A browser feeds a Google Tag Manager container that distributes to several Google destinations; a removed element remains dashed below the container

Published by Thomas dans la catégorie Tracking on 26.10.2026 Last update : 17.11.2026 à 16h20


Table of content

Since 8 October 2026, Google has been changing how Google Tag Manager works under the bonnet. Your container can now send data directly to Google's measurement products without loading a separate gtag.js file, and your Google Analytics, Google Ads or Floodlight accounts attach to the container as destinations. Nothing switches over automatically, but two of these changes can already alter how your tags behave without you touching anything: the standardisation of gtag('config') commands, and the arrival of gtag.config events in the dataLayer.

The essentials

  • On 12 October 2026, Google announced that Tag Manager containers can send data straight to Google measurement products, with no extra gtag.js load.

  • On 8 October 2026, Google standardised gtag('config') behaviour: every gtm.js container initialises on container load, and configuration commands now surface in the dataLayer as gtag.config events.

  • The immediate consequence: a tag sitting on a custom event trigger with a loose regular expression can fire on its own.

  • Container optimisation is not automatic. Google states explicitly that no changes are made without your consent.

  • The governance point not to miss: each linked destination gets Read access to your container by default, and those permissions are not versioned.

  • Server-side tracking is not replaced. This change is about browser-side loading.

What exactly changed in Google Tag Manager in October 2026?

Three announcements landed in quick succession, and they are worth separating, because their practical consequences are not the same.

The first is dated 20 August 2026. Google describes an interface overhaul (a streamlined Overview page, a new Settings tab, an Advanced tab that groups triggers, variables, templates and folders) and, more importantly, the convergence of the Google tag and Tag Manager. Setups that only used the Google tag become full Tag Manager containers, with an interface, debugging and version history. Google notes that this upgrade will not change the in-page behaviour of the Google tags.

The second is dated 8 October 2026, and it is the delicate one, because it touches code already live on your pages. Google is standardising how tagging snippets initialise. According to the Tag Manager release notes, all gtm.js container snippets now initialise on container load, regardless of any gtag('config') commands on the page. Those configuration commands then appear in the dataLayer as gtag.config events.

The third is dated 12 October 2026. Google opened up container optimisation: containers can send data directly to Google measurement products instead of loading a separate gtag.js script. The stated goal is threefold: site performance, better data collection, and settings managed in one place.

For a Swiss SME running WordPress, with a GTM container installed by an agency three years ago and a consent banner bolted on afterwards, the second announcement is the one that can cause trouble this week. The third is a project to schedule.

What is a Google destination, and how does it differ from a tag?

The vocabulary matters here, because the word "tag" already does too much work. In Google's documentation, a destination is a Google measurement product account that shares a Google tag's configuration and receives its data. Four types are supported: a Google Analytics 4 web data stream, a Google Ads account, a Floodlight configuration used in Campaign Manager 360 and Display & Video 360, and a Merchant Center conversion source.

Previously, your GTM container loaded a Google tag, which in turn loaded the gtag.js file, which then talked to each of those accounts. Two loading layers, two places to set parameters, and a painful diagnosis whenever something went missing. After optimisation, the container itself carries the instructions and settings for its destinations and speaks directly to the Google products.

Simo Ahava, author of the reference blog on Google Tag Manager, documented the change on 13 October 2026 and sums up the shift: the container takes on the capabilities of a Google tag, destinations are no longer loaded as separate gtag.js resources, and new Destination tag types (Google Analytics, Google Ads, Google Floodlight) replace the old Google tag type. He also flags that the Conversion Linker moves to legacy status.

The diagram in this section compares the two loading chains:

  • Legacy model: web page, then the gtm.js container, then a Google tag, then the gtag.js file loaded on top, then the Analytics, Ads and Floodlight accounts.

  • Optimised model: web page, then the gtm.js container carrying its destinations' settings, then a direct send to the Analytics, Ads, Floodlight and Merchant Center accounts.

  • What you gain: one fewer script request, a single place for shared settings, and an explicit link between each destination and the tags feeding it.

  • What to watch: settings that were duplicated across several Google tags have to be reconciled, and access permissions change.

Why can your custom event triggers start firing on their own?

This is the technical point to check first, and it is not hypothetical. Google writes it into its 8 October 2026 release notes: configuration commands now appear in the dataLayer as gtag.config events, which may cause tags on wildcard custom event triggers to fire.

In plain terms: many containers hold a custom event trigger whose event name is a deliberately loose regular expression. The classic agency case is an "all e-commerce events" trigger set up with regex matching on something like .* or view|add|purchase, because it was quicker than listing events one by one. Google's documentation notes that the "use regex matching" option lets that field handle regular expressions, and that a trigger set to all custom events responds to any custom event meeting the name condition.

A gtag.config event arriving in the dataLayer meets a loose pattern. The tag fires. Depending on what that tag does, the consequences range from statistical noise to a real bill: a Google Ads conversion sent on every page load, a GA4 event inflating an engagement rate, an advertising pixel firing before the visitor has made a choice and putting your compliance in breach.

The diagram in this section sets out the sequence to verify:

  1. The page loads the gtm.js container, which now initialises without waiting for a gtag('config') command.

  2. A configuration command on the page is pushed into the dataLayer as a gtag.config event.

  3. A custom event trigger with a loose regular expression recognises that event name.

  4. The tag attached to that trigger fires, with no visitor action involved.

  5. The result shows up in your reports: duplicated conversions, phantom events, or a third-party tag that should never have fired.

The check takes ten minutes per container. Open GTM preview, load a page of the site, and look at the event list in the left-hand panel. If a gtag.config event appears, list every tag that fired on it. For each custom event trigger using a regular expression, replace the permissive pattern with the explicit list of your event names, or add an exclusion condition on gtag.config.

The special case of sites mixing gtm.js and gtag('config')

There is an even more fragile setup, and it is common among SMEs that changed provider somewhere along the way: a Google tag ID that does not start with GTM (so a G-, AW- or DC- ID) loaded through the gtm.js snippet path, with gtag('config') calls added alongside. Google calls this an unsupported implementation and warns that sites using it may experience unexpected changes to their tag behaviour. Affected containers were notified by email.

The fix depends on how you use it:

  • You use gtag('config') and nothing else: deploy the standard Google tag (gtag.js) snippet, which loads googletagmanager.com/gtag/js, immediately after the opening head tag on every page. Then remove the gtm.js snippet and the standalone gtag('config') commands tied to it.

  • You use GTM for other tags: keep the GTM snippet, but replace the G-, AW- or DC- ID with your GTM container ID, then deploy the Google tag and set its options inside the Tag Manager interface.

Two checks are then enough, and they are the ones Google recommends. In Tag Assistant, your tag ID should appear under the Google tags found, with a green or blue status. In the Network tab of Chrome DevTools, requests should go to googletagmanager.com/gtag/js rather than the gtm.js path.

What does the Settings tab change for tag governance?

This is the most useful part of the new model for a marketing team, and it is underrated. Until now, settings shared across your Google tags (cross-domain domains, cookie lifetime, URL parameter redaction, server-side transport) were copied into each tag, sometimes differently, with nobody noticing. Three Google tags, three divergent configurations, and gaps between Analytics and Ads that everyone blamed on attribution.

Shared settings become a container resource, with version history like tags and triggers. Destination tags can override them but, as Simo Ahava notes, only towards a stricter state. In other words, a destination can be more cautious than the shared setting, never more permissive.

Two caveats, and they matter for a serious rollout:

  • Setting conflicts are not flagged in the interface. If two Google tags hold different values, the call is yours. Google does provide a separate path in that case: instead of a simple add button, a wizard has you choose and confirm which settings to consolidate.

  • Your event tags and triggers are kept. The optimisation documentation is explicit on this point, which limits the risk of the exercise. It is the Google tag configuration that moves, not your tagging plan.

A-Track tip

Before starting the optimisation, export the current version of your container as JSON and archive it. The interface lets you roll back to an earlier version, but an export outside GTM gives you a readable reference to compare settings before and after, and a safety net if several people work in the same workspace.

Who gets access to your container after optimisation?

This is the question nobody asks, and it is the most important one for a Swiss company working with several providers. Google is clear: the links between the container and the destination accounts are created automatically, with Read access by default. In practice, users of your media agency's Google Ads account can see your GTM container, meaning your tagging plan, your variables and the way you measure your forms.

Two details in the optimisation documentation change how you should approach this. First, Tag Manager admins specify the level of access each destination has to the container during the optimisation: it is a decision to make at the time of the work, not afterwards. Second, and this is the trap, those permissions are not versioned: rolling back to an earlier container version does not restore them. They have to be changed by hand, in the External Account Links section of the container's admin settings.

For a Geneva SME sharing its Google Ads account with an agency, its GA4 with a consultant and its Floodlight with a media house, that deserves ten minutes of thought before clicking. The exercise is also an opportunity: most containers we audit hold forgotten access, granted to a provider who left two years ago. Our 15-point tracking audit checklist covers this governance angle.

From a FADP standpoint, the subject is not neutral either. Who accesses what in your measurement setup is part of what you must be able to document, alongside your processors and your transfers. Read access to a container is not access to personal data, but it reveals your collection architecture, and it falls within the scope of your record of processing activities.

Destination tags remain consent-aware, and that is the baseline for working cleanly in Switzerland as in the European Union. Beyond that, Simo Ahava describes in his 13 October 2026 analysis a new data transmission setting that acts as a Basic Consent Mode control layered on top of Advanced Consent Mode: hits can be withheld until consent is granted, then refired. He attaches a major caveat to that reading, and it is decisive for your compliance: the mechanism only covers Google's own tags.

In other words, nothing happening in that tab disciplines your Meta pixel, your LinkedIn Ads tag or your chat tool. Those tags still ignore Consent Mode and must be gated on the visitor's choice through the dataLayer, exactly as before. This is the single most common mistake we meet in audits: a team convinced consent is handled because the Google tab is green.

The diagram in this section places the three layers to check:

  • Layer 1, the banner: your CMP collects the choice and pushes it to the four Consent Mode v2 parameters. Nothing changes here.

  • Layer 2, the Google tags: Analytics, Ads, Floodlight and Merchant Center destinations, consent-aware, now steered from the container settings.

  • Layer 3, the third-party tags: Meta, LinkedIn, TikTok, chat, maps, session replay. Outside the scope of Google's settings, to be gated tag by tag.

  • The decisive check: refuse everything, reload, and confirm that no cookie and no request from any of the three layers goes out. The gcs and gcd parameters should accompany Google requests.

If this is new ground for you, our guide to Google Consent Mode v2 in Switzerland details the four parameters and the tests to run.

What does this change for server-side tracking?

Less than the headlines suggest, and that needs saying plainly to avoid a misreading. What changes in October 2026 concerns browser-side loading: one script fewer, settings grouped together, destinations attached to the container. Your server container keeps receiving requests, enriching them and redistributing them exactly as before.

Three concrete consequences all the same, for anyone running a server-side setup:

  • The transport URL remains an overridable field. In the new model, the parameter that sends requests to your collection domain is set like the other shared settings, with the option to override it at destination level. That simplifies the common case where a single container feeds a server-side GA4 and a client-side Google Ads.

  • The Conversion Linker becomes a legacy object. If your setup depends on it for click identifier persistence, check what the migration offers in its place before publishing.

  • Monitoring stays essential. A container changing its loading model is precisely the kind of event that breaks tracking without warning. That is the subject of our article on why server-side tracking breaks after launch.

And the underlying question has not moved: consolidating how Google tags load does not make you the owner of your data collection. A faster container still sends its data to the same recipients, from the visitor's browser, with the same cookie lifetime limits and the same exposure to blockers. The reasons to move server-side are unchanged, and we set them out in our complete guide to server-side tracking and FADP compliance. The first-party container, for its part, remains the answer to blockers, as explained in our article on the first-party GTM Custom Loader.

How do you run the migration without breaking measurement?

The reassuring part, and Google puts it in writing in its 20 August 2026 announcement, is that no changes are made automatically: you choose whether to adopt the new configuration. Optimisation starts from a banner on the Overview page, and changes stay in your workspace until you publish. Here is the order we apply with our clients.

The diagram in this section lays out these six steps:

  1. Deal with the 8 October urgency first. Look for gtag.config events in preview, and tighten every custom event trigger using a loose regular expression. This step does not depend on optimisation and should be done now.

  2. Inventory the container's Google tags. How many Google tags, which IDs, which settings, and which ones diverge. That work decides whether you follow the simple path or the conflict resolution path.

  3. Decide on access before clicking. For each destination, what level of container access, and for whom. Remember: those permissions are not versioned.

  4. Preview the changes, then add to the workspace. The interface offers a preview of all planned changes before carrying them into the workspace.

  5. Test in Preview mode, not in production. A full purchase or enquiry journey, refusing consent and then granting it, checking requests in the Network tab.

  6. Publish, then monitor for a fortnight. Compare conversion volume by source in Google Ads and in GA4 against the previous period. A gap on only one of the two platforms points to a destination problem.

For an agency managing thirty or forty client containers, the advice is to work in batches, starting with a simple, low-stakes container, and to document the consolidated settings for each client. Containers mixing several Google Ads accounts with a Floodlight configuration should be handled last, with time to spare.

How A-Track can help

We are a Geneva team specialising in marketing measurement, server-side tracking and FADP and GDPR compliance. On this particular change, we work in three ways.

First, checking gtag.config events and over-permissive triggers, because that is the only genuine urgency this October and it can be handled in a single pass per container. Then the container optimisation itself, with the arbitration of divergent settings, the decision on destination access, and acceptance testing before publication. Finally, a check of the whole setup, Consent Mode, third-party tags and server container, as part of our tracking and compliance audits or our Tracking and Compliance engagement.

If your setup is stable but nobody is watching these changes, that is exactly what our annual maintenance is for: active monitoring of Google's release notes, and fixes applied before your reports go wrong.

Conclusion: three actions this week

The October 2026 change is good news in the medium term. One fewer script request, versioned shared settings and an explicit link between destinations and tags: that is a cleaner architecture than today's stack. But the standardisation of gtag('config') is a behaviour change applying to code that is already live, and that is where the immediate risk sits.

  1. Today: open GTM preview on your site, look for a gtag.config event, and list the tags firing on it.

  2. This week: tighten every custom event trigger using a loose regular expression, and check in the Network tab that you are not mixing a G-, AW- or DC- ID with the gtm.js path.

  3. This month: inventory your Google tags and decide on destination access before launching the optimisation, testing in Preview mode before any publication.

If you would rather have someone look at the container with you, our flash tracking audit is delivered within 48 hours and will tell you on a single page whether your container is affected.

Related services

Tracking & Compliance

Server-side tracking and FADP / GDPR compliance: around 40% more data recovered on average (measured across 100+ installations), from CHF 1,200.

Tracking & Compliance Audits

Tracking audit, FADP and GDPR compliance audit, data governance audit. Custom quote.

Annual Maintenance

Your tracking monitored and your compliance kept up to date, from CHF 900 a year.

Questions fréquemment posées

What is a destination in Google Tag Manager?

▾
According to Google's documentation, a destination is a Google measurement product account that shares a Google tag's configuration and receives its data.

Four types are supported: a Google Analytics 4 web data stream, a Google Ads account, a Floodlight configuration used in Campaign Manager 360 and Display & Video 360, and a Merchant Center conversion source.

In the optimised model announced on 12 October 2026, the Tag Manager container itself carries its destinations' settings and sends them data directly, without loading a separate gtag.js file.

What is a gtag.config event in the dataLayer?

▾
Since the 8 October 2026 announcement, every gtm.js container initialises on load regardless of any gtag('config') commands on the page, and those commands appear in the dataLayer as events named gtag.config.

The risk Google flags is that a tag sitting on a wildcard custom event trigger will fire on that event.

The check: open GTM preview, load a page, look for a gtag.config event in the left-hand list, and list the tags firing on it. Then replace over-broad regular expressions with the explicit list of your event names.

Is optimising the GTM container mandatory?

▾
No. In its 20 August 2026 announcement, Google states that no changes are made automatically and that you choose whether to adopt the new configuration.

Optimisation starts from a banner on the container's Overview page, the interface offers a preview of the planned changes, and everything stays in your workspace until you publish.

The standardisation of gtag('config') behaviour from 8 October 2026 is different: it applies to code that is already live. It does not depend on your decision to optimise, so that is the one to check first.

Who can see my container after GTM optimisation?

▾
The links between the container and the destination accounts are created automatically, with Read access by default. Users of your agency's Google Ads account can therefore see your container, your tagging plan and your variables.

Two things to watch. Tag Manager admins specify the level of access each destination has during the optimisation, so it is a decision to make at the time of the work.

More importantly, those permissions are not versioned: rolling back to an earlier container version does not restore them. They have to be changed by hand, in the External Account Links section of the container's admin settings.

What is GTM (Google Tag Manager) Server-Side?

▾
It is the server version of GTM. It allows for receiving web data, manipulating it (cleaning PII, hashing emails), and distributing it to partner APIs (Meta, Google Ads, TikTok).

Is it mandatory if I only target Switzerland?

▾
Yes, if you use Google Ads. It is not Swiss law that requires it but Google's policy: since 31 July 2024, Google has applied its EU user consent policy to Switzerland. Even for a small Vaud business selling only locally, without Consent Mode v2, remarketing, audiences and conversion measurement are limited.

On the legal side, Switzerland requires information and a right to refuse (Art. 45c of the Telecommunications Act); prior consent is required for EU visitors.

Does this replace Server-Side Tracking?

▾
No, the two are complementary. Consent Mode passes the visitor's choice on to Google tags. Server-side handles how data travels: first-party measurement on your own subdomain, more resilient to browser restrictions and ad blockers, and data filtering before sending.

Combined, they give reliable measurement that always respects the visitor's choice: a refusal is still a refusal.