47%
Overall adoption rate
72%
Tech industry adoption
3.2x
Salary premium for experts
84%
Growth in prompt engineer jobs
In 2023, only 12% of knowledge workers had ever used a prompt engineering technique. By early 2025, that number has jumped to 47%—a nearly 4x increase in just 18 months, according to a new study by the Coursera Global Skills Report (2025) and proprietary surveys from Workings.me. But the real story isn't just the headline number—it's the uneven distribution across industries, roles, and demographics.
Key Findings
- Adoption is highest in tech (72%) and lowest in healthcare (18%), but the fastest growth is in education (+320% year-over-year).
- Job postings requiring prompt engineering grew 84% from 2024 to 2025 (LinkedIn Hiring Data).
- Only 9% of adopters can be considered “experts” (able to craft complex multi-step prompts reliably).
- Salary premium: Professionals with verified prompt engineering skills earn 32% more on average (Glassdoor 2025 salary analysis).
- Self-taught vs. formal training: 68% learned through online courses or tutorials; only 12% through employer-provided training.
- Generational divide: Gen Z (ages 18–27) adoption rate is 61%, compared to 29% for Baby Boomers (ages 60+).
- Tool-specific adoption: ChatGPT remains the most used (78% of adopters), followed by Claude (34%) and Gemini (22%).
Data Section 1: Adoption by Industry
The table below pulls from our own survey of 5,400 professionals across 12 industries, conducted January 2025, supplemented by the McKinsey State of AI 2025 report.
| Industry | Adoption Rate (%) | Year-over-Year Change | Avg. Proficiency Score (1-10) |
|---|---|---|---|
| Technology & Software | 72% | +38% | 6.8 |
| Marketing & Advertising | 58% | +45% | 5.4 |
| Education & Training | 43% | +320% | 4.1 |
| Financial Services | 35% | +40% | 4.8 |
| Healthcare | 18% | +25% | 3.2 |
| Legal | 22% | +60% | 4.0 |
| Manufacturing | 27% | +80% | 3.7 |
| Government / Public Sector | 14% | +30% | 3.0 |
Manufacturing's 80% year-over-year jump is notable—driven by AI-driven automation in production planning and quality control. Education's 320% spike likely reflects the rapid integration of AI tools into curriculum design and personalized tutoring.
Data Section 2: Adoption by Job Role & Seniority
We cross-referenced job title data from PayScale and our own user base. The following table breaks down adoption by role category.
| Role Category | Adoption Rate (%) | Median Salary with Skill | Salary Premium |
|---|---|---|---|
| Software Developers | 81% | $145,000 | +28% |
| Data Scientists / Analysts | 74% | $138,000 | +22% |
| Digital Marketers | 63% | $82,000 | +35% |
| Designers (UX/UI/Graphic) | 51% | $95,000 | +18% |
| Writers & Content Creators | 66% | $68,000 | +42% |
| Project Managers | 34% | $98,000 | +12% |
| HR & Recruiting | 25% | $72,000 | +15% |
| Executive / C-Suite | 38% | $220,000 | +8% |
Content creators and writers show a surprisingly high salary premium (42%), likely because prompt engineering directly amplifies output quality and speed—making skilled practitioners stand out dramatically. In contrast, C-suite adoption is lower (38%) but the premium is modest (8%), probably because many executives delegate prompting to assistants.
Data Section 3: Geographic Adoption
Regional data is drawn from Statista's Global AI Skills Index 2025 and our own survey. North America leads but adoption is skyrocketing in Southeast Asia and Africa.
| Region | Adoption Rate (%) | YoY Change | Top Learning Platform |
|---|---|---|---|
| North America | 54% | +40% | Coursera |
| Europe | 41% | +35% | Udemy |
| Asia-Pacific (excl. China) | 38% | +62% | Khan Academy / NPTEL |
| China | 45% | +70% | Baidu AI / Jike |
| Latin America | 29% | +55% | Platzi |
| Middle East & Africa | 22% | +85% | DataCamp / ALX |
The Middle East & Africa region is growing fastest from a low base, fueled by initiatives like ALX's AI training programs. China's rapid adoption (45%, +70%) is partly due to government mandates and the proliferation of domestic LLMs.
What The Data Tells Us
Three dominant patterns emerge:
- Adoption is superficial. While nearly half of professionals have “tried” prompt engineering, only 9% achieve advanced proficiency. Most still use simple single-turn prompts rather than multi-shot or chain-of-thought techniques.
- Salary premium is real but uneven. For roles where AI directly enhances output (writers, marketers, developers), the premium is 22-42%. For roles where AI is more of a support tool (PMs, executives), the premium is much smaller.
- Self-directed learning dominates. Over two-thirds of adopters learned on their own via online courses or YouTube. This suggests a huge opportunity for employers who invest in formal training—only 12% currently do.
"I took a weekend course on prompt engineering and immediately saw my output quality double. But when I applied for a promotion, my manager said the skill wasn't 'certified'—so now I'm spending more time on credentials than on actually using AI better." – Maria K., Senior Content Strategist at a fintech startup
Consistent with Maria's experience, many professionals report that while they can use prompts effectively, they struggle to signal that competency to employers. This is where tools like our Career Pulse Score can help—it measures your skill adoption against industry benchmarks and gives actionable feedback on where you stand relative to peers.
Methodology Note
Data for this report comes from three sources: (1) A proprietary survey of 5,400 professionals conducted via Workings.me in January 2025; (2) The Coursera Global Skills Report 2025 (n=98,000 learners); (3) Job posting analysis from LinkedIn's Economic Graph team (2024 vs. 2025). Industry and role classifications follow standard O*NET categories. Adoption is defined as having used prompt engineering (structured, multi-step instructions to an LLM) at least three times in the past month. Proficiency was self-reported on a 1-10 scale and cross-validated with a short test in a subset of respondents.
Disclaimer:
All data is aggregated and anonymized. Salary figures are median values for roles with and without the skill, adjusted for experience and location. Individual results may vary.
Deep Dive: Why Proficiency Lags Behind Adoption
One of the most striking findings is the gap between adoption (47%) and advanced proficiency (9%). Why are so many people trying prompt engineering but failing to master it?
Part of the reason is the rapid evolution of the tools themselves. A technique that worked six months ago—like "role prompting"—may now be obsolete as models become more capable of inferring context. According to research from Liu et al. (2024) at Stanford, the half-life of prompt engineering best practices is roughly 8 months, meaning half of what you learn today will be outdated within a year. This forces continuous learning, which many professionals aren't equipped for.
Another factor is the lack of effective training. While 68% of adopters used online courses, most of those courses are basic tutorials that teach simple one-shot prompts. Fewer than 10% of courses cover advanced techniques like chain-of-thought, few-shot prompting, or output structuring. The result: a lot of people know how to ask ChatGPT to write an email, but can't use it to generate complex data analysis reports or automated workflows.
Scenario 1: The Self-Taught Generalist
You’ve watched a few YouTube videos and can get ChatGPT to draft blog posts, summarize articles, and generate boilerplate code. But you struggle when the output needs to be precise, consistent, or adhere to specific formatting. You’ve never used system prompts or temperature settings. This describes about 38% of all adopters—the “casual users.”
What to do: Invest in structured learning. Platforms like DeepLearning.AI offer free courses on advanced prompting. Also, practice with a variety of models—Claude, Gemini, and open-source tools like Llama 3—to understand differences in behavior.
Scenario 2: The Power User
Approximately 12% of adopters fall into this category. They use multi-turn conversations, define personas, and adjust parameters. They can get reliable outputs even for complex tasks. But they often hit a ceiling: they don't know how to integrate prompts into APIs or automate workflows.
Next step: Learn to code (even basic Python) or use no-code automation tools like Zapier or n8n. Prompt engineering becomes exponentially more valuable when embedded into a pipeline—like generating weekly reports on autopilot.
Scenario 3: The Expert (Top 9%)
Experts write prompts that are robust, reusable, and maintainable. They use structured outputs (JSON, XML), create few-shot examples dynamically, and troubleshoot failures systematically. They often build custom GPTs or use LangChain to orchestrate multiple calls.
To reach this level: Study prompt design patterns—like those from the DAIR Prompt Engineering Guide. Practice on varied tasks: data extraction, creative writing, code generation, reasoning. Track your success rate over time.
The Future of Prompt Engineering: Will It Even Be a Separate Skill?
A popular debate is whether prompt engineering will become obsolete as models improve. If you can just tell an AI “what to do” in natural language, why learn special techniques? Data from our survey suggests that even with GPT-5 and Claude 4, expert prompters outperform novices by 40% on complex tasks (e.g., multi-step reasoning, domain-specific content). The gap narrows but persists.
In fact, the Anthropic research team found that as models improve, prompt engineering evolves rather than vanishes. Early techniques (e.g., “think step by step”) are absorbed into the model's training, but new techniques emerge (e.g., “use chain-of-thought with structured output”). So the skill is not going away—it's changing.
How to Benchmark Your Own Skill Level
Given the fast pace, periodic self-assessment is crucial. That's where our free Career Pulse Score comes in. It asks a series of questions about your AI tool usage, proficiency, and learning habits, then compares you to others in your industry and role. You'll see exactly where you stand—and whether you should invest more time in prompt engineering or focus on other skills.
Insider Tip:
Don't just rely on course certificates. Build a small portfolio of prompt examples that solve real problems—like a custom content generator or a data extraction script. Employers (and the Career Pulse Score) value demonstrated ability over credentials.
FAQ
Q: What exactly counts as “prompt engineering” in your data?
We define prompt engineering as the practice of deliberately crafting instructions or input sequences to elicit desired outputs from large language models. This includes techniques like role assignment, system prompts, few-shot examples, chain-of-thought prompting, and output formatting. Simple single-turn questions (e.g., "Write a poem about cats") without any structure or context setting are not counted as prompt engineering—that’s just basic usage.
Q: How accurate are self-reported proficiency scores?
We validated self-reported proficiency through a small pilot study where a subset of respondents (n=400) completed a standardized test with 10 tasks of increasing complexity. The correlation between self-reported score and test score was r=0.74, indicating reasonable accuracy. However, most people slightly overestimate their ability (by ~0.7 points on average on a 10-point scale).
Q: Which industry had the biggest surprise in adoption growth?
Education, by far. The sector went from essentially zero (3% adoption in 2023) to 43% in 2025—a 14-fold increase. This is driven by teachers using AI to generate lesson plans, quizzes, and personalized learning materials. Schools and universities are now increasingly including prompt literacy in their curricula.
Q: Is it worth learning prompt engineering if you're not in tech?
Absolutely. While tech has the highest adoption, the highest salary premiums are actually in non-tech roles like content creation (42%), marketing (35%), and even legal (+22%). As AI tools spread, professionals who can skillfully direct them will have a significant edge in any knowledge-based role. Use our Career Pulse Score to see how prompt engineering skills affect your specific career path.
Q: What is the best way to get started if you have zero experience?
Start with a course like ChatGPT Prompt Engineering for Developers (free) or the Vanderbilt Prompt Engineering Specialization. Then practice daily for 15 minutes, trying to solve real tasks you encounter at work. Track what works using a simple log.
Q: How do I know if my prompts are “good”?
Define success criteria before writing the prompt: accuracy, completeness, style, format. Then iterate. A good metric is the percentage of times your first attempt produces a usable output—experts achieve >70% on familiar tasks. Novices often need 3-5 attempts. Regularly testing yourself with varied tasks (e.g., from Awesome ChatGPT Prompts) can accelerate improvement.
Q: Will AI replace the need for prompt engineers?
In our opinion, no. As models become more capable, the skill shifts from “how to phrase a prompt” to “how to design a task decomposition and validate outputs.” The best analogy is the evolution from assembly language to high-level programming—programming didn't disappear; it became more abstract. Prompt engineering is becoming a core digital literacy, like typing or using spreadsheets. The data shows that adoption is accelerating, and those who ignore it risk falling behind.