Ethereum transaction is the primary way for users to engage in on-chain activities, involving transferring funds and invoking smart contracts. The input data is an optional field in transactions that can store arbitrary data, and therefore, it can be used to store toxic content. Due to Ethereum’s immutability and transparency, toxic content in transactions is visible to everyone and cannot be removed, causing lasting negative impacts in the real world. Currently, Ethereum transaction analysis mainly focuses on the behavior of on-chain transactions to detect security issues such as phishing scams, lacking attention to the content safety of Ethereum transactions. This paper provides a comprehensive examination of toxic content in Ethereum transactions. We developed a two-phase content extraction method to retrieve arbitrary content embedded in Ethereum on-chain transactions, resulting in the extraction of 96,687 text segments, 9,931 files, and 1,198,889 URLs from 2,224,215,240 Ethereum transactions. We examined the extracted content, and the results indicate that approximately 3.2% of the data is toxic. Our work provides new insights into the types and distribution of toxicity in Ethereum transactions, contributing to further efforts to detect and mitigate this threat.