Bangladesh Leads AI Clothing Edits at 35.2%, India People Prompts at 65.5%, Pakistan and US Meme Prompts at 0.7%, Google Study Finds

Fact-Checked by Irfan Ahmad

People in different countries use AI image tools in different ways, with country-level patterns emerging in requests involving people, clothing, lighting, memes and image composition, according to a Google research paper submitted to arXiv on November 5, 2026, authored by Will Tucker-Ray, Marika Swanberg and 10 other researchers.

Pakistan recorded a 0.7% prevalence of meme-style AI image prompts, followed by the United States at the same percentage. The Philippines, Thailand and India each recorded 0.6%, while Vietnam registered 0.5%. Mexico, Indonesia, Egypt and Brazil each recorded 0.4%, according to the study of Gemini conversations involving Nano Banana.

India had the highest prevalence of people in prompts, at 65.5%, followed by Iraq at 65.3%, Somalia at 65.0%, Nigeria at 63.5% and Indonesia at 63.0%. Malaysia recorded 62.6%, Suriname 62.1%, Nepal 61.8%, the Philippines 61.7% and Congo 60.8%.

Bangladesh led clothing and apparel modification prompts, with 35.2%, followed by Morocco at 29.3%, Egypt at 28.4%, Sri Lanka at 27.9% and Iraq at 27.7%. India ranked seventh at 26.2%, followed by Saudi Arabia at 26.0%..

Somalia recorded the highest prevalence of lighting modifications at 28.4%, followed by Malaysia at 28.3%, India at 28.1%, Nigeria at 27.7% and Sri Lanka at 27.5%.

The United States ranked tenth for precise positioning prompts, at 28.1%. It also ranked third for prompts involving imitation of art, people or characters, at 4.9%, and fifth for absurdity or humor, at 2.1%.

The researchers explained that they examined "some differences in the geographic distributions of the topics" and "styles of image generation and editing (lighting, meme formats, precise positioning)." They also compared how users interacted with the tools, including "full re-prompting or not."

Google Study Identifies Country-Level Differences in AI Image Generation and Editing
Image: How are viral trends different from educational and commercial uses of LLM image generation and editing? A privacy-preserving descriptive analysis. CC BY

In social and viral conversations, "The highest prevalence (at nearly 80%) is the inclusion of people." The researchers also found that "Full reprompts are about twice as prevalent as simple iterations (e.g., “try again”) in social and viral prompts." Meme style or format and absurdity or humor were the least prevalent features in this category.

The study analyzed a random worldwide sample of 2.1 million Gemini conversations containing more than 5.2 million individual prompts, collected over 90 days in the first quarter of 2026.

The researchers cautioned that their country comparisons included only "countries with" more than "100 conversations with those features." They also stressed that "These are measures of intents, not outputs," and described their findings as "directional and to an extent ordinal, but not necessarily precise."

DIW's Two Cents: Whether creating AI images for personal profiles, entertainment or creative expression, users should respect the dignity, privacy and rights of people depicted. Altering someone's appearance or identity without consent, or to humiliate, mislead or damage their reputation, raises ethical concerns. Responsible use also means respecting modesty and personal boundaries.

Read next: 

• Influencers are supposed to influence us – so why are they asking us what they should do?

• I don’t want AI in my online search. What options do I have?
Previous Post Next Post