How to use AI in your job search without sounding like a robot

01

Where does AI actually help in a job search?

Career centers describe the same division of labor. Harvard’s Mignone Center for Career Success calls generative AI a useful tool in the editing process, one that can help you brainstorm revisions to bullet points and work in relevant keywords from a job description. Its starting point is your own draft. Ready-made versions of these tasks are in ChatGPT prompts for your job search, and the research itself is covered in how to research a company.

02

Where does AI hurt?

It invents. The U.S. National Institute of Standards and Technology lists confabulation among the risks of generative AI: confidently stated but erroneous or false content. In a job search that means a tool you never used, a percentage you never measured, or a fact about a company that is out of date. Yale’s Office of Career Strategy warns that AI does not always pull from authoritative sources and may provide inaccurate information.

It sounds like everyone else. Harvard’s guidance says generative AI should not be the primary author, not least because its output will likely be very generic. The University of Illinois engineering career office adds that employers can often recognize overly AI-generated wording. Whether employers can tell you used AI matters less than whether the text says anything specific about you.

It can break the rules of an application. Harvard advises reading all application instructions, because some organizations have rules about whether AI can be used in preparing materials. Yale notes that employers may scan materials to check whether an AI tool was used, which may result in the denial of an application.

It does not belong in the live interview. Carnegie Mellon’s career center tells students never to use live AI tools during an actual interview. An interviewer is assessing how you think and speak. Use AI beforehand, for mock interviews.

03

How do you keep AI-assisted writing sounding like you?

1. Write the facts first. Before you open a chatbot, put down in rough notes what you did, for whom, and what changed. The notes can be ugly. They are what makes the result yours.
2. Give it a sample of your writing. A few paragraphs from an email or a report you wrote show it your level of formality and your sentence length.
3. Ask for options, not a final version. Three ways to phrase a bullet give you something to choose from and adapt.
4. Remove the tells. See the table below.
5. Add one detail only you would know. A named project, a specific obstacle, the reason you chose the company. Each paragraph needs one.
6. Read it aloud. If you would not say a sentence to a person, change it.

You are editing, not writing. Below are (1) a sample of my own writing and (2) a draft.

Rewrite the draft so it sounds like the sample: the same level of formality, similar sentence length, plain words.

Rules:
- Do not add any fact, number, skill, tool, employer or result that is not already in the draft.
- Cut filler and any phrase that could appear in anyone's application.
- Keep it under [number] words.
- After the rewrite, list every change that altered the meaning, so I can check it.

My writing sample:
[paste 150 to 300 words you wrote yourself, with names removed]

Draft:
[paste the draft, with personal details removed]

04

How do you keep the facts true?

Treat the chatbot as a fast writer who has never met you. It tends to fill a gap with something plausible, so close the gaps before it can.

1. Keep one source of truth. A master resume or notes file with every job, date, title, tool and result you can stand behind.
2. State the rule in the prompt. “Use only the facts below. If something is missing, write [NEEDS DETAIL] instead of guessing.”
3. Check the output against the source. Numbers, dates, job titles, tools and employer names first. The prompt below makes the chatbot do a first pass, but the final check is yours.
4. Confirm outside facts elsewhere. Carnegie Mellon’s advice is to ask chatbots for links and citations and cross-reference what they say with employer sites and people in the field.
5. Apply the interview test. Harvard’s version: be able to speak to every line of your resume if asked. Delete what you could not explain.

Compare the draft below with my source notes. The source notes are the only facts that are true.

List, as a table:
1. Every claim in the draft that does not appear in the source notes.
2. Every number, date, job title, tool or employer name in the draft, and whether the source notes support it exactly.
3. Any phrase that exaggerates what the source notes say.

Do not rewrite anything. Do not treat a claim as true because it is plausible.

Source notes:
[paste your master resume or notes, with personal details removed]

Draft:
[paste the draft]

The same routine applies to the resume itself, which is where invention does the most damage. See should you use AI to write your resume?.

05

What should you keep out of a chatbot?

Career centers are consistent on this. Harvard’s guidance is not to share personal or proprietary data when using generative AI. Yale’s is to leave out your name and other identifying information when you submit documents for review.

• Your identifiers. Full name, home address, phone number, date of birth, and any government ID number. Replace them with placeholders such as [NAME] and put them back afterwards.
• Other people’s details. References’ names and contact information, and colleagues or clients named in your notes.
• Your employer’s confidential information. Client names, financial figures, source code, unreleased plans, and anything covered by a confidentiality agreement. Describe the work in general terms instead.
• Sensitive personal facts. Health, immigration or legal matters, unless you have decided the tool’s privacy terms are acceptable for them.

Before you paste a resume into any tool, look at its privacy settings and find out whether your conversations are kept or used to train its models.

FAQ

Frequently asked questions