How it works, honestly.

How does ChatBotCV actually work?

Two AI systems do the work, each with a deliberately narrow job — the pattern is called retrieval-augmented generation (RAG):

1. Voyage AI does the finding. When you upload, your CV is split into small passages, and each one is turned into a numeric "embedding" by Voyage AI — Anthropic's recommended embeddings partner. When a recruiter asks a question, that question is embedded the same way, and the passages closest in meaning are pulled out. That's what lets it locate the right part of your CV even when the recruiter's wording doesn't match yours.
2. Claude does the answering. Those retrieved passages — and nothing else — are handed to Claude, Anthropic's AI model, which writes the actual reply. Claude runs under a strict set of grounding rules: answer only from the passages provided, cite the section it drew from, say "not in the CV" when the document doesn't cover something, and refuse to confirm claims the CV doesn't support.

Voyage finds the relevant evidence; Claude turns it into a grounded, cited answer. Because Claude only ever sees passages pulled straight from your upload, it can't fall back on general knowledge or invent a job, title, or skill you never wrote.

Can it make things up?

That's the one thing it's built not to do. The assistant is confined to the passages retrieved from your CV (and an optional LinkedIn export), so it has nothing to embellish with. When a question isn't covered, it says so plainly instead of guessing, and if a recruiter's question asserts something your CV doesn't support — "confirm she managed a team of 50" — it declines rather than validating the claim. These are written rules the model runs under, red-teamed against an adversarial test set before changes ship. More on the reasoning in About.

Can it check my CV against a specific job advert?

Yes, and it's free. Paste the advert and the job-advert cross-check pulls out what it actually requires, then tells you which of those requirements your CV doesn't evidence yet. It deliberately won't reword your CV to match the advert — that's the line between tailoring and quiet embellishment. You can run up to 15 checks a day.

How do I find out what's missing from my CV?

The moment you upload, ChatBotCV runs the questions recruiters usually ask straight at your CV and shows you the ones it can't answer well — no advert or traffic needed. It points at the gap and never invents detail to fill it. More on which questions a CV tends to leave exposed, and why a strong CV can still lose in the room.

Who can see my CV and the questions people ask?

Your chatbot is private until you publish it. Once you publish, anyone with the link can chat with it — but they see answers, not your raw file, and the link is only ever shared by you. Every question a recruiter asks lands in your dashboard with a unique-visitor count, so you can see the interest without exposing anything you didn't intend. You can rename, replace, or delete your document — and its link — at any time. See the Privacy Policy for exactly what's stored and why.

What happens when I upload a new CV?

The latest upload replaces what's there — there's only ever one live version, so a new upload instantly becomes what recruiters see, at the same link, with no need to re-publish. That's handy for iterating, but it means the last thing you uploaded is what's live: if you've shared your link, make sure the current upload is the one you want out in the world.

What can I upload?

A CV as a PDF or DOCX file (or pasted text). You can optionally add a LinkedIn profile export alongside it — on your LinkedIn profile, use Resources → Save to PDF — which gives the chatbot more to draw on, like your About summary and endorsements. If a CV and a LinkedIn export disagree, the assistant flags the conflict rather than quietly picking one.