AI & ChatGPT · 2026-08-19 · 13 min read

Improve a CV with AI: what a tool does (and what a chat does not)

How to optimise a résumé with AI when you already have a PDF: analysis, ATS template, and limits. Chat versus a CV tool such as ClarCV.

Improving a CV with AI is not pasting the file into a chat and hoping for a miracle. A CV tool with artificial intelligence works on your document: it extracts text, flags clarity and structure issues, and can regenerate a PDF built for ATS filters. A chat (ChatGPT, Gemini, Claude) drafts paragraphs if you give context. It does not parse Canva well, does not guarantee selectable text, and is not your employment history.

If you want an AI CV tool or to optimise a CV with AI, intent is usually “I have a file and want the next step”, not “I want a prompt”. This guide maps that intent to the ChatGPT cluster on this blog, lists a safe flow (facts → AI → you review → ATS file), and places ClarCV: PDF/Word upload, analysis, templates, download. Start free, €7/month if Pro.

AI does not replace real job ads or metrics you can defend. Optimising means aligning what is true with the role’s language and with a format software can read. Inventing 40% churn is the most expensive failure, whether a model or a human does it.

Chat vs tool: two different jobs

• Chat: rewrites bullets, summarises, translates, extracts keywords if you paste the ad.
• CV tool: input and output file, template, A4 margins, document analysis.
• Chat: copy-paste into Canva or columns breaks ATS (see the Canva guide).
• Tool: the value is the parseable PDF, not a “professional” tone in a paragraph.

If you only need sentences, chat is enough and cheaper. If you need to apply to a portal tomorrow, you need the file. Mixing both is fine: chat for a draft of achievements, ClarCV for the ATS shell. There is a dedicated ChatGPT vs CV-tool guide on this blog; here the focus is “optimise the one I already have”.

Flow to optimise the PDF you already have

First, the text test (select and paste). Second, one ad beside it: 8–15 real keywords. Third, structure analysis (sections, dates, skills as words). Fourth, rewrite weak bullets with verb + scope + result. Fifth, a one-column PDF. AI speeds up the third and fourth; you sign off the fifth.

Before: a 2019 CV, two columns, “responsible for various functions”. After: the same job, “Coordinated 3 distribution routes (12 stores) and cut delivery incidents from 9 to 4 per week.” The tool did not “invent logistics”; you supplied the fact and AI helped order it.

Privacy: not the same product

Pasting a CV into a consumer chat is not the same as an enterprise API with a no-training policy and an account. Do not upload ID numbers or exact addresses where they are not needed. Detail in the privacy and AI guide. In ClarCV the file travels to analyse and generate; it is not sold to advertisers.

What no serious AI promises

• A job. Only a clearer document.
• An official ATS score from that company.
• A certified Europass or an “approved” Canva file.
• Better lying. Worse lying, actually: more fluent and more checkable.

When to use ClarCV in this flow

When the bottleneck is the file: upload, read the analysis, apply an ATS template, download. Three analyses and two downloads per 30 days on the free plan. If the bottleneck is “I do not know what I was in that job”, inventory on paper first; AI does not remember your life. UI in ES/EN; output CV can follow the ad language.