Can AI Apps Really Identify Mushrooms? The Science Says Not Yet

The study, conducted by researchers at the University of Vienna and published in Nature Scientific Reports, represents the most rigorous independent evaluation of mushroom ID apps to date. The researchers photographed mushrooms in the field under realistic conditions — varying angles, lighting, and stages of maturity — and submitted the images to the top-ranked apps on iOS and Android.

The apps tested included Picture Mushroom (the most downloaded app globally), Mushroom Identificator, Shroomify, iNaturalist, and Google Lens. Each image was submitted multiple times to account for variation in results.

Key findings:

• The best-performing app (Picture Mushroom) failed to provide the correct species as its top result in approximately 15% of tests.
• No app consistently returned a single correct answer. Most presented a ranked list of possibilities, and the correct species was sometimes buried below incorrect suggestions.
• Performance was worst for species that are visually similar to unrelated species — exactly the cases where accurate identification matters most.
• Apps performed significantly worse on immature specimens, old or degraded specimens, and mushrooms photographed from above (without gill or stem detail visible).
• None of the apps provided reliable warnings when presented with deadly species like Amanita phalloides or Amanita virosa.

The 2026 study builds on several earlier evaluations that painted a similarly concerning picture.

A 2022 study by Pocock and colleagues tested three popular apps using 78 images of UK mushroom species. The best performer (Picture Mushroom) achieved only 50% accuracy overall. For toxic species specifically, accuracy dropped to 44%. The study found that apps were most likely to fail on the species where a mistake would be most dangerous.

A 2024 study published in Toxicon focused specifically on apps' ability to identify the 15 most commonly involved species in European mushroom poisonings. The results were alarming: several apps identified deadly death caps as edible field mushrooms, and Galerina marginata (deadly) as honey mushrooms (edible).

A separate 2023 analysis by the French poison control network reviewed mushroom poisoning cases where patients reported using an app. Over a two-year period, they identified 17 cases of hospitalization where patients had specifically stated that an identification app had confirmed their mushroom as safe before they ate it.

The concern about mushroom ID apps is not theoretical. Poison control centers across Europe and North America have documented cases where patients cited app results as a factor in their decision to eat a toxic mushroom.

In 2022, a couple in France were hospitalized with amatoxin poisoning after eating mushrooms they had identified using an app as Macrolepiota procera (parasol mushroom). The mushrooms were actually Chlorophyllum molybdites, the most commonly misidentified toxic mushroom in North America and increasingly found in Europe. Both survived, but required several days of hospitalization.

In British Columbia in 2023, an experienced hiker ate mushrooms he identified using Google Lens as chanterelles (Cantharellus cibarius). They were actually Omphalotus olearius (jack o'lantern mushroom), which causes severe gastrointestinal distress. He spent two days in the hospital.

Perhaps the most insidious risk is overconfidence. A 2023 survey of app users found that 62% said they would feel “confident or very confident” eating a mushroom identified as edible by their app. Among users who had the app for more than six months, that number rose to 78%. The apps are not just failing to identify mushrooms accurately — they are creating a false sense of security that leads people to take risks they otherwise would not.

In 2023, the New York Mycological Society raised an alarm about a flood of AI-generated foraging guides appearing on Amazon. These books, many of them written entirely by large language models with AI-generated illustrations, contained dangerously inaccurate information about mushroom identification and edibility.

Some books paired images of toxic species with descriptions of edible ones. Others described species that do not exist, or blended characteristics of multiple species into fictional composites. A particularly egregious example described a “forest chanterelle” with characteristics matching Amanita muscaria as “a choice edible with a mild, nutty flavor.”

The problem highlighted a fundamental limitation of generative AI: these systems produce plausible-sounding text without any understanding of whether it is correct. When applied to mushroom identification, where a single error can be lethal, this limitation becomes life-threatening.

Amazon eventually removed some of the most flagrantly dangerous titles after media coverage, but new AI-generated foraging books continue to appear. The Mycological Society of America issued a formal statement in 2024 warning against relying on any foraging guide published after 2022 without verifying the author's credentials and the publisher's reputation.

Mushroom identification is one of the hardest problems in visual classification, for reasons that go well beyond what a camera can capture.

Despite their limitations, mushroom ID apps are not useless. They can be a valuable starting point — a way to narrow down possibilities and point you toward the right section of a field guide. The key is understanding what they can and cannot do.

• Never eat a mushroom based solely on an app result. This is the single most important rule. An app should be the beginning of your identification process, not the end.
• Use apps as a “second opinion” tool. If you have a tentative ID based on field characteristics, an app can help confirm or challenge that ID. But an app result alone is not sufficient.
• Photograph from multiple angles. Capture the cap from above, the gills from below, the stem, the base (including any volva or bulb), and a cross-section if possible. Single top-down photos produce the worst results.
• Note what the app cannot see. Record the smell, substrate, habitat, bruising reaction, and any other sensory details. These are often the characteristics that separate deadly from edible.
• Cross-reference with field guides. Use the app's suggestion as a starting point, then verify against a reputable regional field guide by a credentialed mycologist.
• Join a local mycological society. There is no substitute for learning from experienced people in the field. Most local societies offer free or low-cost forays where you can learn to identify mushrooms hands-on.

AI identification technology is improving. Newer models trained on larger, more diverse datasets are showing incremental gains in accuracy. Some apps are beginning to incorporate geographic and seasonal data, which helps narrow identifications to locally plausible species.

Multi-image analysis, where an app requests photos from multiple angles before attempting identification, is being adopted by some platforms. This approach better mimics how a human expert examines a specimen and reduces errors caused by single-angle ambiguity.

Community-powered platforms like iNaturalist combine AI suggestions with human expert review, which produces significantly more reliable identifications than AI alone. The tradeoff is speed — expert review can take hours or days, which does not help someone standing in the woods deciding whether to pick a mushroom for dinner.

Some researchers have proposed that apps should include explicit confidence scores and danger warnings. If an app cannot distinguish between an edible species and a potentially lethal lookalike, it should say so clearly rather than presenting its best guess as a definitive answer.

At Orangutany, we approach mushroom identification with this philosophy: our AI scan is designed as a starting point for learning, not as a definitive field guide for foraging decisions. We present multiple possible matches, link to detailed species information including lookalikes and toxicity data, and explicitly warn users never to eat a mushroom based on any app result alone.