AI Disclosure
ResumePay uses AI models to read documents, compare your experience with jobs, and draft wording. This page explains where AI is involved, what it sees, the safeguards around it, and its limits.
Where AI is used
- Reading your resume: pulling employment, projects, skills, achievements, education, and certifications out of an uploaded file, each tied to the words it came from.
- Strengthening an achievement: asking follow-up questions, and turning your answers into a result with the evidence behind it.
- Job analysis: reading a posting into requirements, finding evidence in your profile for each, and judging how strong it is. Matching may also turn text into embeddings (number lists that capture meaning) to find related evidence.
- Resume tailoring and rewrites: suggesting changes for a target job, from your evidence only.
- Localization: rewriting a resume the way resumes are written in another country and language.
- Interview preparation: likely questions, and STAR stories built from your verified achievements.
Some parts don’t use AI at all: the internship listings and how they match your skills, the ATS parse preview, exports, and the application tracker.
What the AI sees
Each task sends the AI provider only what it needs, such as the text of your resume, the profile items relevant to a job, or the job posting. Which providers this deployment uses, and where they are, is on the Subprocessors page. We don’t give any provider permission to use your information to train its models; Anthropic and OpenAI, under the terms of their business APIs, don’t use data sent through them for training. The “AI model improvement” switch in Settings is off by default, and today ResumePay doesn’t share anything for model training whatever it says.
Safeguards
- Your evidence comes first. Everything in your profile shows where it came from. What the AI proposes but your document doesn’t say is marked as needing your review, and isn’t used until you verify it.
- Writing from verified facts. Generated text may draw only on facts you verified or that your own documents state.
- Fact Lock. Each line the AI writes is checked against your evidence. A claim, number, or name that can’t be matched is flagged for you to confirm or remove before it can be used.
- Numbers come from you. When strengthening an achievement, a figure is used only if it came from your answers.
- Documents are content, not commands. The AI is told to treat the text of an uploaded resume or a job posting as material to read, and to ignore instructions hidden in it; Fact Lock catches claims that get through anyway.
- You decide. You accept, edit, regenerate, or reject every suggestion. Nothing generated is exported, saved to your profile, or sent anywhere without your action, and ResumePay never submits anything to an employer for you.
Limits
AI makes mistakes. It can misread a document, miss a relevant fact, judge a match too strongly or too weakly, or phrase something in a way that overstates it, and the checks above reduce these errors without ruling them out. Match strengths and other analyses are estimates, not predictions of how an employer will see you. Read everything before you use it: you’re responsible for what you submit, as the Terms of Service say.
Questions or problems
If the AI gets something wrong in a way that worries you, or you have questions about how it works, see the Contact page. The Privacy Policy covers how your information is handled.