AI is everywhere in conversations about software engineering. Developers are learning to use AI coding assistants, companies are adding AI tools to their workflows, and job seekers are wondering which AI skills they need to stay competitive.

But job-market discussions often mix together three different things:

• Employers asking for AI-related skills.
• Engineers using AI tools in their day-to-day work.
• Engineers building AI-powered systems.

These are related, but they are not the same.

I wanted to investigate what software engineering job advertisements actually say.

Looking at 6,439 job descriptions

I analysed a dataset of 6,439 software engineering job descriptions and examined the AI-related skills appearing in their extracted skill profiles.

In this dataset, 44.3% of job descriptions had at least one selected AI-related skill in their extracted profile.

That is a useful signal, but it needs context.

An extracted skill profile is not the same as a list of mandatory qualifications. A skill appearing in a profile does not automatically mean the employer requires every applicant to have it.

The figure also depends on which skills are classified as AI-related and how they are extracted from job descriptions.

So the finding is best understood as a measurement of what appeared in this dataset—not proof that 44.3% of all software engineering jobs require AI expertise.

What about using AI tools at work?

A job advertisement might mention machine learning, while another might describe using AI assistants to improve an engineering workflow. Those are different kinds of signals.

To explore that distinction, I manually reviewed a balanced sample of 30 job descriptions.

Descriptions with AI-related skills in their extracted profiles

8 of 15 job descriptions (53.3%) explicitly discussed AI tools in engineering work.

Descriptions without AI-related skills in their extracted profiles

3 of 15 job descriptions (20%) explicitly discussed AI tools in engineering work.

This was a balanced sample of 30 job descriptions, not a representative survey of the entire job market.

In this small sample, descriptions with AI-related skills in their extracted profiles were more likely to explicitly discuss AI tools in engineering work.

But the sample is small and was not designed to represent the entire job market. These percentages should not be treated as population estimates.

The comparison is useful as an exploratory observation, not a definitive conclusion.

It also shows why simply counting mentions of AI may not tell the whole story.

Three questions worth keeping separate

When evaluating AI’s place in software engineering, I think it helps to ask three separate questions.

1. Which AI-related skills appear in job advertisements?

This tells us what employers mention in the roles they publish.

2. Which AI tools do engineers use in their work?

This concerns how software development itself is changing. An advertisement may discuss AI-assisted workflows without requiring candidates to build machine learning systems.

3. Which roles involve building AI systems?

These roles may require more specialised knowledge, such as machine learning, model evaluation, data engineering, or deploying AI-powered applications.

Treating all three as one category can make the market seem more uniform than it is.

What does this mean for early-career developers?

If you’re trying to land your first software engineering role, it’s tempting to interpret every new technology trend as another requirement you must master immediately.

I wouldn’t draw that conclusion from these findings.

Instead, use job advertisements to investigate the kinds of roles you’re targeting. Look at the actual responsibilities, distinguish essential qualifications from desirable ones, and examine whether AI appears as a development tool or as the core of the product being built.

Most importantly, don’t mistake a skill’s presence in job advertisements for proof that every employer expects every candidate to have it.

The goal is to make decisions using better evidence, not to chase every trend.

How I’m exploring this further

I built Glassbox to help developers investigate what software job advertisements ask for.

It collects and analyses public software job listings, counts requested skills, and lets people compare their CVs with skills appearing in job descriptions.

The idea is straightforward: make job-market signals easier to examine so developers can make more informed decisions about their next steps.

The data won’t make career decisions for you. It can help you ask better questions.

We show, you decide.

I’d be interested to hear how other developers are seeing AI appear in software engineering job descriptions. Are employers asking you to build AI systems, use AI tools in your workflow, or both?