nudgecv
Sign inGet started

AI Disclosure

Last updated: 16 June 2026

NudgeCV uses generative AI to draft and improve resume bullets, cover letters, LinkedIn copy, keyword lists, and interview-prep questions. This page explains how it works and what that means for you.

Which model?

We use Claude Haiku 4.5 from Anthropic, accessed through Amazon Bedrock in the us-east-1 AWS region. The same model handles every AI feature in the app — there is no separate "upgrade" tier with a different model.

What we send to the model

  • The resume or cover letter content you ask us to optimize.
  • The job description you paste in (for tailored optimizations).
  • A system rubric we wrote that defines what counts as a good output (ATS-friendly, honest, specific, no fabrication).
  • A short instruction telling the model what task to run (rewrite a bullet, score an ATS match, draft a cover letter, etc.).

We do not send your email, password, payment info, or any data from other users.

What we get back

Suggestions — rewritten bullets, scores, lists of likely interview questions, etc. — that we display in the app for you to review, edit, or discard. Nothing is sent to a recruiter or posted on your behalf. You decide what to use.

What it's not

  • Not legal, medical, financial, or career advice. The model is good at rewording text and pattern-matching to job descriptions. It is not a substitute for a career coach, recruiter, or counselor.
  • Not guaranteed to be accurate. The model can invent details, get dates wrong, or confidently produce incorrect information ("hallucinate"). Always check its output against the truth before using it.
  • Not a way to fake credentials. Our system prompt explicitly tells the model not to invent jobs, schools, or skills you didn't enter. If you do see a fabrication, please report it — that's a bug, not a feature.

Your responsibility

You are responsible for reviewing every AI-generated line before using it in an application, on your LinkedIn profile, in an interview, or anywhere else. Submitting AI-generated content that contains false information is your responsibility, not the model's and not ours.

We design every AI feature to surface the suggestion alongside the source text so you can compare line by line. We also flag exaggeration ourselves with an "honesty score" on each suggested bullet.

Training

We do not train any model on your data. AWS Bedrock's contract with Anthropic states that prompts and completions sent through Bedrock are not used to train Anthropic's public models. We do not save prompts or completions for any training purpose, public or internal.

Opt-out

AI is the core of NudgeCV — there is no way to use the optimizer features without sending content to the model. If you don't want AI involvement, you can still use NudgeCV as a manual resume editor + tracker without ever clicking "Optimize".

Questions or concerns?

Email hello@nudgecv.com. If you believe the model produced harmful output, please include enough detail for us to reproduce it; we treat AI output bugs as priority issues.