API Format
Analyzes survey responses to detect fraudulent activity with detailed request and response structure.
Headers
API key for authentication. Include this in the header of your request.
Body
A unique identifier for the participant.
Records the changes made by the participant while answering the survey. Each key is the id of the question, and the value is the array of changes generated by our Javascript tracker. See here for more instructions.
Contains the survey questions. Each key is the id of theu0009question, and the values are the text of the open-ended survey questions as shown to the participant.
Records the participant's final responses to each question.
A unique identifier for the survey.
Threshold for flagging a response as low effort. Must be between 0 and 10.
Response
Indicates if there was an error in processing the request.
Indicates the type of fraudulent activity detected. This example shows an off-topic response detected.
Number of checks that failed.
Dictionary where key corresponds to question_id and value corresponds to the cluster of duplicated responses.
Dictionary where keys correspond to questions, and every entry is an array of strings, each string representing a failed check. Possible checks include:
- "Automated test: Gibberish" - The response is not coherent or grammatical.
- "Automated test: Off-topic" - The response does not address or relate to the question.
- "Automated test: GPT" - The response is flagged as likely generated by a GPT model.
- "Low-effort" - The response relates to the question but provides minimal information. This is flagged if the effort score is less than or equal to the
low_effort_threshold
. - "GPT paste artifacts" - The text is formatted similarly to ChatGPT outputs.
- "Programmatic entry" - The response was entered programmatically.
- "Response pasted" - The response was pasted or entered all at once.
- "Self-duplicate response" - A similar or identical answer was given to more than one question.
- "Cross-duplicate response" - A similar or identical answer as another participant was provided.
- "Unnatural typing" - Typing behavior flagged as likely non-human.
Dictionary where key corresponds to question_id and value is an effort score from 1-10, with 1 being minimal effort and 10 being high effort.
String specifying the model type.