To the Editor,
There is a significant lack of diversity in skin tones when depicting dermatological conditions in educational and online materials. Google Images is a reference tool often used by both physicians and patients to diagnose skin, hair, and nail conditions.1,2 A 2020 study on the representation of skin of color (SOC) in Google Images of common skin conditions found that 90.5% of searches underrepresented SOC patients concerning the racial demographics of the United States.3 This study highlighted the overall dearth of melanin-rich skin representation in dermatology education and reference materials, potentially causing substandard quality of care, treatment delay, and misdiagnosis for patients with darker skin tones.4 Recently, websites such as the American Academy of Dermatology (AAD), DermNet NZ, and VisualDx have made efforts to be more inclusive with their photos.4 The purpose of our study was to analyze the representation of melanin-rich skin tone photos in Google Images relative to the US epidemiological data for dermatological conditions. Additionally, we assessed whether there has been an increase in the inclusion of photos of melanin-rich skin since the previous 2020 study.
Forty-two common dermatological conditions compiled based on similar research articles were searched on Google Images using "Incognito Mode" (ie, private browsing), blocking location data, and clearing browsing data and cookies. Three independent reviewers analyzed screenshots of the first 100 photos for each search and categorized them as either light-medium skin tone (Fitzpatrick I-IV) or melanin-rich skin tone (Fitzpatrick V-VI).
Forty-two common dermatological conditions compiled based on similar research articles were searched on Google Images using "Incognito Mode" (ie, private browsing), blocking location data, and clearing browsing data and cookies. Three independent reviewers analyzed screenshots of the first 100 photos for each search and categorized them as either light-medium skin tone (Fitzpatrick I-IV) or melanin-rich skin tone (Fitzpatrick V-VI).
A total of 4,182 photos were analyzed; only 11.7% of these photos were classified as melanin-rich. In 52% of the searches, there were no photos of melanin-rich skin within the first 10 images. Of the first 100 photos in each search, it was found that conditions with classically equal racial distribution such as contact dermatitis, diaper dermatitis, cherry angioma, urticaria, and tinea versicolor showed drastic underrepresentation of melanin-rich skin tones. Furthermore, conditions with a higher incidence in people of African descent, such as central centrifugal cicatricial alopecia which occurs almost exclusively in Black women, only 64% of photos depicted melanin-rich skin. Similarly, traction alopecia, discoid lupus, hidradenitis suppurativa, dermatomyositis, acne, HSV, and subungual melanoma, still heavily feature photos of light-medium skin tones despite having predominance in people of African descent. The only disease in which melanin-rich skin was overrepresented was vitiligo (Table 1).
Representation of all skin colors within dermatology educational and reference materials is paramount to providing quality care to every patient. Although our search revealed that there was an increase in melanin-rich skin representation since 2020 (from 5.7% to 11.4%) there is still work to be done towards creating more inclusive search results on Google Images, especially for conditions in which there is no racial/ethnic predominance.3 Notably, images featuring melanin-rich skin tones with those conditions were represented on the first page only after adding the search term "dark skin" to the query, indicating that these photos do exist but the algorithm still places a preference on lighter skin tones.
