AI is quickly becoming a source of career advice. Students and adults can ask a large language model (LLM) to recommend occupations, compare degree programs, improve a resume, or plan a career transition in seconds.
That accessibility has real value. At a time when personalized career counseling remains difficult to provide at scale, AI can help people find information and explore unfamiliar options.
But career guidance isn’t just about organizing information. It’s about uncovering potential.
There is an important limitation: An LLM typically begins with what the user already knows—or thinks they know—about themselves.
Career guidance should help people discover what they don’t yet know about themselves.
LLMs Can Only Work With the Story They Are Given
Ask an AI system, “What career is right for me?” and it will usually ask about interests, favorite subjects, experiences, goals, and skills.
Those are reasonable questions. They are also incomplete.
Interests are influenced by exposure. Experience depends on access. A person’s description of their abilities may reflect confidence, stereotypes, or years of feedback from others.
A student who has never written code may not identify technology as an interest. Someone who has never met an engineer may not see engineering as attainable. A young person repeatedly told that they are “not a math person” may exclude analytical fields without ever receiving objective evidence of their capabilities.
An LLM can make inferences from the information it receives. It cannot independently measure a capability the user has never experienced, recognized, or mentioned.
That is not simply a prompting problem. It is the limitation of building guidance primarily on self-reported information.
AI Can Reinforce the Limits of a Person’s Environment
AI systems learn from patterns in existing information. That makes them useful, but it also creates the risk that they will reproduce assumptions already present in society or in the data used to develop them.
The National Institute of Standards and Technology identifies three broad categories of AI bias: systemic bias, computational and statistical bias, and human-cognitive bias. These forms of bias can arise even without prejudice or discriminatory intent. (www.airc.nist.gov)
Career guidance is particularly vulnerable because career exposure has never been distributed equally.
Consider a student who says she enjoys helping people and has primarily encountered nursing, teaching, and social work. An LLM may reasonably suggest similar occupations.
But what if she also has strong spatial visualization, numerical reasoning, or pattern-recognition aptitudes? Those capabilities could point toward careers in engineering, architecture, data science, and other fields she may never have considered.
The AI response might be supportive and logically sound while still reinforcing the boundaries of her current environment.
Interest Matters, but It Is Not the Same as Aptitude
Interests should remain part of career guidance. They tell us what currently attracts or motivates someone.
Aptitudes answer a different question: How is this person naturally inclined to process information and solve problems?
Skills show what the individual has already learned. Experiences provide evidence of how those capabilities function in real settings.
Aptitude is not destiny, nor should it prescribe a single career. It is one important source of objective evidence that can expand possibilities before narrowing choices.
Strong guidance should incorporate objective aptitude evidence alongside interests, demonstrated skills, experiences, values, and goals.
No single signal should become a verdict.
This distinction matters because job titles are becoming less stable. Someone who understands only that they are “a programmer,” “an accountant,” or “a welder” may feel stranded when a role changes.
Someone who understands the capabilities beneath that title can see how those capabilities transfer. Careers change. Human capability endures.
Job Titles Are Shifting Faster Than Guidance Models
The World Economic Forum’s Future of Jobs Report 2025, based on input from more than 1,000 employers representing over 14 million workers globally, offers useful context for the scale of this change.
The report estimates that 39% of workers’ existing skill sets will be transformed or become outdated by 2030. Employers also expect greater demand for creative thinking, resilience, flexibility, agility, and technological literacy. (www.weforum.org)
The Organization for Economic Co-operation and Development (OECD) has reached a related conclusion. Its research finds that most workers exposed to AI will not need specialized expertise in fields such as machine learning or natural language processing.
AI will, however, change many of the tasks they perform and the broader skills their jobs require. (www.OECD.org)
This makes static job matching increasingly inadequate.
Career guidance cannot simply select an occupation from today’s list and construct one linear route toward it. It must help people understand the durable capabilities beneath occupational labels.
The better question is not only, “Which career matches me?” It is also, “What kinds of problems am I equipped to solve, and where else could those capabilities create value?”
Students Need Evidence About Themselves
The readiness crisis among recent graduates shows what happens when people make consequential decisions without enough self-knowledge.
Recent national research on post-graduation readiness found that 77% of recent graduates felt only moderately, slightly, or not at all prepared for what came next. Sixty-nine percent lacked strong confidence in their post-graduation plans.
The same research found that graduates were following more varied pathways than the guidance system often assumes. Only 37% were pursuing a traditional four-year degree, while the remaining 63% were entering two-year programs, technical education, the workforce, gap years, or other routes.
They were not asking for less rigor. They wanted a stronger connection between school and the decisions ahead of them.
Fifty-five percent wanted more real-world work experience before graduation. Forty-three percent wanted more hands-on, career-connected coursework, and 40% wanted structured help understanding their aptitudes and direction.
The same research also found that graduates who received structured career guidance connecting their demonstrated aptitudes with education and career pathways reported substantially higher levels of preparedness and confidence than those who did not.
Students do not need another system that asks them to make a better guess. They need evidence that expands their understanding of what they can do.
Good AI Guidance Should Open Options, Not Issue Verdicts
The future of AI-supported guidance should not be a chatbot that declares, “You should become an engineer.”
That would replace one rigid recommendation model with another.
A better system would begin with a richer understanding of the individual—including objective aptitude evidence—then incorporate interests, skills, experiences, and personal goals.
AI could use that foundation to explain why certain career functions may fit, identify adjacent pathways, and show where further education or experience may be needed.
It should also explain its reasoning. Which capabilities does a role use? What other occupations rely on similar strengths? Where might the person need additional practice or support?
The OECD has found that digital technologies can expand access to career guidance and create opportunities for more personalized support. However, those benefits depend on how the tools are designed and used. (www.OECD.org)
Technology does not automatically produce better guidance simply because it can reach more people.
AI should function less like an oracle and more like an evolving guide. After someone completes a project, certification, job shadow, internship, or first job, the system should help interpret that experience and identify what it opens next.
The objective is not to predict one perfect career. It is to help someone make a stronger next decision.
Career Guidance Must Build Adaptability
In a more stable labor market, career guidance could focus heavily on matching people to established occupations.
That is no longer sufficient.
People will need to develop new skills, move between functions, and reconsider how their capabilities create value as work changes.
The strongest foundation for that adaptability is not certainty about a job title. It is a deeper understanding of one’s own capabilities.
When the ground is shifting, people need bedrock.
LLMs can help explain careers, compare pathways, identify adjacent roles, and turn insight into action. But they should not be mistaken for instruments that can independently reveal human potential.
The future of career guidance should combine the reach and responsiveness of AI with objective aptitude data, real-world experience, and human judgment.
AI can help people navigate the map. First, we must help them understand the traveler.
