Search Results in ChatGPT, Claude, and Perplexity LLM Responses API

The sources an AI model cites tell only part of the story. The web results it discovers along the way can be just as valuable, revealing which content gets discovered even when it doesn’t make it into the final answer.

With the new search_results field in DataForSEO’s ChatGPT, Claude, and Perplexity LLM Responses APIs, you can go beyond citations and explore the web results retrieved by LLMs during the answer-generation process.

This update will help you better understand AI-driven content discovery and uncover new GEO opportunities.

What’s new

We’ve introduced the search_results field in the following LLM Responses endpoints:

➤ ChatGPT LLM Responses: Live and Task GET.
➤ Claude LLM Responses: Live and Task GET.
➤ Perplexity LLM Responses: Live.

The new field returns an array of web search results retrieved by the LLM while it was gathering information for its response.

Each search result includes:

• title: title of the retrieved web page.
• url: URL of the retrieved web page.
• date: publication or indexing date of the page.
• snippet: text excerpt from the search result.

Example:

{
  "search_results": [
    {
      "title": "Amusement Parks Market Size, Forecasts Report 2026-2035",
      "url": "https://www.gminsights.com/industry-analysis/amusement-parks-market",
      "date": "2026-10-30 00:00:00 +00:00",
      "snippet": "# Amusement Parks Market Size & Share 2026-2035n## Amusement Parks Market SizenThe amusement parks market was valued at USD 80.5 billion in 2025, reached USD 86.1 billion in 2026, and is expected to grow at a 5.7% CAGR (2026–2035) to reach USD 142.1 billion by 2035.n...n## Amusement Parks Market Reg..."
    },
    {
      "title": "Pourquoi la France continue d’attirer toujours plus de parcs",
      "url": "https://www.24matins.fr/pourquoi-la-france-continue-dattirer-toujours-plus-de-parcs-1419544",
      "date": "2026-10-10 00:00:00 +00:00",
      "snippet": "Le *SNELAC* recense pourtant un marché encore en progression, avec environ **650 parcs de loisirs** et **70 millions de visiteurs** attendus en 2025, contre seulement 3 millions en 1980.n...n*Disneyland Paris* reste de loin le poids lourd, avec 16 millions de visiteurs pour ses deux parcs.n*Le Parc ..."
    },
    {
      "title": "Sales for financial year 2024/25",
      "url": "https://echanges.dila.gouv.fr/OPENDATA/AMF/ECO/2025/10/FCECO080060_20251028.pdf",
      "date": "2026-08-20 00:00:00 +00:00",
      "snippet": "Paris, October 28, 2025 – (Unaudited data) – Compagnie des Alpes reported record sales of €1,397 million for fiscal year 2024/25 (ended September 30, 2025), an increase of 12.8% compared with the previous year.n...nOur mountain operations posted significant growth for the third consecutive year, and..."
    }
  ]
}

The key difference between search_results and sources:

• search_results includes all web results retrieved by the LLM during its search process, regardless of whether they appear in the final answer.
• sources represents the results the LLM actually cites in its generated response.

Together, they provide a broader view of what an LLM finds and what it ultimately chooses to reference.

Why it matters

By looking beyond cited sources, you can better understand where your content appears in the AI search process and identify opportunities to improve its visibility.

With search_results, you can:

• Discover which websites and pages LLMs retrieve, even when they aren’t cited.
• See which retrieved pages are referenced in the answers and which are left out.
• Identify competitors in LLM search results and examine their content.
• Find pages and topics that may help inform your AI visibility strategy.

Explore our updated API documentation to get started.