The Memory Crisis: How AI Is Changing Our Brains

My grandmother knew her entire extended family's phone numbers by heart. Every birthday. Every address. Today, I can barely recall my own mobile number. This is not a failure of memory. It is an adaptation, a reallocation of cognitive resources away from information storage because external storage became cheap and reliable.

The question is whether all cognitive reallocation is equally benign.

What memory actually is

Memory is not a filing cabinet. It is an active, reconstructive process that shapes how we think, what we notice, how we make connections, and who we are. The memories we consolidate and maintain are not neutral records. They are the substrate of reasoning, the raw material from which we generate new ideas by recombining existing ones.

When we outsource memory to AI, we are not just freeing up storage space. We are changing the composition of our mental substrate. We remember less, which means we have less material for the generative, associative thinking that produces insight.

The attention crisis underneath

Perhaps more serious than the memory shift is the attention shift. Memory requires attention: you cannot remember what you never truly attended to. As AI generates information faster than we can process it, the tendency is to skim, scan, and summarize rather than read deeply. The result is a layer of superficial familiarity with many topics and genuine depth in very few.

Deep thinking, the kind that produces genuinely novel work, requires sustained attention over time. It requires holding a problem in mind long enough for non-obvious connections to emerge. This process is slow, uncomfortable, and entirely incompatible with the attention patterns that AI-accelerated information consumption encourages.

The research we should be watching

We are, collectively, running a large-scale uncontrolled experiment on human cognition. The results will not be clear for years, possibly decades. But early indicators, such as declining performance on certain memory tasks, reduced reading comprehension depth, and increasing difficulty with extended concentration, suggest that the effects are real, even if their magnitude and permanence are not yet established.

What this looks like in practice

It rarely shows up as a dramatic loss. It shows up as a client call where you used to hold the shape of the problem in your head between sessions, and now you open the AI summary to remember where you left off. It shows up as finishing an article and being able to repeat the conclusion but not the argument that got there. None of these moments feel like decline. Each one feels like efficiency. The pattern only becomes visible in hindsight, once the habit of holding things in mind has quietly stopped forming.

Not every outsourcing is a loss

Memorizing phone numbers was never valuable for its own sake. It mattered because it kept people reachable, and once that function moved to a device, nothing of substance was lost. The same is true for turn by turn directions, unit conversions, or the exact wording of a contract clause you can pull up in seconds. Offloading trivial recall frees attention for harder work. The real question is not whether we are remembering less. It is whether we are remembering less of the material that thinking is actually built from: the arguments, the patterns, the half formed connections that only surface because something was sitting in memory long enough to bump into something else.

Why this matters more for the next generation

Adults who lean on AI for memory and attention are reallocating a muscle that already developed. Children doing the same thing may never build it in the first place. Sustained attention and the discomfort of holding an unresolved problem are learned through repetition, usually the hard way, over years of school assignments and boring afternoons. If AI removes the friction before that repetition happens, the concern is not that a skill atrophies. It is that it never forms. This is one reason the conversation about AI in classrooms cannot only be about academic integrity. It has to include what the developing brain is, and is not, practicing.

The muscle you never build cannot atrophy. It is simply absent, and you will not notice its absence until the moment you need it.

Practices that keep the muscle

None of this requires rejecting AI. It requires a few deliberate habits that run against the grain of convenience.

• Write the first draft of a hard email or argument yourself, badly if needed, before opening an AI tool.
• Read one long piece a week without summarizing it afterward. Let it sit unresolved instead.
• After a meeting, recall the three points that mattered before you check your notes or a transcript.
• When a problem is genuinely hard, give it a night before querying AI for the answer. The overnight sitting is where a surprising number of real insights come from.

What to watch in yourself

A few honest signals are worth noticing: reaching for AI to answer something you actually already know, feeling unable to sit with an open question for more than a few minutes without searching for closure, or being able to summarize a topic fluently but struggling to explain it out loud without notes. None of these are catastrophic on their own. Together, and left unchecked, they describe a mind that has quietly stopped doing the work that makes it a mind.

This is not an argument for limiting AI use. It is an argument for intentional AI use, for maintaining deliberate practices that exercise memory, attention, and deep thinking, as a conscious counterweight to the cognitive tendencies that AI-abundant environments naturally encourage.