I’m running into a confusing behavior with AI outputs that look structurally identical, but behave very differently depending on whether the output column is labeled “Response” or not.
What I’m seeing
I have two AI outputs that are both JSON objects and both contain lists.
Example A (not labeled “Response”)
{
"locations": [
{ "type": "Other", "country": "Australia" },
{ "type": "Other", "country": "France" }
],
"confidence": 8
}• locations is displayed as a list in the cell
• I can run actions on the list inside the cell
• BUT:
In Formulas, I can only select the main column, not the sub-items
(locations does not appear as selectable input)
Example B (column labeled “Response”)
{
"profiles": [
{
"jobTitle": "Partner",
"profileUrl": "https://www.linkedin.com/in/...",
"confidenceScore": 85
}
],
"confidence": "high"
}Column is labeled “Response”
All sub-items (e.g. profiles) are selectable in Formulas
The question
Why do these outputs behave differently if both are valid JSON objects with lists?
Specifically:
What determines whether an output becomes a “Response” column or not?
Why are sub-items accessible in Formulas only when the column is labeled “Response”?
Is there a way to make non-Response outputs expose their sub-items to Formulas as well?
From a user perspective, both outputs look the same, but only one is usable in formulas.
Would love some clarification on the underlying logic here.