Contents
At a glance
- Category
- Automation mechanics
- Stands for
- JavaScript Object Notation
- Six ingredients
- Objects, arrays, strings, numbers, booleans, null
- Array indexing
- Starts at zero - the first item is [0]
- Common rejections
- Trailing commas, comments, single quotes - none are valid strict JSON
- For AI output
- Ask via structured output or function calling with a schema, and enable strict mode
What JSON is
JSON - JavaScript Object Notation - is the lingua franca of APIs. It is a text format for structured data with six ingredients: objects (named fields in curly braces), arrays (ordered lists in square brackets), strings, numbers, booleans and null. Almost every webhook payload, API response and automation platform variable you meet is JSON, so being able to read one is a genuinely useful non-technical skill.
Reading it is mostly about following the nesting. In {"customer": {"name": "Ada", "phones": ["+447700900123"]}} the path to the phone number is customer.phones[0], and that path expression is exactly what you type into a field mapping. Objects are addressed by name, arrays by position starting at zero.
The details that catch people out
Four. Types are not interchangeable: "1200" is a string and 1200 is a number, and an API that wants one will often reject or misinterpret the other. Null is not absent: a field present with value null behaves differently from a field that is missing, and many mapping expressions treat them differently too.
Arrays might have zero, one or many items, and a workflow built against a single-item example breaks the day two arrive - if the platform does not loop, only the first is processed silently. There are no comments and no trailing commas in strict JSON; a trailing comma is the most common reason hand-written JSON is rejected.
JSON and AI output
JSON is also how you get usable output out of a model. Asking for prose and parsing it is fragile; asking for JSON against a declared schema, using the provider's structured output or function calling mode, gives you something a downstream step can consume reliably. Turn the strict mode on where it exists - it converts an entire class of parsing bugs into a guarantee.
Do note the cost angle: JSON tokenizes less efficiently than prose, because every brace, quote and field name is tokens. Verbose field names in a large schema are a real cost on high-volume workflows, so prefer short, clear names over long descriptive ones when volume is high.
What you should know
Log the raw JSON at every boundary before you parse it. When something breaks, the raw payload is the artefact that tells you whether the sender changed format or your mapping was wrong, and it takes minutes rather than hours to answer that question with it in hand.
When a payload is unfamiliar, paste it into a formatter or your platform's own JSON viewer and expand the tree rather than reading it as a wall of text. Ninety per cent of "the mapping is not working" is a path that is one level off, and seeing the structure laid out makes that obvious immediately.
JSON FAQ
What is JSON used for in automation?
It is the format almost every API, webhook payload and platform variable uses to carry structured data. When you map a field from one step to another, the path you type is navigating a JSON structure, so reading one makes automation work considerably easier.
How do I read a JSON path?
Follow the nesting. Objects are addressed by field name with a dot, arrays by position in square brackets starting at zero. So customer.phones[0] means the first phone in the phones list inside the customer object.
Why is my API rejecting my JSON?
Most often a trailing comma after the last item, a comment, or single quotes instead of double quotes - none of which are valid strict JSON. After that, the usual cause is a type mismatch: sending a number as a quoted string, or a date in the wrong format.
What is the difference between null and a missing field?
A null field is present with an explicitly empty value; a missing field is not there at all. APIs frequently treat them differently - null may clear a value while omission leaves it unchanged - and mapping expressions handle them differently too, so it is worth being deliberate.
Why does my workflow only process the first item?
Because the payload contains an array and the workflow was built against an example with one item. Without an explicit loop or split step, only index zero is used and the rest are silently ignored. Always check whether a field is an array before mapping into it.
How do I get JSON out of an AI model reliably?
Use the provider's structured output or function calling mode with an explicit schema, and turn on strict mode where it exists so the response is guaranteed to match. Then still validate before use, since a schema guarantees shape but not that the values are correct.
Primary source
RFC 9110 - HTTP semantics (the transport JSON payloads travel over)
Checked August 2026. Anything priced or rate-limited moves, so verify against the source before you budget.
Want this built rather than explained?
Knowing what json means is one thing; having it running in your business is another. We scope one workflow at a time, quote a fixed price in writing after a free consult, build it, and hand it over with documentation you own.
