Last updated: August 10, 2026
Undertext scans a file or pasted text you upload (a take-home assignment brief) for hidden or invisible content — things like tiny/white text, hidden spreadsheet rows, or suspicious metadata. Most of this scanning happens directly in your browser.
To check for vague or ambiguous instructions, the text of your document is sent to Anthropic's Claude API for analysis. We don't send images or spreadsheet data for this step — only extracted text from documents you scan.
We log basic scan metadata — the type of file scanned, whether the scan succeeded, how many issues were found, and how long it took. We do not store the content of your file, the text of your assignment, or any personal information about you.
To understand how the tool is being used and where it breaks, so we can improve it. This is not used to train any AI model.
Since we don't collect identifying information, there's no account or personal profile to request access to or delete. If you have questions, reach out at swabhig@gmail.com.