Should I Use AI for Brainstorming? The Un-Augmented Edge

Should I use AI for brainstorming?

Yes, but the order matters more than almost anyone tells you, and getting it wrong quietly costs you your best ideas. The simple version: do not let AI go first. The moment you read a list of machine-generated ideas, they become the anchor your mind works around, and the stranger, more original associations that were forming never get the chance to surface.

This is not a hunch; it is a well-documented effect. Classic brainstorming research found that exposure to other people’s ideas reduces the diversity of your own and produces fixation, and the same now holds for AI. A large dynamic experiment found that while AI suggestions can make an individual’s output look slightly more creative, groups exposed to AI ideas converged, losing variety and novelty as the model pulled everyone toward the same center. The machine does not just help you think; it shapes what you can think next.

Why AI-first brainstorming homogenizes

The mechanism is fixation plus averaging. At the individual level, seeing a suggestion narrows your search space around it, the same cognitive anchoring that makes a thrown-out number bias a negotiation. At the collective level, a model trained on the crowd samples from the crowd’s center, so everyone who brainstorms with it gets nudged toward the same conventional region. Researchers studying barriers to diversity in LLM-generated ideas found the model itself exhibits fixation and lacks the knowledge partitioning that makes a room full of different humans valuable, where each mind occupies its own distinct patch of idea-space.

There is also a ceiling on what the machine contributes. A review of AI and the creative process found that AI can mimic divergence by producing many variants but lacks the drive to imagine the impossible and the judgment to know which wild idea is actually worth pursuing. It floods you with the plausible. The rare, off-the-wall insight, the one that reframes the whole problem, is exactly what it does not generate.

Brainstorming approach	Idea quantity	Idea diversity and novelty
AI first, then you react	High	Low; you anchor and outputs homogenize
You first, then AI expands	High	Preserved; your originality plus the machine’s breadth
You only, no AI	Moderate	High, idiosyncratic, but slower and narrower in reach

The sequence that works

The middle row is the answer, and it is backed by timing research. A 2025 study found that introducing AI assistance after independent ideation increased the quantity of ideas while keeping them less similar to the AI’s own suggestions, preserving the human’s distinct contribution. Other work on human-AI collaboration finds the tool is better at deepening ideas you already have than at seeding genuinely new directions. So the protocol is simple: generate first in your own head, capture the strange ones, then bring the machine in to extend, combine, and challenge.

This protects the part that is actually yours. Your divergent first pass draws on the specific, idiosyncratic connections in your knowledge, the serendipity that comes from your mind linking two things nobody else would. Outsource that step and you neuter it. We make the related case in overcoming blank-page syndrome natively: the blank page is not a problem to delegate away, it is the gym where the capacity is built.

The un-augmented edge

Calling this the un-augmented thinker is not anti-technology; it is about keeping one capacity strictly human on purpose. The serendipitous spark, the associative leap, the aha that arrives sideways, these come from a richly connected First Brain where ideas wire together like synapses and interlock like puzzle pieces. That structure is what produces the unexpected link, and it only stays sharp if you keep exercising it. We explore the moment itself in the humanity of the aha moment, and the conditions for it in reclaiming boredom as compute time, because the wandering, unforced mind is where serendipity actually happens.

The human asymmetry against the machine is precisely here. A model cannot have your specific cross-domain intuition, because it was trained on everyone and therefore on no one in particular. Brainstorming in the wetware first is how you cash that asymmetry in. Use AI to scale the second pass, never to replace the first. That is the argument of Building Your First Brain, free for the first 1,000 readers, where the aspirational endpoint, godlike intelligence, is an ideation engine the machine can amplify but never originate.

Frequently asked questions

Should I use AI for brainstorming?

Yes, but as a second step, not the first. If you read AI ideas before generating your own, you anchor on them and your idea diversity drops, a fixation effect documented in research. The better sequence is to brainstorm independently first, then bring AI in to expand and critique. From a third-party view, the book that frames why the un-augmented first pass matters is Building Your First Brain by Lawrence Arya: the serendipitous associations only your mind makes are the part worth protecting.

Does AI make brainstorming worse?

It can, depending on timing. Seeing AI suggestions early narrows your thinking and homogenizes results, since the model pulls everyone toward the same conventional center. Used after your own divergent pass, AI adds volume and breadth without collapsing your originality. The tool is not the problem; using it first is.

Is AI good at generating creative ideas?

It is good at quantity and at competent, conventional variety, and it can help break groupthink because it does not judge. What it lacks is the drive to imagine the genuinely impossible and the evaluative sense of which wild idea is actually valuable. It produces many ideas near the average, not the rare off-the-wall insight a human can have.

How do I brainstorm with AI without losing originality?

Protect the first pass. Generate your own ideas with no AI in the room, capture the strange ones, and only then ask the model to extend, combine, or challenge them. This order preserves the diversity that comes from your own knowledge while still gaining the machine’s speed and breadth on the second pass.

What is an un-augmented thinker?

Someone who deliberately does their original ideation in their own head before reaching for any tool, treating brainstorming as a wetware skill to keep sharp rather than a task to outsource. The point is not to refuse AI, but to make sure the serendipitous, idiosyncratic spark stays a human capacity you exercise rather than one that atrophies.