-
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 2 replies
-
|
I had a closer look at this against the current implementation and Google docs. Nuxt Scripts is doing the expected Consent Mode queueing here: The likely mismatch is Consent Mode semantics rather than the Nuxt Scripts API. With One other thing to check: your config has: $development: {
scripts: {
registry: {
googleTagManager: "mock",
},
},
}If Tag Assistant is pointed at a dev build, that mock config means the real GTM container/default consent setup is not what you are testing. I would test against a production/preview build without the mock, then inspect ["consent", "default", { ...denied state... }]
{ event: "gtm.js", "gtm.start": ... }
["consent", "update", { ...granted state... }]So my recommendation is:
References I checked:
|
Beta Was this translation helpful? Give feedback.
-
|
Hey @buffalom, both pieces of this are now on main. #771 (yours) fixes the underlying bug where the gtag helper was pushing a real Array onto the dataLayer instead of an Arguments object, which is what Tag Assistant and the Analytics Debugger key off, so your consent rows should now show up properly in both. Thanks for the deep investigation and the screenshots, that made the fix obvious. #772 builds on top by adding a consent.default() method for runtime-derived defaults (handy when you resolve region/CMS state after mount), and runs every consent.* call through the canonical GCMv2 schema, warning via consola on typo'd keys or non granted/denied values, which should make this whole class of issue noisier next time. Try the next release and let us know if anything still looks off. |
Beta Was this translation helpful? Give feedback.




Hey @buffalom, both pieces of this are now on main. #771 (yours) fixes the underlying bug where the gtag helper was pushing a real Array onto the dataLayer instead of an Arguments object, which is what Tag Assistant and the Analytics Debugger key off, so your consent rows should now show up properly in both.
Thanks for the deep investigation and the screenshots, that made the fix obvious. #772 builds on top by adding a consent.default() method for runtime-derived defaults (handy when you resolve region/CMS state after mount), and runs every consent.* call through the canonical GCMv2 schema, warning via consola on typo'd keys or non granted/denied values, which should make this whole class…