It was an unexpected problem. Earlier this year, Christina Qi, the chief executive of market data company Databento, noticed almost every job application included exactly what she was looking for.
The company prompts candidates to write a tweet and a press release about microwave towers, a niche topic that requires research, Ms. Qi said. Normally, most candidates fail the test. This time all five passed.
The tests—four from internship applicants and one from someone seeking a full-time content strategist role—were all so similar, “as if it was written by one person,” she said. Suspicious, Ms. Qi put the prompt into ChatGPT, the artificial-intelligence chatbot from OpenAI, to see what it could produce.
“Lo and behold, I got pretty much the same answer that all five candidates had submitted to me,” she said.
Since its launch in November, ChatGPT has been a hot topic at dinner tables and water coolers. Microsoft, Google, Snap and other companies have incorporated artificial intelligence into their products. People have experimented with using ChatGPT at work. Some have even started using it when looking for new roles, tapping the chat assistant to help write cover letters, tweak résumés and formulate responses to anticipated interview questions—without necessarily telling the people doing the hiring.
Employers, who have long used AI to screen potential employees, aren’t always disqualifying applicants who use ChatGPT, but they are scrambling to figure out how to assess candidates who may be using the tool to give them an edge.
Attention-grabbing applications
After being let go by his previous employer in January, Kyle Mickey started job hunting for roles in software engineering—alongside thousands of laid-off tech workers.
The 38-year-old from Colorado said he turned to ChatGPT for help, first sharing job descriptions and his résumé with the chatbot to see what it would tweak. Then he asked ChatGPT to write a recommendation letter for a role he coveted. The chatbot deemed him perfect for the job, as his technical skill set “aligns well with the requirements.”
Mr. Mickey sent the remarks to a recruiter, saying ChatGPT endorsed his skills.
“The recruiter was immediately like, ‘Let’s chat, I like the creativity,’ ” he said. Mr. Mickey didn’t get that job, but was hired at another company without ChatGPT’s help.
SHARE YOUR THOUGHTS
What part of the job-application process would you use AI for? Join the conversation below.
Ryan Stringham, 31, who lives in Utah and works in product management, used it to help write cover letters, including one that got him a foot in the door, and later hired, at a smart-tech company.
“You’re always looking, you’re always applying and you’re getting drained,” Mr. Stringham said of job hunting.
He said the bot broke his writer’s block, distilling his long-winded cover letter into four tight paragraphs. He said it also helped him prepare for job interviews by suggesting new ways for him to ask about company culture and expectations for the role. Instead of asking a vague question about what he should do to excel at the prospective job, ChatGPT suggested Mr. Stringham be more specific about the time frame and metrics for determining success.
Mr. Stringham has encouraged others to use chatbots in the job-hunt process, posting about them on LinkedIn and giving advice to other job seekers.
The only place he hasn’t disclosed his ChatGPT use: at work.
“It helped me get past the application process, and the recruiter never asked about it,” Mr. Stringham said, adding that he edited the cover letter himself and aced the interviews on his own.
‘How we present ourselves’
Programs, websites and other tools to help people fix their résumés and cover letters aren’t new. Microsoft Office and Google Docs offer résumé and letter templates, while companies such as Jobscan promise to optimize résumés to grab the attention of hiring managers, recruiters and hiring-system algorithms.
Candidates need to combine ChatGPT’s edits with their own editing and voice, said Sarah Baker Andrus, chief executive of Avarah Careers, a career coaching firm in Delaware. Whatever a candidate submits for a job should accurately reflect their skills, she said.
“We’re responsible for how we present ourselves,” Ms. Andrus said. “If you decide to use ChatGPT, it’s worthwhile to ask, ‘Is that representing the me that I want to present?’ ”
Employers are already finding ways to catch applicants who cheat with AI.
Engineers applying to San Francisco-based Cobalt Robotics take part in a remote one-hour coding interview where they are paired with an employee to test collaboration and problem-solving skills. If candidates need more than an hour, they can finish on their own, but a screening program called CoderPad tracks their work.
Last month, one candidate went from showing no work in CoderPad to suddenly having a complete solution, said Erik Schluntz, Cobalt Robotics’s chief technology officer and co-founder. He suspected the applicant had sought AI assistance and then copied and pasted its response.
The company declined to move forward with the candidate without telling the person why, though Mr. Schluntz tweeted about it.
Mr. Schluntz said Cobalt can’t properly evaluate candidates who use AI helpers today, but said he can envision giving applicants more challenging tasks in the future if they want to use tools like ChatGPT as an assistant.
“Giving a problem to someone that ChatGPT can solve doesn’t assess someone—it just assesses ChatGPT,” Mr. Schluntz said.
About a week after first spotting the AI-boosted applications, Ms. Qi started letting potential Databento hires use ChatGPT. The new prompt requires candidates to perform additional research and make edits to supplement what the AI tool spits out, and Databento gives “extra points” to people who complete the test bot-free.
Though Ms. Qi said she can usually spot when something was written by ChatGPT, the company also enlists the aid of a bot detector.
“It’s better to be ahead of the game and accept that people are using this rather than try to deny it,” Ms. Qi said.
—Cordilia James contributed to this article.
—For more WSJ Technology analysis, reviews, advice and headlines, sign up for our weekly newsletter.
Write to Ann-Marie Alcántara at ann-marie.alcantara@wsj.com
Copyright ©2022 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8
It was an unexpected problem. Earlier this year, Christina Qi, the chief executive of market data company Databento, noticed almost every job application included exactly what she was looking for.
The company prompts candidates to write a tweet and a press release about microwave towers, a niche topic that requires research, Ms. Qi said. Normally, most candidates fail the test. This time all five passed.
The tests—four from internship applicants and one from someone seeking a full-time content strategist role—were all so similar, “as if it was written by one person,” she said. Suspicious, Ms. Qi put the prompt into ChatGPT, the artificial-intelligence chatbot from OpenAI, to see what it could produce.
“Lo and behold, I got pretty much the same answer that all five candidates had submitted to me,” she said.
Since its launch in November, ChatGPT has been a hot topic at dinner tables and water coolers. Microsoft, Google, Snap and other companies have incorporated artificial intelligence into their products. People have experimented with using ChatGPT at work. Some have even started using it when looking for new roles, tapping the chat assistant to help write cover letters, tweak résumés and formulate responses to anticipated interview questions—without necessarily telling the people doing the hiring.
Employers, who have long used AI to screen potential employees, aren’t always disqualifying applicants who use ChatGPT, but they are scrambling to figure out how to assess candidates who may be using the tool to give them an edge.
Attention-grabbing applications
After being let go by his previous employer in January, Kyle Mickey started job hunting for roles in software engineering—alongside thousands of laid-off tech workers.
The 38-year-old from Colorado said he turned to ChatGPT for help, first sharing job descriptions and his résumé with the chatbot to see what it would tweak. Then he asked ChatGPT to write a recommendation letter for a role he coveted. The chatbot deemed him perfect for the job, as his technical skill set “aligns well with the requirements.”
Mr. Mickey sent the remarks to a recruiter, saying ChatGPT endorsed his skills.
“The recruiter was immediately like, ‘Let’s chat, I like the creativity,’ ” he said. Mr. Mickey didn’t get that job, but was hired at another company without ChatGPT’s help.
SHARE YOUR THOUGHTS
What part of the job-application process would you use AI for? Join the conversation below.
Ryan Stringham, 31, who lives in Utah and works in product management, used it to help write cover letters, including one that got him a foot in the door, and later hired, at a smart-tech company.
“You’re always looking, you’re always applying and you’re getting drained,” Mr. Stringham said of job hunting.
He said the bot broke his writer’s block, distilling his long-winded cover letter into four tight paragraphs. He said it also helped him prepare for job interviews by suggesting new ways for him to ask about company culture and expectations for the role. Instead of asking a vague question about what he should do to excel at the prospective job, ChatGPT suggested Mr. Stringham be more specific about the time frame and metrics for determining success.
Mr. Stringham has encouraged others to use chatbots in the job-hunt process, posting about them on LinkedIn and giving advice to other job seekers.
The only place he hasn’t disclosed his ChatGPT use: at work.
“It helped me get past the application process, and the recruiter never asked about it,” Mr. Stringham said, adding that he edited the cover letter himself and aced the interviews on his own.
‘How we present ourselves’
Programs, websites and other tools to help people fix their résumés and cover letters aren’t new. Microsoft Office and Google Docs offer résumé and letter templates, while companies such as Jobscan promise to optimize résumés to grab the attention of hiring managers, recruiters and hiring-system algorithms.
Candidates need to combine ChatGPT’s edits with their own editing and voice, said Sarah Baker Andrus, chief executive of Avarah Careers, a career coaching firm in Delaware. Whatever a candidate submits for a job should accurately reflect their skills, she said.
“We’re responsible for how we present ourselves,” Ms. Andrus said. “If you decide to use ChatGPT, it’s worthwhile to ask, ‘Is that representing the me that I want to present?’ ”
Employers are already finding ways to catch applicants who cheat with AI.
Engineers applying to San Francisco-based Cobalt Robotics take part in a remote one-hour coding interview where they are paired with an employee to test collaboration and problem-solving skills. If candidates need more than an hour, they can finish on their own, but a screening program called CoderPad tracks their work.
Last month, one candidate went from showing no work in CoderPad to suddenly having a complete solution, said Erik Schluntz, Cobalt Robotics’s chief technology officer and co-founder. He suspected the applicant had sought AI assistance and then copied and pasted its response.
The company declined to move forward with the candidate without telling the person why, though Mr. Schluntz tweeted about it.
Mr. Schluntz said Cobalt can’t properly evaluate candidates who use AI helpers today, but said he can envision giving applicants more challenging tasks in the future if they want to use tools like ChatGPT as an assistant.
“Giving a problem to someone that ChatGPT can solve doesn’t assess someone—it just assesses ChatGPT,” Mr. Schluntz said.
About a week after first spotting the AI-boosted applications, Ms. Qi started letting potential Databento hires use ChatGPT. The new prompt requires candidates to perform additional research and make edits to supplement what the AI tool spits out, and Databento gives “extra points” to people who complete the test bot-free.
Though Ms. Qi said she can usually spot when something was written by ChatGPT, the company also enlists the aid of a bot detector.
“It’s better to be ahead of the game and accept that people are using this rather than try to deny it,” Ms. Qi said.
—Cordilia James contributed to this article.
—For more WSJ Technology analysis, reviews, advice and headlines, sign up for our weekly newsletter.
Write to Ann-Marie Alcántara at ann-marie.alcantara@wsj.com
Copyright ©2022 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8