AI-altered images on birdwatching forums putting research at risk

For many birdwatchers, seeing a bird species in a location where it doesn't normally live is a very exciting event. In the UK, such sightings often make national news. For instance, the western reef heron, which is typically found in Africa and southern Europe, was seen in a coastal town in north Wales in June and shared widely on birding websites.

However, a new problem is threatening to spoil the enjoyment: the misuse of AI in image editing, sometimes called AI slop.

Scientists are asking birdwatchers to reduce their use of AI when editing photographs. They are concerned that this could damage the trustworthiness of popular citizen science platforms. Websites like iNaturalist and Macaulay Library are regularly used by scientists to study where different species live.

The development of generative AI tools, such as ChatGPT and Google Gemini, has resulted in a significant rise in fake or improved images of rare and remarkable birds appearing on wildlife photography sites. People can create very realistic fake images in just seconds. They can also enhance a photo by asking the AI to remove objects like a branch or leaf that might be blocking the view of the bird. This process can unintentionally change the image in important ways.

In a recent article published in the scientific journal Nature, researchers expressed concerns that hundreds of fake images have already been discovered on popular databases used for recording species. They stated that the full extent of this problem is unknown, as many fake images might not be identified, and this could affect the accuracy of data collected by the public.

Dr Alexander Lees, an ecologist at Manchester Metropolitan University and one of the authors of the journal article, commented that much of the wildlife photography seen on social media nowadays appears to be entirely AI-generated. He added that this makes it very difficult to use these photos to understand the geographical distribution and timing of species populations.

Lees mentioned that completely fake images are still uncommon and usually easy to detect; for example, no one is likely to believe a report of a toucan sighting in Siberia. However, he explained that birdwatchers often use AI to edit and improve their pictures. During this process, the AI might add features from different bird species, creating a new, inaccurate image.

Lees gave an example of a false sighting of a red-winged blackbird in central Brazil. This species is normally found in North America and had never been recorded in that part of Brazil before the supposed sighting. In reality, the bird was an epaulet oriole, a common bird species native to the Americas. The photographer had asked an AI platform to make the picture look better, and this led to parts of a red-winged blackbird being added, resulting in the inaccurate report.

Lees cautioned that while wildlife photographers often aim for beautiful images, there is a danger that edited photos could cause problems later, especially when AI has been used in the editing process.

Organisations involved in citizen science are still trying to figure out how widespread this issue is. On iNaturalist, a social platform where nature lovers can log sightings of plants and animals, only about 1,400 out of over 610 million images have been marked as potentially using AI. Citizen science has led to many discoveries, from tracking how plants and animals move due to climate change to observing new behaviours in species.

Tony Iwane, iNaturalist’s director of community support and a co-author of the Nature paper, suggested that most of these cases are probably not intentional deception. However, he urged users to remain watchful.

He explained that on platforms like iNaturalist, ordinary people share information that scientists would find very difficult to gather in large quantities. He also described this data as being like a real-time sensor of global events: noting if plants are flowering earlier or if species are migrating northwards as the climate warms. He emphasised that knowing the location of species more accurately helps conservationists make better decisions, but this information must be reliable.