OpenAI: Web search in the Responses API now returns images alongside text results

On June 9, 2026, OpenAI announced an upgrade to the web search tool within the v1/responses endpoint of the Responses API: in addition to the usual text results, search now also returns image results. The change is intended for applications where visual context makes responses higher quality and more complete.

What changed in the web search tool?

Until now, the web search tool within the Responses API returned exclusively text results — page titles, relevant content snippets, and URLs. Users who also needed visual materials had to implement separate image-retrieval steps or redirect users to external visual searches.

Starting today, the same web search tool call can return images found on the web alongside text. Images are integrated into the standard tool response — there is no separate endpoint, no need for a special flag or beta access key. The change is automatically active for all existing and new implementations using web search within v1/responses.

The images returned are those that the web search finds in real time on publicly accessible pages — they are photographs from the web, not images generated by an AI model. The quality and availability of results is subject to the usual variables of web search: the quality of indexed content and source availability.

Use cases where images alongside text add value

OpenAI identifies several categories of cases in which returning images alongside text results concretely improves the user experience:

Product photos — e-commerce, price comparison, or catalog review applications can display a visual representation of an item alongside its text description, without a separate call to an image API or photo database.

Landmarks and locations — travel guides, trip planners, and city exploration apps can display a photo or illustration of a location alongside its text description, without any additional API calls.

Events and people — journalism tools, news summarizers, or assistants monitoring current events can display visual context — an event photo or a person’s portrait — together with textual information, without additional API calls.

Visual references — any application where the user asks “what does X look like” or “show me Y” can now respond in an integrated manner, rather than redirecting the user to a separate visual search.

Technical notes for developers

For developers already using web search within the Responses API, the change is transparent at the code level — it requires no changes to existing implementations. Images are returned as part of the normal tool response.

Complete documentation on the response structure when image results are included is available in OpenAI’s web search guide, which was updated alongside this changelog entry.

---

The upgrade is targeted and practical: for developers building visually rich applications, it reduces the number of required API calls and simplifies solution architecture. Instead of separately orchestrating image retrieval as an additional workflow step, the web search tool now handles that within the same call. For applications that rely on web-based visual content — especially in the product, location, and event categories — this is a small but welcome improvement.