AI transparency
Every AI feature in the product: what it reads, what it produces, and whether it can decide anything about you. Written because running models over a CV is profiling, and profiling deserves a straight answer.
On this page
This is the complete list. If a feature is not here, it does not use AI.
| Feature | What it reads | What it produces | Can it decide anything? |
|---|---|---|---|
| CV generation and improvement | Your CV content, target role | Rewritten bullet points, a formatted CV, an ATS score | You accept, edit or discard every suggestion. Nothing is applied automatically. |
| Role match and gap analysis | Your CV, a job description you supply | A fit score across seven dimensions, missing skills, an evidence map | Advisory only. It does not decide whether you may apply, and no employer sees it unless you apply. |
| Cover letters | Your CV, the job description, a tone you choose | Draft letter text | A draft. You edit and send it yourself. |
| Interview prep and coaching | Your CV, the job description, your recorded answers | Practice questions, focus areas, scoring on clarity and delivery | Practice feedback. It is never shared with an employer. |
| Career health, diagnostics and Pathfinder | Your applications, outcomes, skills and profile | An explanation of where your search is failing, with citations; ranked next roles | Advisory. Declines to answer when your history is too thin to support a conclusion. |
| Skill path, study guides and salary signal | Your CV, target role, real job postings | A phased plan, study material, a market range | Advisory. Salary figures come from real postings, never invented. |
| Autopilot auto-apply | Your CV, your standing preferences, open postings | An application submitted on your behalf | Off unless you switch it on. It applies only above a qualifying score, and every application appears in your pipeline. |
| Applicant ranking and screening (employer side) | Applications received, screening answers, the job description | A rank order and a pass / review / fail suggestion with reasons | A suggestion to a human recruiter. No candidate is rejected by the system alone. |
| Job description generation (employer side) | The employer's own company profile, indexed role knowledge | Draft job description text | A draft the employer reviews and edits before publishing. |
| Recruiting agents (employer side) | Pipeline data, candidate profiles, the employer's templates | Draft outreach, prep packs, offers, client summaries | Every agent drafts for approval. None contacts a candidate on its own. |
GDPR Art. 22 gives you the right not to be subject to a decision based solely on automated processing that produces legal effects or similarly significantly affects you. A hiring rejection is exactly that kind of decision, so we designed around it rather than arguing about it.
If you believe an AI-derived score has affected you unfairly, you can ask for human review by emailing privacy@neuralcareernetwork.com. We will look at the underlying data, explain the score, and correct it if it is wrong.
The diagnostic features are built to be grounded in your real activity, and to say so when they cannot be. Career Health cites the applications and outcomes behind each finding. Salary figures come from real job postings — if there is no signal for a role and location, the product says there is no signal rather than inventing a range. Job descriptions are grounded in an indexed role-knowledge base, and the generator will not invent a salary the employer did not enter.
This is a deliberate product decision. A confident invented number is worse than an honest gap, because you might act on it.
AI output can be wrong. ATS scores are an estimate of how a parser is likely to read your CV, not a guarantee of how any particular employer's system will. Fit scores are advisory. Generated text should be read before it is sent.
You are always the last step. Nothing generated is sent to an employer without you choosing to send it, except an Autopilot application you configured.