In News That Surprises No One, LLM Usage Leads to Less Brain Activity

Last year, MIT published a paper titled, "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task." It's over 200 pages, but you can read it yourself. (What else have you got going on this weekend?)

The study's findings are not surprising. If you use an LLM like Claude or ChatGPT to write something, it results in less creativity, less originality, less retention, greater potential for bias, and decreased overall brain activity.

And this can happen fast—in this case, in just one writing session.

What Did MIT's Study Find?

In this study, 54 people were split into three groups. One group used an LLM to write essays, another used Google, and a third used their own brains. They wore EEG monitors that tracked their brain activity in real time across four sessions over four months. Each session had a distinct set of three topic prompts, drawn from SAT tests. Researchers were interested in finding "the cognitive cost" of LLM usage. They used OpenAI's GPT-4o.

To measure brain changes and results, constraints were swapped in the fourth session. The 18 participants in this session were given essay choices from previously written topics, so they were familiar with them. Those from the original LLM group were asked to write with no tools (LLM-to-Brain), and the brain-only group members were given an LLM to assist them (Brain-to-LLM).

Researchers concluded:

Brain-to-LLM participants showed higher neural connectivity than LLM Group's sessions 1, 2, 3 ... This suggests that rewriting an essay using AI tools (after prior AI-free writing) engaged more extensive brain network interactions.

In short, brain-only participants still had their brains fire up during the process, even when they used an LLM in the last session. Those who had used an LLM and suddenly had to go without it demonstrated less brain activity.

Overall, the search engine group fell in the middle, but they were not participants in the final session.

As for the content of the writing, the LLM users did provide longer essays with more specifics and details, but many of the resulting essays contained similar content (the study calls them "homogenous"). The brain-only writers produced shorter essays with more errors. Search engine users created essays that fell somewhere in the middle in terms of length, with some similarity in writing (likely due to finding similar search results), while also having higher brain interconnectivity than the LLM users.

After the writing, participants were asked to quote from the work they had just produced. And 83% of LLM users could not accurately quote what they had "written." Or copied and pasted.

According to the research, "These results indicate that reliance on an LLM substantially impaired participants' ability to produce accurate quotes, whereas search‑based and unaided writing approaches yielded comparable and significantly superior quoting accuracy."

When those participants later had their LLM access removed and were asked to quote from their brain-only work, they still failed. "78% failed to quote anything," per the study.

Relying on LLMs too early, without bothering to learn first, "may result in shallow encoding," the study wrote.

MIT researchers landed on the term "cognitive debt" to describe the issue of LLM overreliance, leading to a decrease in brain activity and engagement that results in independent thinking. Which, if you're a writer, is pretty important.

Cognitive debt defers mental effort in the short term but results in long-term costs, such as diminished critical inquiry, increased vulnerability to manipulation, decreased creativity. When participants reproduce suggestions without evaluating their accuracy or relevance, they not only forfeit ownership of the ideas but also risk internalizing shallow or biased perspectives.

What Does This Mean for Screenwriters?

Well, a lot. We hope you aren't using an LLM to create an entire screenplay. Why would you want to outsource the work of coming up with a whole world and the people in it? That's the fun part.

And why would you want to risk homogeneity when a script needs to stand out in a sea of pitches?

How can you expect to revise if you never did the work of setting things up in Act One to build toward a climactic Act Three? You won't know where to start.

How can you expect to perform in a pitch if you can't recall character arcs or important lines of dialogue?

The study doesn't necessarily argue against the wholesale use of LLMs. But it does frame it as a tool within a framework where you, as a writer, have already learned the basics, found your voice, and discovered what excites you. An LLM can't do that for you.