Does search_context_size set an exact source count?

Definition: search_context_size is an optional setting on OpenAI's web search tool. It gives high-level guidance about how much context from search results is made available to the model before it generates a response; it does not specify an exact number of sources.

Key facts

• OpenAI's web search documentation defines search_context_size as a control for the amount of search-result context available before response generation.
• The setting accepts low, medium, or high; OpenAI's Responses API reference says that medium is the default.
• The documentation describes low for simple lookups, medium as a balanced default, and high for answers that may need more detail from search results.
• The web search guide states that the setting does not set an exact token count or a specific number of sources or citations.
• The model's response can include inline citations and url_citation annotations for URLs found in web search results.

What does search_context_size control?

OpenAI's web search guide describes the setting as guidance about the amount of context taken from web search results before the model generates its answer. The three values describe intended levels of search-result detail: low suits a simple lookup, medium is the balanced default, and high suits an answer that may require more detail.

The Responses API reference describes the field as optional and lists the same three allowed values. It also states that medium is the default. The labels therefore describe a context-size preference for the search tool, rather than a result-list length that you can count in advance.

Does it determine how many sources appear?

No exact source count follows from this setting. OpenAI's documentation says that search_context_size does not set an exact token count or specify a specific number of sources or citations. A request using low can therefore not be read as a request for one source, and high does not define a fixed upper bound.

The guide describes citations as response data. A web-search response contains a web_search_call item describing the action and a message item containing the text and citation annotations. By default, the model response includes inline citations for URLs found in the search results, while the url_citation annotation carries the URL, title, and location of the cited source. The output shows what the request produced; the context setting alone does not provide a source-count promise.

How do you set it in a request?

OpenAI's guide places the setting inside the web_search tool object in the Responses API tools array. A request can use this shape:

{
  "type": "web_search",
  "search_context_size": "low"
}

The same guide says that the model can choose whether to search the web based on the input prompt, just as it can choose whether to use another tool. Enabling the web search tool and selecting a context level therefore describes the available search configuration; it does not require every request to produce a search call.

Related terms

• ChatGPT search
• Where can I see the sources behind a ChatGPT search answer?
• ChatGPT deep research
• Does ChatGPT always search the web before answering?

FAQ

Does low mean that OpenAI uses one source?

No. The guide defines low as a choice for simple lookups and separately says that the setting does not specify a number of sources or citations.

Is medium the default?

Yes. OpenAI's Responses API reference lists low, medium, and high as the allowed values and states that medium is the default.

Does setting a context size force a web search?

No. The web search guide says that the model can choose whether to search based on the input prompt. The setting applies when the web search tool is used.

Sources

1. OpenAI Developers, "Web search", undated
2. OpenAI Developers, "Create a response", undated

---

Check how a page reads to answer engines. Open the GEO score analyzer