You open an AI-generated cover letter and it looks impressive. The grammar is clean, the sentences flow—and absolutely anyone could have written it.

Used badly, AI can polish an application while removing the person an employer needs to meet. Used well, it can organise experience, broaden a search and support practice. The difference is whether it supports your judgement or replaces it.

Give AI a task, not your identity

Treat AI as a research assistant, editor and rehearsal partner—not a career oracle.

It can compare an advert with your experience, group skills and improve clarity. It cannot know which workplace will suit you or which trade-offs your family can manage.

Before prompting, decide what job you want the tool to perform. “Fix my career” invites generic advice. A focused request is more useful:

“Compare this public job advert with my factual career summary. Identify strong matches, missing evidence and questions I should answer. Do not invent experience, qualifications or numbers.”

Clear boundaries leave less space for confident guesses.

Build a truth pack before you prompt

Create a redacted career brief with confirmed titles and dates, achievements you can explain, tools you have used, target roles, practical constraints and any claims still needing verification.

Remove ID, passport and bank numbers, medical or family information and confidential employer data. A public vacancy plus a sanitised career summary is usually enough.

Tell AI to use only supplied facts, mark missing information and separate inference from evidence. An unmeasured “30% efficiency gain” does not become true because it sounds plausible.

Search by meaning, not only by job title

Job titles are inconsistent. Similar work may appear as customer success, client services or account support.

Where an AI-powered search feature is available, describe the work you want in normal language and add context. For example:

“Entry-level customer support roles in Cape Town that use retail service experience, offer hybrid work and do not require a university degree.”

LinkedIn says its AI-powered search can interpret intent beyond exact keywords, although access is gradual and some searches remain unsupported. Treat results as leads: open the original advert and confirm the requirements and application channel.

Translate experience without inventing it

AI can translate workplace language when it remains anchored to evidence.

Suppose you managed queues, handled returns and calmed dissatisfied shoppers in a busy Gauteng store. A customer-support advert may use terms such as case management, de-escalation and service resolution. Ask AI to map the real activities to the advert’s requirements, then choose wording you can defend in an interview.

A good process is to extract the five most important requirements, match each to a real example, mark missing evidence, draft only from those matches and check every verb, tool, number and outcome.

Ask the model to explain a gap accurately, not conceal it, and to clarify your real responsibility rather than “make me sound senior.”

Edit until the application sounds like you

AI often defaults to phrases such as “results-driven professional.” They are too general to carry your story.

Give the tool a short sample you wrote naturally and request plain South African English, short sentences and specific evidence. Then read the result aloud, replace phrases you would never say and remove claims that are only technically true after a long explanation.

Your final CV and message do not need to sound casual, but the same person should appear on paper and in the interview.

Use AI as a demanding practice partner

Ask the model to run a mock interview from a public advert, one question at a time, and challenge vague answers for relevance, evidence and clarity.

Practise aloud before reading the suggested answer. Otherwise, you may train yourself to recognise good answers rather than produce them.

AI can turn a genuine experience into a STAR outline, but you must supply the situation and result. It can help rehearse practical questions about notice, location or salary expectations; it cannot know an employer’s scoring or secret “correct” response.

Protect your data and verify every opportunity

POPIA sets conditions for lawful processing, including purpose, quality, openness and security. Consent is not the only possible lawful basis. Practically, share only what is needed, know who receives it and keep sensitive material out of experimental prompts.

Never paste another employer’s confidential reports, customer lists, passwords or internal assessments into an AI tool. Review the provider’s privacy and retention settings before using personal documents.

AI also cannot certify that a vacancy is real. The South African Department of Employment and Labour warns work seekers not to pay recruitment fees or surrender personal information to suspicious recruiters. Verify the company, email domain and application route independently.

Build a repeatable weekly cycle: discover broadly, select carefully, map evidence, draft from your truth pack, edit in your voice, verify and practise. The goal is not maximum applications; it is less avoidable work without surrendering judgement or personality.

Sources and limits

Product-provider guidance explains intended use but does not independently prove that AI improves hiring outcomes. Features, availability and privacy settings can change. AI may produce errors or fabricated details and cannot guarantee an interview or offer. This article is general career and privacy guidance, not personal legal advice.