Advice for Employers and Recruiters
Ending the keyword arms race: How human-centered job descriptions improve your talent pool
We recently shared a list of the 10 things that students, recent graduates, and others who are early in their careers hate the most about AI-powered hiring systems. Today, we’re going to dive more deeply into the fifth: the understanding that virtually all medium- and large-sized employers use applicant tracking systems that rank, match, or score based in large part on the use and even placement of somewhat arbitrary keywords.
Imagine a brilliant graduating senior—let’s call her Sarah. Sarah has a 3.8 GPA in Data Science, completed a high-impact internship at a major tech firm, and teaches coding workshops to high school students on weekends. She’s exactly the kind of early-career talent you want to hire.
Now, imagine Sarah’s frustration. She opens a job description for an entry-level Analyst role at your company. Instead of seeing an opportunity to launch her career, she sees a wall of corporate jargon, acronyms, and a list of 25 mandatory requirements that look like they were generated by a thesaurus.
She knows she can do the job. But she’s been told by her campus career center, her mentors, and endless internet articles that to get past your AI-powered Applicant Tracking System (ATS), she must “match the keywords.”
So, Sarah spends three hours. She doesn’t spend them reflecting on her achievements or crafting a compelling narrative. She spends them “optimizing”—painstakingly replacing her own natural, authentic descriptions of her skills (“Collaborated with cross-functional teams to analyze dataset ‘X'”) with the robotic phrases in your JD (“Demonstrated synergy within matrixed organizational structures utilizing dataset ‘X'”).
This is the Keyword Optimization Arms Race. It is a process that early-career candidates hate, that strips authenticity from resumes, and that is actively damaging your ability to identify true talent.
In part five of our series on AI hiring hurdles, we’re shifting the focus from the candidate’s resume to the employer’s responsibility. The fix isn’t teaching candidates how to lie better; it’s teaching employers how to write better. We must stop using jargon-filled laundry lists and start using human-centered job descriptions that invite natural language applications.
The Cause of the Crisis: Why Jargon Creates Bots
To fix the keyword arms race, we must understand its origin: the bad job description.
Most job descriptions in the corporate world are ancient artifacts. They are copied and pasted, year after year, with more technical jargon added by each successive manager. They are “exclusive” documents, listing everything a candidate must already have, rather than “inclusive” documents, detailing what a candidate will achieve and learn.
When your JD is a rigid list of acronyms (SQL, Python, SEO, PPC, CRM, ERP, KPI) and corporate cliches (“highly motivated self-starter,” “think outside the box”), the candidate has two options:
- Apply using their authentic voice: This is high-risk. Their story of learning Python during a summer boot camp might not use the exact phrase “Mastery of Python Syntax,” causing the older or poorly configured AI models of 2026 to down-rank them, despite their massive potential.
- Exaggerate or “Keyword Stuff”: This is the logical choice. They repeat your JD phrases verbatim, sometimes even “white-fonting” them (hiding keywords in white text so only the AI sees them).
When you force candidates to optimize, you don’t get the best candidates; you get the candidates who are best at gaming the bot. In 2026, where early-career talent is hyper-aware of AI, writing a jargon-heavy JD is effectively a signal to the most calculating (and perhaps least authentic) candidates.
The Solution: Rewriting for “Context” and “Impact”
You don’t have to abandon keywords entirely. But you must shift how you ask for them. The goal is to move from keyword matching to semantic matching—from “Do they have this exact word?” to “Do they demonstrate understanding of this concept and context?”
Modern AI systems, when properly configured, are excellent at understanding intent. But they can only work with the context they are given. Your JD provides that context.
Here is how you rewrite your job descriptions to encourage natural language and eliminate the need for candidates to stuff their resumes.
1. Prioritize “Context” Over Acronyms
Instead of listing a software tool as a prerequisite, list the purpose it serves in natural language.
- Jargon JD: “Must possess high proficiency in Oracle NetSuite ERP.”
- Human JD: “We use Oracle NetSuite to manage our core financials and inventory. While we don’t expect you to be an expert on day one, we are looking for someone who is quick to learn complex database systems and understands basic accounting principles.”
The Difference: The “Jargon JD” requires a candidate to have NetSuite experience, which few entry-level people do. It invites exaggeration. The “Human JD” describes the why. It allows a candidate with experience in any database tool to explain how their “transferable skills” make them a strong candidate, without needing to stuff the resume with “NetSuite.”
2. Focus on “Impact” and “Learning,” Not a Past Duties Laundry List
Early-career talent should be hired for where they are going, not just where they have been. Stop describing the “tasks” and start describing the “impact.” This invites a candidate to tell their story.
- Jargon JD: “Duties include managing Excel spreadsheets for KPI tracking and CRM database entry.”
- Human JD: “Your main impact will be to ensure our marketing team has accurate data to make decisions. You’ll become the owner of our performance dashboards (we use Excel and Salesforce), learning to track key metrics and provide regular insights to leadership. Training will be provided, but curiosity is essential.”
The Difference: The “Duties” list is boring. The candidate only knows to stuff the resume with “Excel” and “KPI.” The “Impact” section tells a story. It invites Sarah to write about her experience organizing a charity event’s budget (managing data) and learning a new scheduling tool quickly (demonstrating curiosity). She can use her own words, and your semantic AI will understand the correlation.
3. Define the “Transferable Skills” in Plain English
For entry-level roles, you are hiring for potential. Soft skills (grit, resilience, critical thinking, adaptability) are paramount. But “critical thinking” is a keyword cliché. Define what it means for this role.
- Jargon JD: “Highly motivated self-starter with excellent critical thinking skills.”
- Human JD: “We move fast, and priorities shift. We are looking for someone who isn’t afraid to ask ‘why’ when something doesn’t look right, who enjoys figuring out puzzle pieces that seem conflicting, and who thrives on feedback. We don’t want robots; we want thinkers.”
The Difference: This JD explicitly signals that personality matters. It invites a candidate to write naturally about their experience—perhaps navigating a difficult group project or self-teaching a new skill outside of class—because you’ve defined the behavior, not just used a generic label.
4. The “Anti-Keyword” Disclaimer: A Radical Move for 2026
If you want to build trust and eliminate resume stuffing instantly, be explicit. Tell the candidate that you value their voice over their vocab-guessing skills.
- The Trust-Builder: Add a short box at the top or bottom of every entry-level JD:“A Quick Note on AI: We use an AI assistant to help our recruiters sort through the thousands of applications we receive. We are configured to value context and skills over exact keywords. Don’t spend hours trying to guess the ‘right’ words or stuffing your resume. Tell your authentic story in your own words. We’re interested in who you are, not how well you can copy our job description.”
The Result: This is the ultimate “fix.” It immediately reduces candidate anxiety, creates a massive advantage for your employer brand on platforms like Glassdoor and Reddit, and ensures you get resumes that are readable by actual humans, not just optimized for bots.
Conclusion: Rewrite Like a Human to Hire a Human
Early-career talent entering the 2026 job market is looking for connection and authenticity. If your first interaction with them is a jargon-laden, robotic demands document, don’t be surprised when you receive a robotic application in return.
Keywords are necessary for filtering, but they are not the point of hiring. The point is finding the human potential. By rewriting your job descriptions to emphasize context, impact, and learning—and by explicitly inviting natural language—you break the optimization arms race.
You don’t just speed up your process; you improve it. You shift the focus from “Do they match the words?” to “Can they achieve the goal?” and in a world where AI is doing the initial sorting, that distinction is everything.
Next in the Series: We’re tackling technical glitches and digital inequality—why an unstable Wi-Fi connection during an AI-monitored assessment is filtering out your most resilient (but perhaps less affluent) candidates.
