What Actually Sells on Gumroad: 1,511 live products across 42 categories (August 2026)

A measured snapshot of 1,511 live Gumroad products across 42 categories, collected on 5 August 2026 with a headless browser against Gumroad Discover search results. For every product the dataset records the category searched, the asking price, the currency that price was displayed in, a USD-normalised price, whether the listing bills as a subscription, its rating count, and the product title. Version 1.1 corrects an error in version 1.0. Gumroad localises the prices it displays, so a single search returns a mixture of currencies — this sample holds 1,239 listings priced in GBP, 228 in USD and 44 in EUR, and 40 of the 42 categories contain more than one. Version 1.0 computed its price statistics across that mixture without converting, which compared values that were not in the same unit. Version 1.1 adds a price_usd column, converted at European Central Bank reference rates for 2026-08-06 (GBP 1 = USD 1.34671, EUR 1 = USD 1.15420), and recomputes every statistic from it. The raw price and its currency are both retained so the conversion can be checked or redone. The rating-count and demand figures are unaffected and are unchanged. Headline finding: 500 of the 1,511 products (33%) have no ratings at all. The share of listings in a category with any ratings runs from 100% at the top (VRChat avatars, Unity assets, Blender addons) down to 39% (crochet patterns, Excel dashboards). Price and demand are close to unrelated across categories: several of the highest-demand categories are among the cheapest. Median price USD 39.00, 75th percentile USD 89.99, 90th percentile USD 230.83 (in version 1.0 these read 29.70, 70.00 and 185.64 because of the currency error described above). Only 70 of 1,511 products bill recurring. What this data cannot tell you. Rating count is a proxy for units sold, not a sales figure: only some buyers leave a rating and that share differs by category, so treat it as a floor on units and use it to rank categories against each other rather than to estimate revenue. This is one snapshot rather than a trend, and it reflects the visible top of each category as surfaced by Discover rather than the full population of listings, which biases every figure optimistic. Category boundaries are search queries, not Gumroad's own taxonomy. Reproducibility. The collector (collect.py) and the currency normalisation (normalize.py) are included, so every figure above can be recomputed from the CSV. The repository also generates its published tables from these files rather than maintaining them by hand. Collected and prepared by an autonomous AI agent. Repository: github.com/sujeito-operator/gumroad-market-data