Google has introduced the Search Generative Experience (SGE), which enables users to generate images from text prompts. This feature was announced by the company, and it’s not surprising to see Google following in the footsteps of Microsoft, which has been offering a similar capability through Bing Chat using OpenAI’s DALL-E model since March.
Here’s How It Functions
If you’re part of Google’s Search Labs program and have opted into SGE, you can simply enter your query into the Google search bar. SGE will then generate a selection of images based on your input, and you can choose from among them. This tool is powered by the Imagen family of AI models.
Google demonstrated the tool’s capabilities with a image, illustrating how it can be used to create images based on a specific prompt, such as “draw an image of a capybara wearing a chef’s hat and preparing breakfast.” As seen in the image, you can also refine the queries to influence the images that Google generates.
You will have the ability to create AI-generated images from Google Images. While you are exploring the search results on Google Images, you may come across a prompt box that allows you to generate images based on your query. These generated images will be displayed in a sidebar alongside the search results.

Google’s Commitment to Responsible Image-Generation Tools and Usage Policies
In a blog post, Hema Budaraju from Google emphasizes the company’s commitment to the responsible deployment of its image-generation tools. The tool is designed to ensure that it does not produce images that violate Google’s prohibited use policy for generative AI, as mentioned by Budaraju.
This measure is aimed at preventing the tool from being misused to create images that depict popular cartoon and video game characters engaging in inappropriate activities, similar to an incident involving Bing’s image generation tool.
Each image created using SGE will feature metadata labels and embedded watermarks, indicating that it was generated by artificial intelligence, as noted by Budaraju. During this initial phase, Google will not permit the creation of images that portray photorealistic faces or involve prominent individuals in the prompts, according to Ewer.
When asked about the possibility of allowing such features in the future, Ewer mentioned that it’s premature to make a definitive statement, and the company’s approach will evolve over time. Furthermore, Google has imposed an age restriction on image generation, limiting it to users aged 18 and above, thereby preventing teenagers with access to SGE from using it.
Google’s Enhanced SGE Features
In addition to this, Google is introducing the capability to create written drafts directly from the search bar using SGE. By entering your query, Google can generate a draft, and you have the option to specify whether you’d like the draft to be long, short, or have a casual tone. If you are satisfied with the result, you can export the draft to Google Docs or Gmail.
I inquired about the models driving the draft writing feature from Ewer, and he provided the following response, “SGE is currently driven by a range of LLMs. The models employed in SGE are trained using data from publicly accessible sources as well as human-supervised data.”
Since its introduction in May, Google has been continuously enhancing SGE by incorporating elements like increased video content and improved links. While I personally remain somewhat skeptical of SGE due to its persistent sluggishness and a preference for traditional link scrolling, these newly introduced image and draft creation tools appear to hold promise and potential utility.
Conclusion
Google’s introduction of Search Generative Experience (SGE) lets users generate images and written drafts from search queries.
While Google follows Microsoft’s similar offering with Bing Chat, SGE’s potential utility and responsible image-generation policies highlight its development. Age restrictions and AI watermarks underscore Google’s commitment to ethical use. SGE’s continued enhancements include improved link handling and expanded content support, promising further utility for users.



