Google mapped 2.1 million Nano Banana use cases without images

Google researchers analyzed 2.1 million Gemini Nano Banana conversations, finding that people were the leading feature in nearly 80% of social and viral cases without examining the images.

They applied Google’s existing Urania framework, already used for insight into LLM chat interactions, to patterns in text prompts and conversation sequences. That gave the paper a privacy-preserving route to study use while people submit sensitive material involving themselves and their families.

The paper states: “Beyond proactive self-reports in traditional media and on social networks, there is limited data on how and why many people are using LLM image generation.” The authors described usage as growing, but the paper supplied no separate adoption figure.

Five categories

The analysis placed conversations in five groups:

• Social & Viral
• Brand & Marketing
• Product Visualization & E-Commerce
• Architectural & Spatial Design
• Educational & Technical Information

Researchers compared each group across artistic style, image editing, composition, interactions and multi-select features. The classifications captured:

• Image editing: Removing objects or people, changing backgrounds, adding elements, resizing and restoring images.
• Artistic style: Watercolors, pencil sketches, photography, oil paintings, minimalist looks, film noir and vintage moods, plus conceptual mashups. Other labels covered absurdity or humor and imitation of an artist, public figure or fictional character.
• Composition: Lighting, perspective, constraints, including what not to show, and requests for people and how many.
• Interactions and sessions: The full conversation, including how users iterated on and edited generated images.
• Multi-select features: Text and typography requirements, transformation types, work context and content-creation stages.

The text taxonomy had four prompt types:

• Full Specification: Precise style and wording.
• Precise Content Only: Supplied text without a specified style.
• Stylistic/Generic Only: Described the intended vibe without specific details.
• N/A: Did not mention text or fonts.

What 2.1 million conversations reveal

Inclusion of people was the most prevalent feature in the Social & Viral group, at nearly 80%. Composition and artistic-style requests were also common. Requests to start over were about twice as prevalent as requests for minor changes.

By contrast, memes and humor were rare. The researchers wrote: “Perhaps surprisingly, meme style or format and absurdity / humor are the least prevalent.”

The study therefore found uses beyond trivial viral content: brand and marketing work, educational illustrations, technical instructions and product visualization.

Precision dominates brand and marketing

Marketing conversations most often controlled composition, including subject placement, people, lighting and instructions about what should not appear. The group also had the highest prevalence of requests for exact text and detailed typography.

The paper interprets that pattern as a demand for promotional images with specific text and design requirements, rather than requests for an attractive image without production constraints.

Brand & Marketing also had the highest prevalence of conversations involving small businesses, individuals or side hustles. Much of the brand activity concerned original, fictional or unrecognized brands rather than established companies. Researchers suggested that some users may be developing branding for new or speculative businesses; that remains an interpretation, not a measured outcome.

Commercial activity extended beyond first-generation images. Users refined logos, changed text and colors, created promotional materials and developed consistent visual identities for online engagement and traditional promotion.

What does this change for marketers and creators?

The practical answer is that Nano Banana already reaches commercial, educational, design and technical work, not only viral entertainment. The researchers said tasks that once required professional experience are now within reach of more people. For practitioners, typography, composition, editing and iterative control matter more than novelty alone.

That does not establish a budget case. The paper reports no prices, campaign reach, conversion rates, output quality or time saved, and the researchers said it remains unclear whether the tool enables entirely new activities or simply makes existing ones less expensive.

They expect additional uses to emerge as users continue experimenting. For teams, the evidence supports testing structured prompts and repeatable brand workflows while treating savings, audience growth and campaign performance as questions their own projects must answer.