 3 is yes, but you can cut the payload until both become no.
- Gate (fire only after valid consent) or remove if it still cannot get to no.
Expect the answer to question 2 to be yes for any tool that sends page-level data to an advertising platform on a health site, because the OAIC found that browsing a health service website generates health information. For those tools the choice is gate or remove, not tweak.
How to modify a tool you keep
These are options to put to your adviser, not a compliant configuration. In your tag manager and ad settings:
- Replace the page URL and page title the tag sends with neutral values, so a treatment page is not named.
- Strip query parameters from the page address.
- Switch off matching features such as Meta’s Advanced Matching, which sends hashed contact details.
- Use neutral event names with no condition, treatment or service in them.
- Send no form field values.
- Fire conversion tags on the confirmation page only, and only for visitors who have consented.
How to switch a pixel off safely
- Pause or remove the tag in the tag manager, or deactivate the plugin that loads it.
- Reload your site in a private window and confirm the pixel no longer fires.
- Leave the data and audiences in the ad account untouched until you have advice on what to destroy. Deleting first removes the record of what was collected.
- Remove the pixel’s access in the ad account, including any agency or partner access you no longer need.
- Update your inventory and your privacy policy.
- Scan again after the next plugin or theme update, because updates can bring tags back.
A worked example
This is a hypothetical dental clinic, built to show the reasoning. It is not a client case.
| Tool | Where it runs | What it could see | Decision | Reason |
|---|---|---|---|---|
| Google Analytics 4 | Every page | Page addresses and titles, including treatment pages | Modify, then gate | Reduce what it collects and fire it only after consent |
| Meta pixel | Every page | Treatment page visits, booking clicks and identifiers | Gate or remove | It is used for retargeting, the use the OAIC treated as most harmful |
| Google Ads conversion tag | Booking confirmation page | The booking event and, if configured, contact details | Modify | Use a neutral event name, send no service name or parameters, and check which identifiers it sends |
| Call-tracking script | Every page | Which source a caller came from, and recordings if enabled | Keep, then review | Check what it collects and whether it records calls |
| Heatmap tool | Every page, including forms | Clicks, scrolling and possibly typed text | Remove | Nothing in the measurement plan depends on it |
| Booking widget analytics hook | Booking domain | Page address and the service chosen | Check before enabling | It is another tag for the inventory, so confirm what it sends |
Cookie banner vs pixel consent: what valid consent looks like
The OAIC found that Medmate’s cookie pop-up, added in the final weeks of the investigation period, was not sufficient. It did not mention pixels, did not name Meta or TikTok, and did not explain that information was being disclosed to those platforms for advertising matching and retargeting.
For sensitive information, consent has to be informed, voluntary, current and specific, and given with capacity. A privacy policy alone did not meet the notice requirement either, and a banner or pop-up at the point of entry is expected.
That gives you three separate checks: consent to collect sensitive information (APP 3.3), notice at the point of collection (APP 5.1), and consent for using it in direct marketing (APP 7). These are the design principles worth putting to your adviser:
- Show a separate notice on arrival, not only the cookie banner.
- Make it specific. Say what is collected (including that the pages visited may indicate health services), who receives it, what it is used for, and name the platforms.
- Load nothing before the choice. The tag should not fire and send first, then ask afterwards.
- Record the choice and honour a change of mind.
- Name pixels in your privacy policy, not only cookies, and run a privacy impact assessment before adding a pixel. Neither provider did one. The OAIC publishes a guide to doing so.
Two practical points follow. First, consent changes your reporting: platforms will only see the visitors who agree, so reported conversions will fall, and the reconciliation ratio later in this guide exists for exactly that. Second, Google’s Consent Mode is not consent. It carries the visitor’s choice to Google’s products, and it is not evidence that your notice was valid. That is our interpretation.
AHPRA’s advertising rules are a separate matter. They govern what your ads say, not what your website collects.
Should your clinic keep retargeting? Measurement vs audience-building
Most tracking questions on a clinic site are really two. Do you need to know what happened? That is measurement. Do you need to show ads to people who visited? That is audience-building. As Hamilton Locke reads the determinations, the most significant harm the OAIC identified was not the collection itself but the downstream use: sensitive information feeding advertiser targeting. That is why this guide treats the two as separate decisions.
Pracxcel runs paid campaigns for clinics, so weigh this section with that in mind. The test below is built so that it can come out against retargeting.
Where we would start, for your lawyer to test
This is our analysis of a sensible default for most clinics, not legal advice.
- No third-party advertising pixels on condition-specific or treatment-specific pages.
- Conversion measurement limited to a neutral event on the confirmation page, fired only after valid consent, with no service name, parameters or identifiers you do not need.
- Calls, enquiries and practice-software bookings as the primary measurement.
- No audience-building from site visits, and no patient-list uploads, without written legal sign-off.
- A named owner for every tag, and a scan every quarter.
The decision tree
- Do you provide a health service and hold health information? If yes, the Privacy Act applies, whatever your size.
- Does any third-party tag fire before consent? If yes, gate or remove it first. Nothing else matters until this is settled.
- What is each tool for?
- Measurement only. Cut the payload, and check whether a first-party method answers the question without sending anything to an ad platform. Keep a platform conversion only if you need it for bidding, and put it behind consent.
- Audience-building. You need pixel-specific consent, platform eligibility and a break-even case. If any of the three fails, drop it.
- Where does the conversion happen? Phone, hosted booking, embedded booking or practice software. Route to the booking section below.
- Reconcile monthly against confirmed new patients.

Google and Meta have their own rules for health data
A tool can be lawful with consent and still be blocked by the platform, and the reverse is also true. Check both.
- Google. Its policy on health in personalised advertising restricts using health conditions to personalise ads. Industry coverage notes it does not restrict conversion tracking itself. Separately, industry coverage reports that accounts in sensitive categories including healthcare can be labelled “Eligible (Limited)”, which removes remarketing lists and Customer Match. Sources disagree on how much general remarketing survives, so check the policy status in your own account.
- Meta. It classifies some advertisers and data sources as health and wellness. For a restricted source, it limits URL parts and custom parameters, audiences can shrink or stop working, and features such as advanced matching can become unavailable. One industry summary reports that Meta flagged custom conversions with names such as “appointment booked”. You can check how your data source is classified in Events Manager.
Our hypothesis is that much of the signal loss health clinics noticed since 2025 came from classification and policy rather than from cookies. We cannot quantify it, and classification varies by site, but it is the first place to look when reporting drops.
A break-even test for retargeting
Keep retargeting only if:
incremental new patients per month × first-year margin per new patient > monthly spend + monthly cost of consent and tooling
The risk sits outside the formula. If you cannot get valid consent and platform eligibility, you never get to run it.
The hard variable is incremental. Platform-reported conversions count people who booked after seeing an ad, including people who would have booked anyway. A campaign that reports 20 bookings has not shown it caused 20. Some of the people retargeting reaches would have booked regardless, and the platform cannot tell you how many.
A simple test is to pause retargeting for long enough to cover your usual gap between enquiry and first visit, and watch confirmed new patients in your practice software, allowing for seasonality. If the numbers do not move, you have learned something no platform report would have told you.
If they do, you have a basis for putting it back on a consented footing. This is a method, not a result. Before you run it, work out what a new patient is worth.
Two special cases
Patient lists. Uploading patient names and contact details to an ad platform is a different activity from a website pixel, and it was part of the finding against Monash IVF, which uploaded Custom Audience lists with names, email addresses, phone numbers and gender to Meta.
If your reactivation work uses lists this way, get legal advice on the mechanism before the next campaign. Email and SMS to your own patients raise different rules, which also need checking.
Sensitive specialties. Fertility, sexual and reproductive health, mental health and oncology are the obvious cases, and Medmate’s examples included contraception and UTI treatment. If your site is mainly about one of these, as with many gynaecology and obstetrics practices, even a generic tag carries more inference.
Cookieless measurement for clinics: tracking new patients without third-party pixels
If you remove or gate third-party pixels, you can still tell which marketing works. You lose detail. Work up this ladder in order.
| Step | What you measure | How | What it can’t tell you |
|---|---|---|---|
| 1. Calls | Which channel produced the call | Call tracking that swaps the number by source, capturing counts and duration, with a separate decision on recording | Whether the caller booked, and the script itself is a tool to inventory |
| 2. Enquiries and booking requests | Source of each enquiry | Store the source (UTM values and the ad click ID) in hidden fields in your own enquiry system, not in an ad platform | Visitors who did not enquire |
| 3. Bookings in practice software | Whether an enquiry became a patient | Match enquiries to bookings in your practice management software, such as Best Practice, Medical Director or Cliniko | Anything beyond what staff record consistently |
| 4. Intake question | Where new patients say they heard of you | A fixed-choice “how did you hear about us?” in your practice software | Self-reporting is imperfect, so use it as a cross-check, not the truth |
| 5. Offline conversion import | Feeding confirmed bookings back to Google Ads | Send the booking event after it happens, matched by the click ID captured at enquiry rather than by email or phone where you can | A click ID still tells Google that a particular ad click led to a booking, so put it on the same consent footing. Sending email or phone discloses patient identifiers directly. |
| 6. Monthly reconciliation | Whether platform reports match reality | The ratio below | Why a gap exists, which the diagnostic table later in this guide helps with |
Keep UTM values free of treatment names. Any tag on the page can read them, and Meta removes URL parts after the domain from restricted data sources anyway.
Local search adds its own first-party signals. Google reports actions such as calls and direction requests on your business profile, so local SEO and Google Business Profile work can be measured without a pixel on your site. Check what your profile reports for your account.
Count new patients separately from existing ones. Search ads can attract people who are already patients, and you do not want to credit a campaign for a patient who would have booked anyway. Where your practice software records attendance, count attended new patients rather than bookings, because show rates can differ by channel.
Meta and Google lead forms keep the form inside the platform, which avoids a website pixel but still hands patient details to that platform, so treat them as a data flow to check, not an exemption.
The reconciliation ratio
Divide the conversions a platform reports in a month by the new patients your practice software attributes to that channel in the same month. Track the ratio for each channel, month by month.
- A rising ratio suggests double counting, view-through credit, or credit for patients who would have come anyway.
- A falling ratio suggests lost signal: a cross-domain gap, a call number that is not swapping, or visitors declining consent.
- Investigate any move you cannot explain by a change you made, such as a new consent notice, a removed tag or a campaign change.
For illustration only: if Meta reports 30 booking conversions in a month and your practice software shows 24 new patients from that source, the ratio is 1.25. That points towards double counting or patients who would have come anyway, so look at the gap before you credit the campaign.
This is our suggested diagnostic, not an industry standard. The number to manage is cost per confirmed new patient: spend on a channel divided by the new patients you can confirm from it.
What you give up
Per-click attribution detail, some of the signals automated bidding learns from, retargeting, and audiences built from site behaviour. What you keep are calls, bookings and cost per confirmed new patient. For most clinics, that last figure is the one that matters.
Where server-side tracking fits
Server-side tracking sends events from your own server to an ad platform instead of from the visitor’s browser. Meta’s Conversions API is the most common example for clinics running Meta ads.
It can make delivery more reliable when browsers or blockers stop a pixel, and that is often why reported conversions rise after the move. If the same event arrives from both the browser and the server, it needs a matching event ID so the platform counts it once. More events delivered does not mean more patients.
It does not change what you disclose. If a server-side event carries the same page, event name or identifier a pixel would, the consent question is the same.
Vendor documentation also notes that platform health restrictions apply however the data arrives. As covered earlier, we have not seen the OAIC address conversion APIs, so treat server-side tracking as a delivery improvement, not a compliance fix.
Why booking systems break clinic tracking: Cliniko, HotDoc and Dentally
A common failure pattern is that the booking happens on another domain. A patient clicks an ad, lands on your site, then books in a system on a different address. Analytics sees the visit end and a new one begin, often labelled “direct” or “referral”. Booking setups differ across GP, dental, allied health and specialist practices, but the break points are the same.
| Setup | What happens | Options |
|---|---|---|
| Hosted booking page on another domain | The session breaks at the handoff. Cliniko publishes a guide to connecting hosted bookings to Google Analytics. | Cross-domain measurement if the system supports it, or have the system send the booking event back |
| Booking form embedded in an iframe | Cliniko’s guidance notes that cross-domain tracking does not work with embedded bookings. | Use a hosted flow, use the system’s own event feed, or import bookings as offline conversions |
| Widget or marketplace booking | Bookings made through a widget or a marketplace listing, such as HotDoc or HealthEngine, happen inside the vendor’s system | Ask the vendor which events it can pass back and what data it sends, and treat the integration as another tag |
| System with no native web analytics link | A Dentally community post from October 2025 said it had no native GA4, GTM or cross-domain support and suggested sending bookings to Google Ads as offline conversions using email or phone | Confirm the current position with the vendor, then apply the privacy point below |
Four rules follow:
- If your booking form is in an iframe, do not force cross-domain tracking. Check whether a hosted option or event feed exists. If not, use offline import or the intake question.
- If you send booking events to Google Ads, send the minimum identifiers, name events neutrally, and put the practice on the consent footing described earlier. Email and phone sent for matching are patient identifiers.
- If your booking provider says it “integrates with Google Ads”, ask three things: what page address it sends, what identifiers, and what service name. That integration is another tag for your inventory.
- For phone, check your own call logs: what share of new patients ring rather than book online? If it is large, call tracking is your most important measurement. It is also a script on every page, and recordings are likely to contain health information. Recording rules differ across states and territories, so check what the vendor collects and any notice requirements with your adviser.
Why your clinic’s ad numbers changed: a diagnostic table
| What you see | Likely cause | What to check |
|---|---|---|
| Calls steady, “calls from ads” fall | The number-swap script is broken, blocked or was removed in a tag clean-up | Click through from an ad and confirm the number changes |
| Bookings show as direct or referral from your booking domain | The session broke at the booking system | How the booking is embedded, and whether the system supports cross-domain tracking |
| Meta audiences shrank or events disappeared | Health and wellness restrictions on URL data, custom events or audiences | Your data source category in Events Manager and the names of your custom conversions |
| Reported conversions fell when a consent notice went live | Only consenting visitors are now measured | Compare with bookings in practice software. Expect platform counts to fall while bookings hold. |
| Reported conversions fell after removing a pixel, bookings unchanged | The pixel was counting patients already counted elsewhere, or the platform now has less to learn from | Cost per confirmed new patient. If it is flat, nothing was lost. |
| Reported conversions rose after moving events to a server | Better delivery of events previously blocked, not extra patients, and possibly duplicates | Bookings, and whether the same event arrives from both browser and server with a matching event ID |
| Google Ads remarketing lists unavailable | Sensitive category limits on the account | The policy status shown in your Google Ads account |
Who owns your tags? A checklist for clinics and agencies
Pixels reach clinic sites in ordinary ways. The OAIC found agencies and web teams commonly installed them as part of advertising campaigns, often assuming the data was anonymous. Medmate reviewed its contracts with external marketing agencies during the investigation. Monash IVF’s pixels ran from 2012 until December 2024. And the OAIC treated both providers as holding the pixel data because they controlled what was collected.
A contract can allocate costs between you and your agency. It does not move your own obligations as the clinic. That is our reading of the decisions, so confirm it with counsel. Tags on pages you do not control, such as a directory listing or a booking page hosted by a vendor, sit outside your inventory but not outside the question of what you have agreed to.
Governance basics
- One named owner at the clinic holds admin rights to the tag manager and ad accounts. Agencies get access, but the clinic owns the accounts.
- No new tag without a note of what it sends and why, and a privacy impact assessment before anything that sends page-level data to an advertising platform.
- A scan every quarter, and after any plugin or theme update.
- Offboarding. When a campaign or an agency relationship ends, remove the tags and revoke access.
Questions to ask any agency, including us
- Can I see every tag and pixel on my site, and who owns each account?
- What does each one send, and on which pages?
- Does anything fire before a visitor consents?
- Do any URL parameters or event names include treatment or condition names?
- Are you building audiences from site visits or from patient lists, and what is the consent basis?
- How do you measure new patients when consent is declined, and what do you reconcile against?
- What happens to pixels and data in the ad accounts if we end the contract?
- Who is responsible if a tag you installed collects something it should not, and what does our contract say?
Take the answers, and the contract wording on tag approval, deletion on exit and who installs what, to your lawyer.
A second set of eyes
If you would rather not run the inventory alone, ask your agency to do it with you. Pracxcel offers a free growth audit. If tracking and consent are your concern, say so when you book. Whoever you choose, the first deliverable should be the same: a list of every tool, what it sends, and who owns it.







