JSON

JSON is a text format for data that looks almost like Python lists and dictionaries, and the json module converts between the two.

Last updated: 13 Oct, 2026 · Python 3.12

To use a model's answer rather than only display it, you ask the model for JSON. Files, from Reading and writing files, often hold JSON too.

Syntax:

json.loads(text)     # JSON text -> Python value
json.dumps(value)    # Python value -> JSON text
json.load(f)         # from an open file
json.dump(value, f)  # to an open file

Reading JSON text with json.loads

import json

answer = '{"category": "billing", "priority": 4}'
data = json.loads(answer)

print(type(answer))
print(type(data))
print(data["priority"] + 1)

<class 'str'>
<class 'dict'>
5

answer is a string that looks like a dictionary. json.loads, load from a string, reads it and builds the real dictionary, so data["priority"] is a number you can add to.

Writing JSON text with json.dumps

result = {"id": 3, "category": "billing", "reply": None, "urgent": True}
print(json.dumps(result))
print(json.dumps(result, indent=2))

{"id": 3, "category": "billing", "reply": null, "urgent": true}
{
  "id": 3,
  "category": "billing",
  "reply": null,
  "urgent": true
}

json.dumps goes the other way. Look at what changed: JSON writes None as null and True as true, and always uses double quotes. indent=2 spreads it over lines for people to read.

Writing and loading the tickets file

The rest of the course works from five tickets saved as JSON. Writing them with json.dump gives a file the next lessons read back:

tickets = [
    {"id": 1, "customer": "Asha", "text": "I was charged twice for one order"},
    {"id": 2, "customer": "Ben", "text": "My parcel has not arrived"},
    {"id": 3, "customer": "Chen", "text": "Can I get a refund for the blue mug?"},
    {"id": 4, "customer": "Dara", "text": "How do I change my password?"},
    {"id": 5, "customer": "Eli", "text": "The parcel arrived but the box was crushed"},
]

# indent=2 keeps the saved file readable; the later lessons load it back
with open("tickets.json", "w") as f:
    json.dump(tickets, f, indent=2)

print("wrote", len(tickets), "tickets")

wrote 5 tickets

[
  {
    "id": 1,
    "customer": "Asha",
    "text": "I was charged twice for one order"
  },
  {
    "id": 2,
    "customer": "Ben",
    "text": "My parcel has not arrived"
  },
  {
    "id": 3,
    "customer": "Chen",
    "text": "Can I get a refund for the blue mug?"
  },
  {
    "id": 4,
    "customer": "Dara",
    "text": "How do I change my password?"
  },
  {
    "id": 5,
    "customer": "Eli",
    "text": "The parcel arrived but the box was crushed"
  }
]

Each ticket is a dictionary with an id, a customer and the text they wrote. json.dump writes the whole list to tickets.json, and reading the file gives the list back.

with open("tickets.json") as f:
    tickets = json.load(f)

print(len(tickets))
for ticket in tickets:
    print(ticket["id"], ticket["text"])

5
1 I was charged twice for one order
2 My parcel has not arrived
3 Can I get a refund for the blue mug?
4 How do I change my password?
5 The parcel arrived but the box was crushed

json.load, without the s, reads from an open file. The file becomes a list of dictionaries, the shape from Lists of dictionaries.

Saving a summary with json.dump

counts = {"tickets": len(tickets), "customers": [t["customer"] for t in tickets]}

with open("summary.json", "w") as f:
    json.dump(counts, f, indent=2)

with open("summary.json") as f:
    print(f.read())

{
  "tickets": 5,
  "customers": [
    "Asha",
    "Ben",
    "Chen",
    "Dara",
    "Eli"
  ]
}

json.dump writes to an open file, and reading the file back shows the JSON it wrote.

JSON vs Python spelling

Where JSON shows up in AI code

• Asking a model to answer in JSON, then reading the answer with json.loads.
• The body of almost every web API request and response.
• Saving results, settings and test cases to files.

Related

• Previous: Reading and writing files
• Next: Stand-in model
• Reference: The json module in the Python docs