AI Disclosure

JobResumAI uses large language models to assist with writing and analysing resumes. This page explains exactly what the AI does, what it doesn't, and where your judgement matters.

Which AI models we use

• Anthropic Claude (Sonnet 4 for most tasks, Haiku for lightweight ones like proofreading) — primary model.
• OpenAI GPT-4o-mini — fallback when Claude is unavailable or for specific tasks.

Both are accessed through commercial API endpoints under contracts that prohibit using your inputs to train their models.

What the AI is instructed to do

Every prompt we send to these models is framed by a system instruction. The key rules, in plain English:

• Never invent metrics or experience.If a number isn't in your input, it won't be in the output. If an experience isn't described, the AI won't claim you have it.
• Describe scope or business impact when the user has no specific metric — not fake numbers.
• Avoid cliches("passionate", "team player", "results-driven", etc.).
• Preserve your facts exactly during bullet rewrites — we reshape sentences, not substance.

What the AI can't do well (yet)

Large language models make confident mistakes. Watch for:

• Subtle factual drift. A rewrite might shift "team of 4" to "team of 5" or swap verbs in ways that imply more ownership than you had. Read each rewrite before accepting it.
• Industry jargon. For very niche roles the AI may produce generic phrasing. Your domain knowledge is the final authority.
• ATS score precision. Our scoring is the best approximation we can build from published ATS vendor documentation — it is not a guarantee your resume will get a specific outcome at any given company.
• Compliance claims. We say our templates are designed to parse cleanly in major ATS systems. We don't promise they'll pass every obscure deployment of every vendor.

Your responsibility

You are the author of your resume. We provide tools. Before exporting, please:

1. Read every line the AI produced.
2. Verify that every number, title, date, and company name is accurate.
3. Correct anything that doesn't match reality. Submitting a misleading resume — whether written by you, by AI, or by anyone else — is your risk.

Prompt injection defense

We sanitize inputs before sending them to the model to defend against prompt-injection attempts smuggled in a job description or resume content. Still, no defense is perfect. If output looks suspicious — especially if it references unrelated topics or instructions — delete it and start over.

Data retention

Your resume content is stored in our database only as long as your account is active. You can delete your account and all associated data from account settings at any time. AI providers retain request data per their own policies (usually 30 days for abuse monitoring, then deleted).

Questions

Email support@jobresumai.com.