Five-second answer: Going blank does not automatically mean poor vocabulary or low working-memory capacity. Producing language requires several partly separable operations: deciding what to express, retrieving relevant knowledge, selecting words, building an utterance, monitoring it, and producing it. Someone may be extremely fluent in live conversation yet dislike composing text, while another person may clearly feel “that was fun” but fail to generate a free-form message. AI is especially useful when it externalizes the intermediate operations without inventing the person’s meaning.[1][2]
1. Why can “I had fun, thank you” be the end of the road?
Imagine someone who enjoyed an outing and can easily say, “It was fun” and “Thank you.” Then ask for a short letter, and the mind goes blank.
Now imagine another person who, face to face, can turn one remark into five branches: a film becomes a discussion of trends, confusion, similar experiences, hobbies, work, and the next question.
It is tempting to call this a vocabulary difference or a working-memory difference. But language production is not one single ability.
Classic psycholinguistic models separate conceptual preparation, lexical selection, grammatical and phonological encoding, and articulation.
Feeling something and converting that feeling into a polished message are therefore different operations.
No words does not necessarily mean no inner content.
2. Vocabulary, retrieval, ideational fluency, and working memory are not the same thing
A computer analogy helps.
- Vocabulary is the size and organization of the database.
- Lexical retrieval is how quickly the right item can be found.
- Verbal fluency is the ability to keep retrieving responses under a rule.
- Ideational fluency is the ability to generate several possible ideas or branches.
- Working memory keeps and manipulates currently active information.
- Monitoring checks whether an output is appropriate.
They interact, but they are not interchangeable.
A 2023 meta-analysis covering 79 studies and 12,846 participants found a small overall relationship between memory and creative cognition, with semantic memory—especially strategic verbal retrieval—playing an important role. Working memory was more strongly associated with convergent than divergent creative thinking.[1]
So a large “RAM” does not automatically produce endless conversation.
A smaller buffer can coexist with a very fast search engine.
3. “Write anything you want” is secretly a hard task
Open-ended prompts sound easy because they remove restrictions.
In reality, they remove the search key.
“Write whatever you felt” requires the writer to choose the topic, decide what is relevant, infer the expected level of detail, organize it, phrase it, and decide when the answer is finished.
“Which moment was the most enjoyable?” is easier because it provides a retrieval cue.
“What did the other person do that you appreciated?” narrows the search again.
“What would you like to do next?” supplies a future-oriented cue.
Writing research has long described composing as an interaction among planning, translating ideas into text, and reviewing the result.
A blank page is not empty freedom.
It is a project with no specification.
4. A blank mind is not evidence of weak feelings
“I know it was fun, but I do not know what to say” can arise from multiple bottlenecks.
Perhaps the person cannot spontaneously select an episode. Perhaps the episode is available but slow to verbalize. Perhaps candidate sentences appear but are rejected immediately. Perhaps open prompts fail while direct questions work well. Perhaps the expected form or evaluation standard is unclear.
Writing-anxiety research also suggests that worry and heavy self-monitoring can compete with the cognitive resources needed for idea generation, organization, and wording.[3]
None of this allows diagnosis from one behavior.
The useful question is not “What is wrong with this person?”
It is “At which stage does output start moving again?”
5. Second-language speech makes the distinction obvious
In a native language, a person may instantly say, “Yes, that is exactly what I mean.”
In a second language, the concept may be perfectly clear while the linguistic route is slow:
“What was that phrase?” “I know the word.” “By the time I build the sentence, the conversation has moved on.”
Research on L2 fluency examines exactly this gap between knowledge and processing speed. Lexical and collocational processing speed and automaticity are associated with fluent speech production.[4]
A strong concept-to-L1 route can coexist with a weak concept-to-L2 route.
Likewise, some people may have a relatively weak route from experience and emotion to free-form written language even in their first language.
The content may be there. The conversion is simply not automatic.
6. How can someone who once hated essays become good at writing later?
A person may struggle with school essays or repeatedly erase creative work because every attempt feels “wrong,” yet become highly effective at reports, proposals, or structured writing later.
One major change is procedural.
Old loop:
generate → judge → dislike → delete everything → restart
New loop:
identify purpose → inspect criteria → list required elements → make a rough version → revise the visible object
Writing processes compete for limited cognitive resources; planning, sentence formulation, and reviewing can all load working memory.
Separating them therefore makes sense.
Using criteria is not “cheating.” It turns an undefined creativity test into a constrained design problem.
Instead of fighting one giant blank-page boss, you split it into requirements, draft, and review.
7. Talking forever can be a retrieval strategy, not just a big memory buffer
People who keep a conversation alive often use the other person’s last sentence as the next retrieval cue.
“People around me love that film, but I did not get it.”
That one sentence can trigger:
popularity → feeling left behind → similar experience → what was confusing? → another example → a personal anecdote.
The speaker is not necessarily carrying a huge prepared topic list.
They are converting incoming speech into new search queries.
That also explains why someone may talk easily but hate writing long messages. The ideas arrive faster than typing, ordering, and editing can keep up.
Conversation skill and tolerance for manual text composition are different variables.
8. AI works best when it separates the stages instead of inventing a personality
A practical AI workflow can be extremely simple.
The human dumps the raw material by voice:
“I want to mention this.” “I already said that part.” “This connects to the other topic.” “I do not know the order.”
AI removes repetition, orders the points, adjusts length, and produces readable sentences.
The division of labor becomes:
human: meaning, observations, judgment AI: holding, ordering, compression, wording options
This resembles cognitive offloading: using an external tool to reduce internal processing demands.[2]
A 2026 systematic review of AI-assisted writing also distinguishes more regulated users, who selectively evaluate and revise AI feedback, from users who accept suggestions with little evaluation.[5]
AI is strongest as extra desk space, not as a replacement mind.
9. Outsourcing everything can separate polished text from actual ability
Cognitive offloading improves performance in many tasks, but research also shows tradeoffs: when information is externalized, immediate performance can improve while internal memory for offloaded material may be weaker.[2]
The same distinction matters with AI.
“I provide the ideas; AI drafts; I inspect and revise” preserves human control.
“Tell me what to think, what to feel, and what to say so people like me” delegates the meaning itself.
Then the text may become impressive while the person represented by it drifts away from reality.
The useful boundary is:
outsource friction, not authorship of meaning.
10. Different bottlenecks need different tools
If you go blank, stop using free-response prompts. Answer three narrower questions:
- What happened?
- How did I feel?
- What would I like next?
If speaking is easy but writing is tedious, speak first and let AI edit without adding new facts.
If a second language is the bottleneck, keep the concept and improve access to frequently reused phrases; fluency is partly a processing-speed and automaticity problem, not merely a knowledge problem.[4]
If you constantly create and delete, separate generation from evaluation. Produce a deliberately rough version first, inspect criteria second, and revise third.
Do not summon the creator and the judge at the same time.
Conclusion
“Words come easily” is not one ability.
Language output contains conceptualization, retrieval, generation, formulation, temporary maintenance, monitoring, and production. Different people can be fast or slow at different points.
That is why both of these are perfectly coherent:
“I know I enjoyed it, but I cannot produce a message.”
“I can talk forever in person, but typing all of that is exhausting.”
AI changes the workflow because the human no longer has to complete every stage internally.
The human can supply meaning. AI can produce a provisional form. The human can detect what feels wrong and revise it.
Someone who once concluded “I am bad at writing” may become very effective once the process is decomposed.
That is not magic. It is a rearrangement of cognitive work.
Sources
- supports the distinction between memory systems, verbal retrieval and divergent/convergent creative cognition
- supports the cognitive-offloading framework and its performance/memory trade-offs
- is used only to support the limited claim that anxiety and self-monitoring can add cognitive load in writing; it is not used to diagnose any individual
- supports the analogy with second-language fluency, processing speed and automaticity
- supports the distinction between selective human evaluation of AI suggestions and minimally evaluated uptake
