LLM Response

Parameters

Example Request

curl -X POST https://api.localseodata.com/v1/ai/llm-response \
  -H "Authorization: Bearer sk_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Who is the best plumber in Austin, Texas?",
    "platform": "chat_gpt"
  }'

Example Response

{
  "status": "success",
  "credits_used": 8,
  "data": {
    "platform": "chat_gpt",
    "prompt": "Who is the best plumber in Austin, Texas?",
    "response_text": "Based on reviews and ratings, some of the best plumbers in Austin include ABC Plumbing (4.9 stars, 500+ reviews), Radiant Plumbing (4.8 stars), and Stan's Heating & AC (4.7 stars)...",
    "model": "gpt-4.1-mini-2025-04-14",
    "sources": [
      { "url": "https://yelp.com/...", "title": "Best Plumbers in Austin", "domain": "yelp.com", "position": 1 },
      { "url": "https://abcplumbing.com", "title": "ABC Plumbing Austin", "domain": "abcplumbing.com", "position": 2 }
    ],
    "fan_out_queries": [
      "best plumber austin reviews",
      "emergency plumber austin tx",
      "cheapest plumber in austin"
    ]
  }
}

Response Fields

Notes

• Query each LLM individually to see how they recommend (or ignore) your business.
• fan_out_queries show what users commonly ask next — use these for content planning.
• sources come from the citations a model attaches when it searches the web, so how full this array is depends on the platform. perplexity searches as part of answering, so sources are populated on essentially every call. chat_gpt, claude and gemini answer from training data unless the question makes them reach for the web — on local business prompts they usually don't, so expect sources to be empty there. That is the model's behaviour, not a gap in the response.
• On gemini, domain is the registrable domain and never the subdomain. Google's grounding data returns a redirect URL plus a title carrying the bare domain, and the subdomain is not transmitted at all — a client hosted at orangecounty.example.com is reported as example.com. Match on the registrable domain when you track a client whose site lives on a subdomain, or those citations will not match. chat_gpt, claude and perplexity return the full hostname, as does /v1/ai/scraper on gemini.
• Web search support differs by platform, and asking for something a platform does not offer returns a 400 saying what it does. claude takes all four search parameters. chat_gpt takes all four, except that GPT-5 models drop force_web_search. gemini takes web_search only — a search can be permitted but never compelled, and cannot be geographically scoped. perplexity takes search_country only and cannot be switched off.
• perplexity is the value option for citation work. Its models search on every call at no extra charge, and in our testing it returned 19 citations across 10 distinct domains for one local query — more sources and more diverse ones than a search-enabled claude call that cost eight times as much. If you are measuring which domains get cited, start there.
• If you want citations for ChatGPT or Gemini without paying the search price, /v1/ai/scraper returns a sources array for those two platforms at 4 credits, less than this endpoint costs.
• We send your prompt and the model name, and leave sampling settings at each platform's own defaults — which differ between platforms, and are not all low. The same prompt therefore returns different wording run to run, by a different amount on each platform. If you track whether a brand gets mentioned over time, expect day-to-day movement that reflects sampling rather than any change in what the model thinks, and note that comparing one platform's trend against another's is affected by this too. Repeat a prompt a few times on day one to see how much movement is normal for the platforms you track, before reading a single day's change as a signal.
• Supports: chat_gpt, claude, gemini, perplexity. Run all 4 to see the full picture.