BotOrNot is a publicly available service that evaluates the likelihood that a Twitter account is controlled by a social bot. It uses over 1,000 features to assess how similar a Twitter account is to known characteristics of social bots. Since its release in May 2014, BotOrNot has served over one million requests via its website and APIs. The service allows users to input a Twitter screen name, retrieve the account's recent activity, and compute a bot-likelihood score. For website users, this score is accompanied by plots of the various features used for prediction. API users receive the classification results in JSON format. The service was initially only available via the website, but later a public API was introduced to allow programmatic access. The API has a rate limit of 180 requests per 15 minutes. The service has been used over 540,000 times since December 2015, bringing the total to over a million queries. BotOrNot's classification system generates over 1,000 features using available metadata and information extracted from interaction patterns and content. These features are grouped into six main classes: network features, user features, friends features, temporal features, content features, and sentiment features. The system uses a Random Forest classifier to train seven different classifiers, one for each subclass of features and one for the overall score. The classifier has a performance of 0.95 AUC (Area Under ROC Curve) using ten-fold cross-validation. BotOrNot aims to lower the entry barrier for social media researchers, reporters, and enthusiasts by providing a free social bot evaluation service. Ready-made reports on individual users are available via the website, or one can use the API to easily check multiple accounts up to the rate limit. The service is designed to be accessible to users without the need to set up their own classifiers. The system is supported by NSF, DARPA, and the J.S. McDonnell Foundation.BotOrNot is a publicly available service that evaluates the likelihood that a Twitter account is controlled by a social bot. It uses over 1,000 features to assess how similar a Twitter account is to known characteristics of social bots. Since its release in May 2014, BotOrNot has served over one million requests via its website and APIs. The service allows users to input a Twitter screen name, retrieve the account's recent activity, and compute a bot-likelihood score. For website users, this score is accompanied by plots of the various features used for prediction. API users receive the classification results in JSON format. The service was initially only available via the website, but later a public API was introduced to allow programmatic access. The API has a rate limit of 180 requests per 15 minutes. The service has been used over 540,000 times since December 2015, bringing the total to over a million queries. BotOrNot's classification system generates over 1,000 features using available metadata and information extracted from interaction patterns and content. These features are grouped into six main classes: network features, user features, friends features, temporal features, content features, and sentiment features. The system uses a Random Forest classifier to train seven different classifiers, one for each subclass of features and one for the overall score. The classifier has a performance of 0.95 AUC (Area Under ROC Curve) using ten-fold cross-validation. BotOrNot aims to lower the entry barrier for social media researchers, reporters, and enthusiasts by providing a free social bot evaluation service. Ready-made reports on individual users are available via the website, or one can use the API to easily check multiple accounts up to the rate limit. The service is designed to be accessible to users without the need to set up their own classifiers. The system is supported by NSF, DARPA, and the J.S. McDonnell Foundation.