JustSERPDocs

ChatGPT

What chatgpt.com answers a logged-out visitor, in a fresh session for each request: the reply, the businesses on its place cards and the sites it cites, also under the fields `markdown`, `text_blocks`, `reference_links`, `cards` and `response_metadata`.

GET
/v1/chatgpt

What chatgpt.com answers a logged-out visitor, in a fresh session for each request: the reply, the businesses on its place cards and the sites it cites, also under the fields markdown, text_blocks, reference_links, cards and response_metadata. location.country sets the country the question is asked from; the coordinates are reported as the visitor's position; language is not read. Name the town in query: a "near me" question is answered for wherever the request was made from.

In the document form. $1 request $2 q and no query is the document-form ChatGPT call (q, country, web_search, expand_entities), answered as the document: search_metadata, search_parameters, text_blocks, markdown, reference_links, cards (knowledge_entity) and response_metadata (message_id, is_web_search_performed, is_truncated); no search_information. country always sets the country the question is asked from (the US when left out). web_search is echoed and may not change the page; expand_entities is accepted and does nothing. Not returned (not on the logged-out page): citations (character offsets), inline_images, web_results, search_queries, a reference's ref_id, response_metadata.model and .conversation_id, a card's details, widget cards.

Authorization

headerX-API-Key<token>

Your API key.

Query Parameters

cancel_token?string

A name you make up for this request (a UUID is fine), so you can later GET /v1/cancel with it. Only you know it, so only you can cancel. Use a new one for every request, retries included: a cancelled token stays cancelled for a few minutes.

Length1 <= length <= 128
query?string

The question, as it is to be asked. 1 to 2000 characters after trimming.

Length1 <= length <= 2000
location?

Where the question is asked from. Every part is optional; a part left out, null or "" is not an error. Latitude and longitude are sent together or not at all.

q?string

The prompt.

Length1 <= length <= 4000
country?string

The country the question is asked from (the US when left out). A country that cannot be used is asked from a default location.

Default"us"
expand_entities?boolean

Accepted and echoed; it does nothing.

Response Body

The answer; with q, the document.

application/json
  1. response

The AI answer, and the same answer in the fields markdown, text_blocks, reference_links, cards and response_metadata.

search_metadata*
search_parameters*

The request's own fields as sent, engine first, without cancel_token.

search_information*

The parser's own, e.g. yelp_search's; otherwise {}.

answer_status?string

no_answer only from google_ai_overview in practice; for the others a page with no reply is parsed: false.

Value in"answered""no_answer"
model?string|null
answer_text?|

Plain text. Null with no_answer.

answer_markdown?|

Headings, lists, bold, tables; code fenced with its language, formulas as $LaTeX$. Prefer it for display.

businesses?array<>
citations?array<>
location_used?|null
markdown?|

The same as answer_markdown.

text_blocks?array<>
reference_links?array<>
cards?array<>
response_metadata?
Example Request
curl -X GET 'https://api.justserp.com/v1/chatgpt?query=Who+has+the+best+pizza+near+Chelsea%2C+New+York%2C+NY%3F&location%5Bcountry%5D=us&location%5Blanguage%5D=en'
Example Response
{  "search_metadata": {    "id": "string",    "status": "Success",    "created_at": "string",    "request_url": "string",    "html_url": "string"  },  "search_parameters": {},  "search_information": {},  "answer_status": "answered",  "model": "string",  "answer_text": "string",  "answer_markdown": "string",  "businesses": [    {      "position": 0,      "name": "string",      "snippet": "string",      "place_id": "string",      "data_id": "string",      "data_cid": "string",      "kgmid": "string",      "website": "string",      "address": "string",      "phone": "string",      "latitude": 0,      "longitude": 0,      "rating": 0,      "reviews": 0,      "citation_indexes": [        0      ],      "maps_url": "string"    }  ],  "citations": [    {      "position": 0,      "url": "string",      "title": "string",      "snippet": "string",      "source": "string",      "date": "string"    }  ],  "location_used": {    "latitude": 0,    "longitude": 0,    "name": "string",    "source": "string"  },  "markdown": "string",  "text_blocks": [    {      "type": "paragraph",      "answer": "string",      "items": [        null      ],      "headers": [        "string"      ],      "rows": [        [          "string"        ]      ]    }  ],  "reference_links": [    {      "index": 0,      "title": "string",      "link": "string",      "source": "string",      "snippet": "string",      "date": "string"    }  ],  "cards": [    {      "type": "knowledge_entity",      "name": "string",      "category": "string",      "disambiguation": "string",      "id": "string"    }  ],  "response_metadata": {    "model": null,    "conversation_id": null,    "message_id": "string",    "is_web_search_performed": true,    "is_truncated": true  }}

Hiring Cafe job description GET

The description of one Hiring Cafe posting, as the company wrote it (HTML), as text and as markdown.

Facebook event GET

One public Facebook event read with no login: title, time, place with coordinates, hosts, description, ticket link, cover image, category, how many people responded, suggested events.