Detect AI-generated text in an uploaded document
Upload a document (.pdf, .docx, .pptx, .odt, .txt, .md, .html or .htm, chosen by file extension, up to 30MB); its extracted text (first 200,000 characters) is scored like /api/v1/aidetect with sentence scores. PPTX text is read in slide order without speaker notes; ODT drops footnotes, comments and tracked deletions; HTML is flattened to its visible text. Billed like /api/v1/aidetect.
Request
- Base URL: https://api.sapling.ai
- URL: https://api.sapling.ai/api/v1/file-aidetect
- Auth: one of:
- HTTP bearer
- none
Query parameters
When true, a successful JSON-object response additionally carries "usage": {"characters": <n>} — the billable character count this request submitted (0 for unbilled requests; the 2.5x adjustment for logographic languages is included, so CJK requests report more than their raw length). Repeat texts inside the billing deduplication window are still counted here even though they are not re-charged, so the number is an upper bound on the final charge. Responses whose body is not a JSON object (e.g. bare-array response formats) cannot carry the field. An existing usage field (such as account/status or completion token usage) is preserved unchanged.
Sapling API key, as an alternative to the "Authorization: Bearer <key>" header.
Response
Same shape as /api/v1/aidetect with sent_scores and score_string enabled.
Changes
No recorded changes to this operation across all 1 revision of this API.