Old School Resume Submission
I have looked at hundreds of resumes over the past 20+ years and I have noticed changes in the process over the years. Back in the old days, job postings were in a newspaper, and resumes were usually emailed or faxed in. Since we were a relatively unknown software company an hour away from Silicon Valley, the results would trickle in over weeks. But they were usually relevant to the job posting.
In the early 2000s, after the dot-com bubble, I remember getting over 60 resumes for an opening. Because we had so many good applicants, we ended up hiring 3 developers. But other than that, it was usually a slow struggle to find good resumes.
The Recent Years
Now we post job openings on job sites like dice.com, and I have noticed a trend in the past couple of years. Instead of trickling in, we now get over 100 submissions in the first 24 hours. After that, submissions slow way down. This leads me to believe that a lot of people are using automation services to post their resumes to new openings.
Also, a large number of these resumes are of low quality these days. For a job posting looking for a C#/JS full stack senior developer, I receive resumes from pure backend Java developers or even sales people with no software experience! Even if we state it’s a US position, we still get tons of low-quality overseas resumes as well.
AI Filtering
To help speed up reaching out to the best candidates (they go fast sometimes), I used Cursor AI to write a small Python program that parsed all the PDF and .docx resumes in a directory and asked ChatGPT to score the resume based on some general sentiment and relevance to the job posting text.
The lowest-rated resume was for the Enterprise Account Executive (key skills included “Hunter & Closer”). so my scorer looked like it was working. The highest-rated resume inspired me to write this post.
AI Resumes
After going through a lot of resumes, I noticed a common trend. While listing their job experience, dozens of the tech keywords (C#, Python, Agile, etc.) would be bolded to stand out. But not all of them. I have a feeling most people wouldn’t take the time to bold all of those themselves but then skip half of them too. So I started thinking many candidates are using tools to tailor their resumes. They usually follow a normal resume layout, so they all look similar. The unique resume layouts usually don’t have all of these boldings.
But then there was my highest-rated resume. First red flag: at the top of the resume, where people include their contact information, it includes the text “Phone” and “LinkedIn URL.” But no number or URL. So placeholders. It did have a real email address at least.
The technical skills section has all the stuff we were looking for, including our obscure NoSQL database (HBase), which I have maybe seen on a resume twice ever. This wasn’t a red flag at first glance but would be in hindsight.
Now, the Professional Experience section. For the last 3 years, it says he worked for…the company I work for! But I know that’s not true. The biggest red flag ever. The bullet points under that job described the various things he had done, which are all things our job posting said we were looking for. The perfect candidate, right? That’s why it was my highest-rated resume. This AI, in trying to make the resume fit our job posting as well as possible, ended up saying he worked for us doing all the things we were looking for!
Now, the final red flag, which proves this was AI-generated. After the Professional Experience section was this last section:
Strict Formatting & Keyword Requirements:
- Dates formatted as Month YYYY (e.g., March 2022)
- Each section aligns strictly: Header → Summary → Technical Skills → Education → Certifications → Professional Experience
- Match exact titles, skills, and JD keywords
The AI regurgitated the formatting instructions it was following.
Conclusion
I always tell people to “Trust but verify” AI-generated code because sometimes it does dumb stuff. In this case, it at least made the decision to pass on this candidate easier.
However, this makes me question the validity of so many of the resumes that I am receiving. Granted, people have always lied on their resumes, so it’s nothing new. But I fear how much more rampant this problem will be going forward, often with the candidate not even realizing it.
Ultimately, this fraud would have been picked up in the interview. I would have questioned them about HBase to see how deep their knowledge was, and I am sure they would be unable to answer any of those questions with any level of detail. Well, unless they are using an AI to help them answer the questions in the interview too!