AI Image Tool Depicts Farms as Idyllic Pastures
A study finds OpenAI's DALL·E 3 rarely shows indoor livestock production, largely because an automatic prompt-rewriting step removes it from generated images.
Updated on Nov. 10, 2026 in Artificial Intelligence
Live Poll
Do you trust AI image tools to show realistic pictures of everyday things?
Researchers found DALL·E 3 depicted dairy farms as green pasture in 94% of images and pig farms with outdoor animals in 91% of images across 4,800 analyzed pictures. An automatic prompt-rewriting step appears to have largely erased indoor farming from the generated scenes.
Why it matters
Farm imagery generated by AI can misrepresent how most livestock are actually raised, potentially shaping public perception of modern agriculture. The researchers warn such scenes could also feed future AI models and mislead users.
Researchers analyzed 4,800 generated images, and when prompt revision was inhibited, 96% of pig farm images and 60% of dairy images showed animals indoors. The study did not examine how the pattern would differ outside Western farming systems.
The players
OpenAI
OpenAI is the San Francisco artificial intelligence company that developed DALL·E 3 and the ChatGPT image generation system.
The details
The rewriting step, powered by GPT-4, expands short user requests into fuller prompts before passing them to DALL·E 3, and OpenAI built it to make images fairer and more varied. When the researchers added a developer-guide line instructing the system not to add detail, indoor farming reappeared in most images, with pasture showing up in only 4% of American dairy images.
Timeline
2012: USDA data showed more than 99% of nursery and grower/finisher pigs were housed without outdoor access.
March 2024: ChatGPT refused battery cage depiction requests in two screenshots.
10 April 2025: OpenAI made GPT-4o image generation ChatGPT's default.
4 November 2026: OpenAI launched ChatGPT Images 2.6.
The Tech Race
The study measures DALL·E 3's farm imagery against 2012 USDA housing data showing over 99% of US nursery and grower/finisher pigs had no outdoor access. The gap suggests that a rewriting layer built to improve image diversity can systematically distort depictions of real-world industries.
Anyone using AI image tools to illustrate food, farming, or agricultural topics may receive depictions that do not match how most animals are raised. Checking generated farm imagery against verified sources before publishing or sharing helps avoid spreading an inaccurate picture of modern livestock production.
The takeaway
Generated images can carry the biases of their rewriting pipelines, so viewers should treat AI depictions of real-world industries as illustrations rather than evidence. Cross-checking against official statistics remains the reliable way to understand how food is actually produced.
Further reading
For more on how generative tools are reshaping visual content, explore Artificial Intelligence coverage.
Source note: This article includes information reported by International Business Times UK.
Live Poll
Do you trust AI image tools to show realistic pictures of everyday things?
