AI that works for you, under constraints you can verify.
HireRoom uses language models for drafts, scores, and a copilot. Each use sits under a hard rule enforced in the product. These are the rules.
The AI cannot invent your history
Resume and cover-letter generation is bound to your verified profile: work history, skills, dates, and metrics you entered. If a job asks for something you cannot honestly claim, HireRoom writes a weaker line and flags the gap instead of filling it with fiction. This is a design constraint in the generation pipeline, not a tone-of-voice guideline.
You approve everything before it moves
No application is submitted under your name without your review and your click. The Chrome extension fills supported forms and stops; the company's own submit button remains yours. There is no auto-submit and no mass-apply feature, and there will not be one.
Scores show their evidence
Fit scores list the signals they used instead of presenting a bare number. Ghost-likelihood is labeled as a heuristic, never as fact. When a score is still computing, the product says so rather than inventing one.
Your data is not training data
Generation runs through Anthropic (default) and OpenAI (fallback, and voice mocks) under commercial API agreements that do not use your content to train their models. The full sub-processor list is on the security page.
The AI is fallible, and the product says so
Models misread postings, misjudge seniority, and miss context a human catches in seconds. Treat every generated document as a draft: read it, fix it, then send it. If HireRoom ever fabricates a line, that is a bug; report it at /support and it will be treated as one.
Where to check these claims
- · Security: data handling and the sub-processor list.
- · Privacy policy: retention, deletion, and your rights.
- · Data sources: where job and sponsorship data comes from.
- · Product: the evidence-bound workflow itself.