Table of content
On 11 January 2027, Google's advertising tags hosted in a Google Tag Manager server container will receive your visitors' full IP address by default. Today, redaction applies container-wide and every tag receives a truncated address. Tomorrow, processing happens tag by tag, and opting out of the full address will not be a checkbox: you will need to build a transformation in your container, test it and publish it before that date.
For a Swiss company, this is not an engineering footnote. Google's consent policy explicitly covers end users located in Switzerland, and Google itself already excludes IP sharing for our country in another of its advertising products. Here is what the notice says, what the official documentation confirms, what it does not say yet, and the procedure to run through in the next eight weeks.
The essentials
A Google notice titled Updates to network signal processing in server-side tagging announces a move from container-wide redaction to tag-level processing, starting 11 January 2027.
The tags concerned are those for Google Ads, Display & Video 360, Search Ads 360 and Campaign Manager 360. Google Analytics tags keep applying partial redaction.
Only server containers are affected. A site running browser-side tracking only, with no server container, is outside the scope.
Opting out is not a setting: it requires a custom template variable and an Augment event transformation that overrides the ip_override field. The procedure is documented by Google.
In Switzerland, three points deserve a written decision: Google's consent policy, the IP sharing exclusion it already imposes elsewhere, and what your privacy policy currently promises your visitors.
As of 16 November 2026, the Tag Manager release notes do not mention this change.
What is Google changing on 11 January 2027?
The starting point is a service notice sent to Google Ads customers, titled Updates to network signal processing in server-side tagging. It was made public on 2 November 2026 by consultant Adriaan Dekker on LinkedIn, then relayed the same day by PPC Land, on 3 November by Search Engine Journal under Matt G. Southern's byline, and a second time by PPC Land on 12 November with an implementation analysis.
The message fits in one sentence. In a server container, handling of the IP address will no longer be a container-wide decision, but a behaviour specific to each tag. And the default behaviour of Google's advertising tags becomes accepting the full IP address.
It is worth measuring what this moves. Until now, a measurement lead could tell their management and their board something simple: our server container truncates IP addresses before passing anything to Google. That sentence, often copied verbatim into a privacy policy or a record of processing activities, will stop being true for some of your tags if nothing is done.
Two methodological notes, which matter to a careful reader. First, the 11 January 2027 date and the description of the new behaviour come from Google's notice, not from a help page. Second, we checked the Tag Manager release notes on 16 November 2026: the latest entry is dated 9 November 2026 and covers page view counting when a page is restored from the browser's back/forward cache. No entry mentions IP address processing. The official documentation therefore does not yet say what the notice announces, which is a good reason to rely on what you observe in your own container rather than on an announcement read second-hand.
Which tags receive the full IP address, and which do not?
The dividing line is not the one most people expect. It does not separate Google from everyone else, it separates advertising from measurement.
Google advertising tags: Google Ads, Display & Video 360, Search Ads 360 and Campaign Manager 360, Floodlight included. They accept the full IP address by default from 11 January 2027.
Google Analytics tags: nothing changes. They already apply partial redaction by default and will keep doing so.
Non-Google tags: Meta, LinkedIn, TikTok, Microsoft and the rest remain in control of the signals they accept. Their behaviour depends on the vendor, not on this notice.
Web containers and sites with no server container: out of scope. If your measurement is entirely browser-side, you have nothing to do.
That last line deserves a second reading, because it produces a counter-intuitive effect: the organisations that must act are the most advanced ones, those that invested in a server container precisely to keep better control of their data. The others are not concerned.
Why does a server container see the IP address in the first place?
An architecture reminder helps here, because it explains why the question only arises in this context. In a classic client-side setup, the visitor's browser calls each platform directly. Each of them sees the visitor's IP address, since it receives the request: there is nothing to decide, and nothing to truncate.
In a server-side setup, the browser sends the event to a collection server hosted on a subdomain of your site. That server receives the real IP address, and it is the one calling the platforms. This intermediate point is exactly what gives the architecture its value: it lets you apply consent, normalise identifiers, hash contact data and, precisely, pass on nothing more than a truncated IP address. That power of arbitration is the whole point of server-side; it is also why a change in default behaviour concerns you directly.
If the subject is new to you, our complete guide to server-side tracking and FADP compliance in Switzerland sets out the basics, and hosting choices are covered in our comparison of Infomaniak, Google Cloud and Stape.
Why does this change weigh more heavily in Switzerland?
Google's consent policy names Switzerland explicitly
Google's EU user consent policy applies to end users located in the European Economic Area, the UK and Switzerland. It requires legally valid consent for the use of cookies or other local storage where legally required, and for the collection, sharing and use of personal data for ads personalisation. It also requires keeping records of the consent given, providing clear revocation instructions, and identifying each party that may use that personal data.
Passing a full IP address to an advertising tag brings that transmission inside the scope of these obligations. The question is therefore not purely technical: it is your consent chain that has to hold, and your documentation that has to describe it accurately. If your banner and your Consent Mode v2 are approximate, this change makes the approximation more expensive.
Google already excludes IP sharing for Switzerland, elsewhere in its products
This is the most useful point in this article, and the least known. In its customer match policy, Google writes that IP and timestamp matching is not supported in the EEA, the UK or Switzerland, and asks advertisers to exclude IP sharing for these regions, to ensure transparent data collection disclosures and to secure all legally required consent.
In other words, for part of its advertising offering, Google itself takes the view that a Swiss visitor's IP address should not be shared with it. That does not make the new tag behaviour unlawful, because it is not the same mechanism. But it creates a tension you will have to document: on one side a product policy that excludes IP sharing for Switzerland, on the other a default behaviour that transmits it. If you choose to let the full address through, write down why, and date that decision.
Note also that in June 2026, Google told its AdSense publishers that IP-based measurement and personalisation solutions would arrive in the EEA, the UK and Switzerland from 3 August 2026. That information comes from an email relayed by PPC Land on 18 June 2026; we have not found a public help page confirming it. It nevertheless points in a direction: the IP address is becoming a measurement signal again, after years of being presented as data to neutralise.
What the FADP says
The Swiss Federal Act on Data Protection defines personal data in art. 5 let. a as all information relating to an identified or identifiable natural person. A full IP address, combined with the other signals in a measurement event, falls within that definition in most configurations. That is the working assumption we apply systematically, and it matches the one Google itself applies by excluding IP sharing for Switzerland in the policy cited above.
Three practical consequences follow.
Proportionality (art. 6 para. 2 FADP). Processing must remain limited to what is necessary. Passing on a full IP address because it has become the default is not a justification. The question to settle is: what does that signal bring you, and is the gain worth the data?
Recognisable purposes (art. 6 para. 3 FADP). Data is collected for specified purposes that are recognisable to the data subject. A visitor who reads that you truncate IP addresses cannot recognise an advertising purpose based on the full address.
Duty to inform (art. 19 FADP). You must inform of the controller's identity, the purpose and the recipients, with additional requirements where data is disclosed abroad. A change in what you send to Google is a change in what you must describe.
Add to this the record of processing activities under art. 12 FADP, which describes categories of data and recipients. If it mentions IP truncation, it becomes inaccurate on 11 January 2027 without anyone touching anything. We detailed these documents in our article on the five documents the FDPIC may request from your DPO.
Does your privacy policy still tell the truth?
This is the most immediate risk, and the easiest to fix. On most Swiss sites running a server container, a sentence was written two or three years ago, at deployment time, and has not been reread since. It says that IP addresses are truncated or anonymised before being passed to third parties.
If that sentence exists on your site, you have two options and a single deadline.
You keep the truncation by building the transformation described below. Your documentation stays accurate, and you simply add scope: which tags, for which events.
You let the full address through to advertising tags. In that case, the sentence has to be rewritten before 11 January 2027, the record updated, and the decision logged with its justification.
What is not an option: leaving documentation that describes behaviour your setup no longer has. An inaccurate statement is harder to defend than a processing operation that is owned and explained. The decisions issued by the FDPIC in 2026, which we analysed in this article, show that the gap between what is promised and what is done has become the entry point for proceedings.
How do you keep a truncated IP address in your server container?
The procedure is documented by Google in its help page on IP address redaction for server-side tagging. That page carries no date and does not mention the 11 January 2027 change; it does describe exactly the mechanism to put in place. Allow an hour for a simple container, half a day for a busy one that is properly tested.
Step 1: create a custom template variable
In your server workspace, create a custom template variable. Its code fits in two lines: require the getAnonymizedIpAddress API, then return the result of calling it. For every request, this variable produces a truncated version of the original IP address.
Step 2: add an Augment event transformation
Go to the Transformations section, create an Augment event transformation, and configure it to override the ip_override field with the variable created in the previous step. That field is the one carrying the IP address in the server container's event model. Google notes that this approach lets you redact the signal before any individual tag accesses the data: that is where the protection happens.
Step 3: scope it to the right tags
Under matching conditions or affected tags, select the tags that should receive the partially redacted address: Google Ads, Floodlight, and any other tag concerned. This is where the operational trap sits: a tag added in six months will not be covered automatically. Build that check into your release procedure, alongside your consent checks.
Step 4: preview, then publish
Run preview mode, fire your tags, and verify in the outgoing request detail that the address is indeed truncated before it leaves your server. Then publish the container. Do it in December 2026 rather than the first week of January: that gives you margin if unexpected behaviour shows up.
What does getAnonymizedIpAddress actually do?
The server-side tagging API documentation describes the function's behaviour precisely, and it is worth knowing before signing any statement about it.
IPv4: the last octet is masked to zero. An address such as 198.51.100.42 becomes 198.51.100.0.
IPv6: the first 48 bits are kept, the last 80 bits are zeroed.
Invalid or unresolvable address: the function returns ::, which, per Google, makes the result always safe to use as ip_override.
Two honest remarks. First, this is not anonymisation: it is truncation. An IPv4 block stripped of its last octet still designates 256 addresses, often in the same organisation or neighbourhood. If your risk assessment assumes anonymisation in the strict sense, it needs revisiting. Second, truncation has a functional cost: it degrades geolocation precision and, depending on the product, the quality of certain matches. That cost is usually moderate for a Swiss SME with a national audience, but it is not zero, and it deserves to be named rather than glossed over.
What about non-Google tags: Meta, LinkedIn, TikTok?
They are not covered by the notice, but they are not protected by your transformation either. Google itself warns that some third-party tags may extract IP addresses from somewhere other than the standard field, for example from a custom header or a specific event parameter.
The approach is as follows.
Check, in the vendor documentation or with Tag Assistant, whether the tag actually reads the ip_override field. If it does, the transformation already in place is enough.
If it reads another field, identify it precisely, then create a second Augment event transformation overriding that field with the same anonymisation variable.
Scope that transformation to the third-party tags concerned, then test and publish.
In our deployments, this is the check that holds the most surprises, because a third-party tag updated by its vendor can change behaviour without warning. It is exactly the kind of silent drift we described in our article on why server-side tracking breaks after launch.
Truncate, blank or let through: three options, three consequences
Three positions are defensible, as long as you choose them knowingly.
Let the default apply. Implementation: nothing to do. Consequence: Google advertising tags receive the full address. To handle: the privacy policy, the record of processing, the legal basis and consistency with Google's consent policy. This is the option that demands the most documentation work, not the least.
Truncate. Implementation: the four-step procedure above. Consequence: tags receive a truncated address, geolocation loses precision. This is what we recommend by default to our Swiss clients, because it preserves consistency with what has already been declared.
Send nothing at all. Implementation: override the field with an empty value rather than with the variable. Consequence: geolocation disappears, and some products may behave unexpectedly. Reserved for the most sensitive contexts, and to be tested seriously.
One point remains open, and we would rather say so than fill it with an assumption: Google has not stated whether requests sent without advertising consent will also carry the full IP address. Until that is documented, the prudent position is to make sure your advertising tags only fire after consent, and that unconsented requests stay redacted.
Five checks to run before 11 January 2027
Inventory your server containers. How many do you have, who has access, and which Google advertising tags live in them? A forgotten container on a secondary site is the most common case.
Reread your privacy policy and your record of processing, searching for the words IP address, truncation and anonymisation. Note every sentence that would become false.
Build and publish the transformation if you choose truncation, in December 2026, with a preview test documented by a screenshot.
Verify consent-conditional firing for all your advertising tags, not only Google's.
Date and log the decision. Which option, for which tags, decided by whom, on what basis. That is the document you will be missing if the question comes up two years later.
If you handle offline conversions or a CRM connection, use this trip into the container to also reread our article on Google Ads Data Manager: both subjects come down to the same judgement call on what you actually pass to Google.
How A-Track can help
We are a Geneva team specialising in server-side tracking and FADP and GDPR compliance. On this specific subject, our work comes in three stages.
A review of your server containers: which advertising tags are in place, what they receive today, and what they will receive on 11 January 2027 if nothing is done. That is the scope of our tracking and compliance audits.
Implementation of the transformation, its preview test and its publication, with verification of third-party tags and an update to your documentation.
Follow-up, because a transformation correctly set in December can be bypassed by a tag added in March. That is the purpose of our annual maintenance.
The next step
Open your server container and answer a single question: which Google advertising tags are in it? If the answer is none, you can close this tab. If the answer is one or more, you have roughly eight weeks to decide what they should receive, put it in place, and write it down somewhere.
This is not a heavy project. It is a decision to make before it is made for you, and that is precisely the difference between a measurement setup you steer and one you merely endure.
Is your server container ready for 11 January 2027?
We check which tags will receive the full IP address, set up the transformation and update your documentation.