Job seekers increasingly turn to AI chatbots to polish their resumes and cover letters, but the evidence paints a nuanced picture: AI can improve writing and even hiring outcomes, yet widespread adoption may erode the very signals employers rely on. Meanwhile, pasting a full CV into a chatbot raises privacy concerns that demand careful handling.
- Hiring boost: A 2023 field experiment found that algorithmic writing assistance raised the probability of being hired by 8%.
- Signal erosion: Later studies show that when AI-generated cover letters become ubiquitous, employers discount personalization and focus on verifiable experience.
- Privacy imperative: Chatbots may retain conversations for up to three years, so removing contact details before pasting is essential.
What the Research Says
The 2023 study by Emma van Inwegen, Zanele Munyikwa, and John J. Horton, published on arXiv on January 19, 2023, ran a field experiment on an online labor market. Job seekers who received algorithmic writing assistance on their resumes saw an 8% increase in hiring probability, and the authors found no evidence that employers were less satisfied. Their explanation: better writing does not prove greater ability, but it helps employers see that ability.
Subsequent work focused on cover letters. Jingyi Cui, Gabriel Dias, and Justin Ye analyzed an AI cover-letter tool on a large online work platform (arXiv, September 2025, revised November). Access to the tool made letters more similar to job postings and raised response rates. But after its introduction, the relationship between that adaptation and employer responses fell by 51%, and employers began paying more attention to other signals, such as prior work history. A practical detail: time spent editing the AI draft correlated positively with landing the job.
Anais Galdin and Jesse Silbert (arXiv, November 11, 2025) used data from a large digital work platform to observe that before large language models, employers valued customized applications highly—and afterward, they did not. In an economic simulation of a scenario where cover letters no longer serve as a signal, individuals in the highest ability quintile would be hired 19% less and those in the lowest quintile 14% more. It is a simulation, not a direct measurement, but it points in the same direction.
The takeaway from these three studies: AI helps you avoid losing points for poor writing, but it no longer earns points for good writing. What counts again is the verifiable: experience, results, references.
How to Use AI—and What to Keep Away From It
Use it to edit, not to invent. The use case best supported by the research is starting from your own text and asking the AI to organize it, shorten it, or adapt it to the vocabulary of the job posting. That human editing correlates with more hiring in the Cui, Dias, and Ye study is no coincidence: it is what separates your letter from everyone else’s.
Prompts that tend to work:
- “Rewrite this paragraph in fewer words without changing the facts.”
- “Compare my resume with this job posting and tell me which requirements are missing.”
- “Flag generic phrases that anyone could have written.”
What you should not delegate:
- Achievements and figures. Language models tend to fill gaps with plausible data—a phenomenon known as hallucination. A percentage improvement that never existed, a language you do not speak, or a tool you have not used will surface in the first interview.
- Tone. If the letter sounds like a brochure (“passionate about excellence,” “dynamic environment”), it resembles all the others.
- Dates and titles. Check every line against your actual documents.
As for whether employers “detect” AI use: no study with a named author and date measuring how many companies use AI-text detectors in hiring was found, so no figure is offered here. What is documented is that such detectors frequently fail. And if you are concerned about the other side—an algorithm filtering your application—candidates have specific rights under data protection rules.
Is It Safe to Upload Your Resume to a Chatbot?
It depends on what data it contains and how your account is configured, because a full CV is a summary of your life: name, phone, address, email, work history, and sometimes a photo. The prudent approach is to remove contact details before pasting and to review the service’s privacy settings.
Here is what two of the best-known services say, according to their documentation as of November 2026:
- Gemini. Its Privacy Hub warns that human reviewers examine some conversations and explicitly asks users not to include confidential information they do not want a reviewer to see. Reviewed conversations are retained for up to three years, even if you delete your activity. Auto-delete is set to 18 months by default. With “Save activity” turned off or in temporary chats, conversations are stored for 72 hours and are not used to train models, unless you submit feedback.
- ChatGPT. According to OpenAI’s data controls FAQ, on personal Free, Plus, and Pro accounts the “Improve the model for everyone” option is enabled by default; you can turn it off without losing your history. Temporary chats do not appear in history, do not create memories, and are not used to improve models.
A simple trick: replace your name, phone, and email with placeholders like [NAME] and [PHONE], work on the text, and reinsert them at the end on your own computer. AI does not need to know where you live to improve a sentence.
The most widespread mistake is treating AI as a shortcut to avoid writing. The data suggest the opposite: the shortcut is now available to everyone, so the advantage returns to what AI cannot fabricate for you. Use the chatbot as a demanding proofreader, not a ghostwriter, and spend the time saved on what actually moves the needle: a concrete achievement, told in your own words, that an interviewer can verify.