AI Agent Tool DateTime Always in America/New_York Timezone (Google Calendar)

Help me Build my Workflow
1

Hi all,

I’m using the AI Agent in n8n with a Gemini model to create events in my Calendar using the Google Calendar tool. However, no matter what I try, the datetime keeps coming through in the America/New_York timezone, even though:

  • My Google Calendar account is set to Europe/London
  • I’ve tried manually setting the timeZone field to “Europe/London”
  • The dateTime returned by $fromAI(‘Start’, ‘’, ‘string’) looks correct, but it’s interpreted as New York time, even though my AI Agent uses correct UK times.
  • My workflow timezone is Europe/London
  • I’ve tried deleting and readding the node

Here’s an example of the output I’m getting:

"start": {
"dateTime": "2025-07-19T21:00:00+01:00",
"timeZone": "America/New_York"
}

Even trying things like this inside the AI Agent Tool expression doesn’t help:

{{ $fromAI('Start', '', 'luxon').setZone('Europe/London').toISO() }}

But the event still shows up in the wrong timezone in Google Calendar.

I am self-hosting N8N on railway.
I have used Google Cloud working credentials on my same google account. I have looked if there’s a way to change the timezone on google cloud but I’m sure it just links to your google accounts which would, again, be Europe/London in my case.

Has anyone managed to force the AI Agent Tool to return a datetime in a specific timezone like Europe/London?

Any tips or working examples appreciated!

Thanks,
Zaine

2

Hey @Zaine_Pilsworth hope all is well, welcome to the community.

So, to re-iterate,

  • your calendar is in europe/london tz
  • your workflow is in europe/london tz
  • the date and time in the result of creating an event comes back correct, but the timezone isn’t.

Quick question, is your GENERIC_TIMEZONE env variable set for the instance? It defaults to America/New_York, so it isn’t set this would be the default. Here is another doc about it.

Just in case I am attaching the workflow I’ve been testing with (and it worked flawlessly):

3

Ich bin auf ein ähnliches Zeitzonen-Problem gestoßen, während ich an einem Projekt arbeitete, das Alter aus Daten berechnet. Alles sah korrekt aus, bis verschiedene Zeitzonen die Ergebnisse in einigen Fällen um einen Tag verschoben haben. Das hat mir bewusst gemacht, wie wichtig die Behandlung von Zeitzonen ist, wenn es um Daten geht. Ich verfolge diesen Thread, weil ich neugierig bin, wie die endgültige Lösung aussieht.