Generative AI tools used in recruitment do more than rephrase bullet points on a CV. For candidates, they reshape the job search, refine the semantic targeting of applications and speed up interview preparation. For employers, automated screening systems analyze far more than keywords: skills parsing, behavioral scoring and chronological consistency checks. Understanding both sides of this process changes how you prepare and submit an application.
What AI-powered ATS actually analyze in a CV
Most articles advise tailoring your CV to match the keywords in a job posting. In reality, the story begins after you submit it. The latest Applicant Tracking Systems (ATS) no longer rely on simple lexical matching between a CV and a job description.
These systems extract structured entities: standardized job titles, technical skills mapped to industry frameworks, experience calculated from dates, and even consistency between the level of responsibility claimed and the action verbs used. A CV that lists “project management” without indicating scope (budget, team size, deliverables) will score lower than one that quantifies these details.
AI screening is classified as “high risk” under the European AI Act, with specific obligations due to take effect on December 2, 2027. However, GDPR and French employment law requirements already apply. Article L. 1221-8 of the French Labor Code requires candidates to be informed about the methods and techniques used to assist recruitment in their case, and those methods must be relevant to the purpose pursued.
In practical terms, a candidate can ask a company which automated tools processed their application. AI cannot make the decision to reject an application on its own when that decision has significant effects: the GDPR requires meaningful human involvement, not merely a formal sign-off on the algorithmic result. Candidates remain largely unaware of this right to human review.

AI-assisted CV writing: fine-tune your prompt to avoid generic results
Platforms specializing in AI-powered CV writing are multiplying. Their value depends directly on the quality of the prompt and the proofreading that follows. A tool such as those listed on https://www.jobassistant.fr structures the process by guiding candidates step by step.
The main pitfall is accepting the model’s first output. LLMs produce statistically likely wording, which means the same phrases thousands of other candidates are getting too. The result is a technically correct but interchangeable CV.
To create a document that stands out, your prompt must include three things the model cannot guess:
- The specific industry context, including the target company’s professional vocabulary (taken directly from the job posting or corporate website)
- Verifiable, quantified results for each role, provided by the candidate: growth percentage, volume handled, time saved
- Any format requirements imposed by the role or industry (one-page CV for finance, project-based CV for consulting, academic CV with publications)
An AI-generated CV without precise personal details looks like every other AI-generated CV. Experienced recruiters can now spot stock phrases such as “passionate about innovation,” “strong team player” and “results-oriented.” These signals make them read the rest of the document more skeptically.
Job searching and algorithmic matching: make the most of smart filters
Job boards are gradually adding AI layers that go beyond simple keyword searches. LinkedIn, Indeed and industry-specific platforms use browsing history, previous applications and the information in your profile to suggest jobs.
We find that the quality of these recommendations depends directly on how complete a candidate’s profile is. A profile with fewer than half of its available fields filled in receives poorly targeted suggestions because the algorithm lacks the signals it needs to refine its matches.
Three technical steps can improve the relevance of your results:
- Enter skills using the exact terms found in job postings in your target industry, rather than approximate synonyms
- Set up alerts using combinations of criteria (location, salary range, contract type) instead of a single generic keyword
- Update your profile regularly: algorithms favor active profiles when making suggestions to recruiters
Matching works both ways: the same algorithm that recommends jobs to candidates also recommends profiles to recruiters. Optimizing your profile for an active job search also improves your passive visibility.
Emotion analysis in video interviews: a practice now regulated
Some tools claimed to analyze candidates’ facial micro-expressions or vocal variations during video interviews to infer stress, motivation or sincerity. Emotion analysis has been prohibited in the workplace since February 2, 2025, except for certain cases provided for under European regulations. Candidates faced with this type of tool have clear legal grounds to challenge the practice.
Personal data protection and generative AI in recruitment
Sharing a CV with a general-purpose AI assistant (ChatGPT, Gemini, Claude) to rephrase it raises an issue rarely covered in practical guides. A CV contains personal data under the GDPR: name, contact details, work history, and sometimes a photo and date of birth.
For candidates, the risk is limited as long as the use remains personal. For employers, using an LLM to screen or rank applications means establishing a legal basis for processing, informing candidates and, according to the CNIL, carrying out a data protection impact assessment if the tool is used to make decisions affecting individuals.
This imbalance in obligations puts candidates in a position to benefit from understanding the legal framework. Asking a company whether it uses automated tools in its recruitment process is not an expression of distrust; it is the exercise of a right provided for under the French Labor Code. The information obtained can help candidates tailor their applications accordingly, ensuring their CV passes the technical filters before reaching a human reviewer.



