AI & Writing

Writing Practice Builds AI Prompting Skills

9 min read AI & Writing
Quick Takeaways
  • AI skills boost job results mainly through creative self-belief and metacognitive skill (79% of the effect)
  • Hayes' 30 years of writing research and 2024 AI studies both point to metacognition as the key limit
  • Six skills transfer: task splitting, clarity, revising, error spotting, metacognitive tracking, working memory control

Knowing AI tricks does not mean we use them well. A 2025 study found ChatGPT helps more with facts than with hands-on skills.[1] Results depend on how students engage. Those who just accept AI output gain less. Those who judge it gain more. Skill alone is not enough.

Research

AI skills boost job results mostly through creative self-belief and metacognitive skill (79% of the effect flows through creative trust). Self-guided learning predicts AI writing results better than tech know-how.

The Metacognitive Demands of AI

Three core needs shape good AI work:

1. Writing the Prompt

We must know our real goals. We break hard tasks into clear sub-goals. We spot vague spots. We state our intent with care.

2. Revising the Prompt

We need metacognitive flex to see when a try fails. Then we shift our plan. This takes well-tuned trust in our own grasp.

3. Judging the Output

We need sharp judgment to catch errors and false claims. We check if the output serves our goal.

The hard part: fewer than half of students use basic metacognitive habits. Few ask for help when stuck or link new tasks to what they know.

What Writing Research Teaches Us

Expert writers loop through planning, writing, and reviewing. They watch their own progress the whole time. Hayes and Flower's 1981 theory (3,800+ citations) found this loop. Hayes' later updates (1996, 2012) put more weight on metacognitive control.

Hayes' 2012 model puts a Control Level at the top. It runs all writing steps. This shift shows that metacognitive control is what sets experts apart from novices.

Six Transferable Skills Developed Through Writing Practice

  1. Task splitting – Breaking big goals into small steps
  2. Clarity – Stating intent with no blur
  3. Revising in loops – Fixing based on what we spot
  4. Error spotting – Seeing when output misses the mark
  5. Metacognitive tracking – Watching our own thinking
  6. Working memory control – Handling cognitive load well

These map right onto what AI work demands.

The Theoretical Bridge: Convergence from Two Directions

Two lines of research found the same chokepoint. Hayes' 30 years on writing and 2024 AI work both point to metacognitive control as the key limit in hard mental tasks.

One caveat: The logic is strong. But no study yet tests if writing practice leads to better AI use. This is both a gap and a chance for new research.

Putting It Into Practice: What Kind of Writing?

Four High-Leverage Practices
  • Timed Freewriting (15 min) – Forces clear words under pressure. Builds speed in turning thought to text.
  • Outlining Drills – Trains task splitting. Breaking big goals into parts is the same skill prompts need.
  • Revision Practice – Builds our inner critic. Write a draft, wait 24 hours, then revise. Grows ease with looped fixes.
  • Explain-It Writing – Practice making hard ideas simple. This trains clarity, a skill we need when prompting AI that lacks context.

What We Know and Don't Know

Established:

  • Writing practice builds metacognitive skills
  • AI work needs metacognitive skills
  • The overlap is large
  • Hayes (30 years) and AI studies (2024) both found metacognition at the core

Unknown:

  • Whether writing practice boosts AI work
  • Which drills help most
  • How long the shift takes
  • How it varies by person

Conclusion

Most AI training teaches tricks. But if metacognitive skill is the real limit, tricks only treat symptoms.

We can teach chain-of-thought prompting in five minutes. But without the awareness to split goals or notice a bad path, the trick will not help.

Writing practice is an overlooked way to build AI skill. It grows the core strengths that make tricks work.

References

  1. ^ Liu, X. (2025). The Impact of ChatGPT on Students' Academic Achievement: A Meta-Analysis. Journal of Computer Assisted Learning. https://doi.org/10.1111/jcal.70096
  2. ^ Hayes, J. R., & Flower, L. (1981). A Cognitive Process Theory of Writing. College Composition and Communication, 32(1), 365-387.
  3. ^ Hayes, J. R. (1996). A new framework for understanding cognition and affect in writing. In C. M. Levy & S. Ransdell (Eds.), The Science of Writing (pp. 1-27). Lawrence Erlbaum.
  4. ^ Hayes, J. R. (2012). Modeling and Remodeling Writing. Written Communication, 29(3), 369-388. https://doi.org/10.1177/0741088312451260
Share