Google is testing new ways to create images and text directly in Search. The experiments would let some users turn a written prompt into images, refine the results, or draft practical text such as a letter or note while researching a topic.
Search prompts can produce images
In the image experiment, a user can describe a specific scene in the search bar. A request for a capybara wearing a chef's hat and making breakfast, for example, produces four AI-generated images as search results.
Users can then add detail to the description to refine the images. That makes the process conversational: rather than choosing only among existing pictures, someone with a particular idea can adjust a prompt to guide what the generator produces.
Google is also testing a more direct route from image searches to creation. When someone looks for inspiration, such as minimalist Halloween table decorations or spooky doghouse ideas, the generator may appear prominently and let them create an image from their query with a single click. The feature is intended to address searches for specific ideas that may not be represented in conventional image results.
Drafts could move from Search to Google apps
The experiments also extend beyond images. While researching a subject, users may be able to ask AI to draft text such as a letter or note. They can adjust the tone and export the result to Google Docs or Gmail.
This places writing assistance within the search process itself. A user could move from exploring a topic to preparing a piece of text without leaving the Google search experience to start from scratch elsewhere. The source describes the drafts as customizable, though it does not specify which tones or document types will be available.
A limited test with safeguards
Both features are initially limited to English-language use in the US and a small group of people taking part in the SGE experiment. Google's image generator also has an initial minimum age requirement of 18.
Google says it plans to add metadata for AI-generated images and presents that as part of its effort to use AI in Search responsibly. The announcement does not provide further detail about how the metadata will appear or how users will encounter it.
The narrow rollout means these tools are experiments, not a general Search feature available to everyone. Their first audience is limited by language, location, and participation in SGE; the image tool also has an age restriction.
What wider adoption could mean
Putting generation inside a core search product could expose many people to AI tools as they look for information or inspiration. The article says that hundreds of millions of people around the world are likely to encounter generative AI over the long term if these tools expand through Search.
That expansion also raises a practical challenge: the cost of running generative AI at the scale of billions of search queries per day. Google will need to address those costs if it wants to make the tools broadly available. The article suggests the pace of its rollout may also be shaped by competition from Microsoft and OpenAI's chatbot offerings.
The tests fit into Google's wider investment in AI described in the source. It says Google will launch the Gemini multimodal model in the fall, is integrating Bard more deeply into its ecosystem, and says Pixel 8 smartphones are optimized for generative AI capabilities. Bringing image and text creation into Search would put another set of those capabilities where users already ask questions and look for ideas.