ChatGPT state targeting API: AI answers by US state

Who it's for

Multi-location and franchise brands, agencies reporting AI visibility per state, regional service businesses, insurance, legal and healthcare marketers whose providers and rules change at the state line, and retailers checking where ChatGPT sends buyers in each market.

What you get

• ChatGPT and Gemini answer from inside the requested state, in all 50 states and DC
• One field, state, a two-letter code that implies country US
• result.provenance.requestedState on every answer, so each row keeps its state
• The businesses ChatGPT names in map, with address, rating and coordinates, and every brand and product in entities
• GET /v1/states lists the 51 codes, and one batch of up to 500 tasks covers every state for 9 prompts
• The same price with or without a state
• City-level answers on Google AI Mode and Google Search with location

Try it

curl -X POST "https://api.answerline.dev/v1/monitor/chatgpt" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Which car insurance company is cheapest for a new driver?","state":"TX"}'

Sync or queue it with POST /v1/async/task. Full field list on the ChatGPT page and in the API reference.

Engines for this workload

What state targeting does

A person in Houston and a person in Boston who ask ChatGPT “which car insurance is cheapest” can get different companies, different sources and, for local questions, different businesses on the map. Add state to a ChatGPT request and the answer comes from inside that state, as a person there sees it. All 50 states and DC are covered.

On 2026-11-09 we asked ChatGPT “best pizza near me” once in each state and once without a state. Each of the 50 state answers was placed in a city inside the state asked, from Anchorage to San Antonio. Without a state, ChatGPT picked New York. The homepage map shows each recorded answer, and Local AI answers has the table.

Who needs it

• Multi-location and franchise brands: see whether ChatGPT names your location in each state you operate in, and which local rival it names instead.
• Agencies: report AI visibility per state for clients with regional footprints, from one prompt set.
• Regional service businesses: plumbers, roofers, clinics and law firms are recommended by place. A national answer says nothing about Ohio.
• Insurance, legal and healthcare: carriers, rules and providers differ by state, so the answer to “best health plan” or “do I need a lawyer for a DUI” does too.
• Retailers and consumer brands: check which stores and brands ChatGPT recommends in each market.

What each engine does with a state

provenance.requestedState echoes the state on every engine, so a mixed run stays labeled. Read an answer from any engine but ChatGPT as a whole-US answer. The coverage page is the reference.

How it works

1. List the codes with GET /v1/states?country=US: 51 entries, each {code, name}.
2. Send state with a ChatGPT request. Leave country out or set it to "US"; any other country is a 400.
3. Read the answer. map holds the businesses the answer names, with address, rating, latitude and longitude. entities lists every brand, product and local business it names. citationPills and sources show the cited domains and pages.
4. Store result.provenance.requestedState with each row.

One request, synchronous:

curl -X POST "https://api.answerline.dev/v1/monitor/chatgpt" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Which car insurance company is cheapest for a new driver?", "state": "TX"}'

The list of codes:

curl "https://api.answerline.dev/v1/states?country=US" \
  -H "Authorization: Bearer $API_KEY"

An excerpt of a state answer. The values are an example, not a recorded answer; the fields follow the POST /v1/monitor/chatgpt result schema:

{
  "success": true,
  "result": {
    "text": "For a new driver in Texas, quotes from ...",
    "citationPills": [
      { "url": "https://www.example-rates.com/texas", "label": "Texas car insurance rates", "domain": "example-rates.com", "citationPillId": 0, "type": "searchResult", "position": 1 }
    ],
    "entities": [{ "type": "brand", "name": "Example Mutual" }],
    "provenance": {
      "prompt": "Which car insurance company is cheapest for a new driver?",
      "requestedCountry": "US",
      "requestedState": "TX",
      "attempt": 1,
      "complete": true
    }
  }
}

All 51 in one batch

POST /v1/async/task/batch takes up to 500 tasks, so one prompt in every state is one call, and 9 prompts still fit. Each task succeeds or fails on its own.

Python, with the answerline SDK:

import os

from answerline import Client

PROMPT = "Which car insurance company is cheapest for a new driver?"
RUN = "2026-11-11"

with Client(os.environ["API_KEY"]) as client:
    states = client.states()  # the 50 states and DC
    batch = client.create_batch([
        {"taskType": "CHATGPT",
         "payload": {"prompt": PROMPT, "state": s["code"]},
         "idempotencyKey": f"insurance-{s['code']}-{RUN}"}
        for s in states
    ])
    for item in batch["results"]:
        code = states[item["index"]]["code"]
        if not item["success"]:
            print(code, item["error"]["code"])
            continue
        done = client.wait_for_task(item["task"]["id"])
        if done["task"]["status"] != "COMPLETED":
            print(code, done["task"]["status"])
            continue
        result = done["response"]["result"]
        names = [e["name"] for e in result.get("entities", [])]
        print(result["provenance"]["requestedState"], names[:5])

TypeScript, with @answerline/sdk:

import { Client } from "@answerline/sdk";

type Answer = { success: boolean; result: { entities?: { name: string }[]; provenance: { requestedState?: string } } };

const client = new Client({ apiKey: process.env.API_KEY! });
const prompt = "Which car insurance company is cheapest for a new driver?";
const run = "2026-11-11";

const states = await client.states();
const batch = await client.createBatch(
  states.map((s) => ({ taskType: "CHATGPT", payload: { prompt, state: s.code }, idempotencyKey: `insurance-${s.code}-${run}` })),
);

for (const item of batch.results) {
  if (!item.success) {
    console.log(states[item.index]!.code, item.error.code);
    continue;
  }
  const done = await client.waitForTask<Answer>(item.task.id);
  const result = done.response?.result;
  if (done.task.status !== "COMPLETED" || !result) continue;
  console.log(result.provenance.requestedState, (result.entities ?? []).slice(0, 5).map((e) => e.name));
}

The tasks run whether or not you wait; the loop only collects them. For a recurring run, add a webhook to each task instead of waiting (webhooks). How to see ChatGPT answers by US state walks through comparing the 51 answers.

What it costs

A state does not change the price. A ChatGPT request costs $0.0015 at the Starter tier, $1.00–$1.50 per 1,000 across tiers.

A failed request is not charged. A new account’s 100 free requests work on ChatGPT with a state too. Tiers are on pricing.

Pitfalls

• Reading another engine’s answer as a state answer. Only ChatGPT and Gemini answer from inside the state; the rest answer for the whole US.
• One run per state. Answers vary between identical runs, so repeat each state a few times and compare rates (answer volatility). Put the repeat number in the idempotency key, or the repeat is refused as a duplicate of the first.
• Restating the place in the prompt. “Best plumber in Austin” measures the prompt. Keep the prompt the same in every state and let state set the place.
• DC answers can name places in nearby Maryland. Check the addresses in map before counting them as DC.
• Asking for include.html or include.screenshot with a state. ChatGPT refuses both with a 400.

For city grids on Google, see local rank tracking. For the full field list, see the ChatGPT engine page.

More use cases