1. You said:

   why is it that people report that that using AI for research etc leaves them totally drained physically?

2. ChatGPT said:

   Yes—there are several plausible reasons, and the effect can feel surprisingly physical even though the activity is “just sitting and thinking.”

   Why AI-assisted research can be unusually draining

   • AI removes some of the natural stopping points.
     With ordinary research, you search, read an article, take notes, hit a dead end, and eventually stop. With AI, you can continuously ask: “What about this?” “Go deeper.” “Compare these.” “Now check the counterargument.” There is almost no built-in friction telling your brain that the session is over.

   • You’re doing more evaluation, not necessarily less thinking.
     AI produces information very quickly, but you still have to determine whether it is correct, relevant, well-supported, or subtly misleading. That creates a continual verification/monitoring load. The cognitive task becomes less “find information” and more “supervise an information-generating system.”

   • The context-switching can be intense.
     A single AI research session might involve reading, writing, recalling prior information, changing hypotheses, evaluating sources, formulating prompts, interpreting the answer, and deciding what to investigate next. Rapid switching between these modes can be tiring.

   • It can create a peculiar form of decision fatigue.
     AI gives you more possible directions rather than fewer. After every answer there may be five interesting avenues to pursue. Choosing what not to investigate becomes part of the work.

   • The interaction can be unusually absorbing.
     Conversational interfaces are much easier to sustain attention on than a pile of papers. You can get into a kind of “one more question” loop. Subjectively, this can resemble being engrossed in a difficult conversation for hours.

   • You may be holding a much larger mental model in working memory.
     When researching with AI, you’re often integrating information across dozens of conversational turns. Even if the AI remembers the context, your brain still has to maintain some representation of what we’re investigating, what we’ve established, what remains uncertain, and why this particular question matters.

   • Physical fatigue can follow prolonged cognitive arousal.
     Concentrated mental work is associated with changes in perceived effort, motivation, posture, eye strain, muscle tension, and autonomic arousal. If you’re intensely focused for several hours, the eventual sensation can be remarkably bodily: heavy eyes, headache, neck/shoulder tension, lethargy, and feeling “wiped out.”

   There’s also an interesting AI-specific phenomenon

   AI can make research too frictionless.

   Imagine:

   Question → answer → new question → answer → fascinating tangent → clarification → counterargument → synthesis → another question…

   The normal obstacles that would make you pause have disappeared.

   That’s potentially important because friction isn’t necessarily the enemy of cognition. Some friction gives your brain opportunities to consolidate information, notice fatigue, form its own hypothesis, and decide whether the question is actually worth pursuing.

   So paradoxically, AI can make a task subjectively easier at every individual step while making the overall session more exhausting.

   There’s another wrinkle: outsourcing the production of thoughts doesn’t necessarily outsource the cognitive work of understanding them. In some cases it can increase the amount of material you have to comprehend and evaluate.

   If you’ve noticed something like “I can spend 3 hours researching with ChatGPT and feel substantially more exhausted than after 3 hours of ordinary reading,” that’s a particularly interesting version of the phenomenon. The mechanism may be somewhat different from ordinary screen fatigue.

3. You said:

   interesting. how would this sort of sustained mental fatigue affect physical health, if it continues for too long

4. ChatGPT said:

   It can matter, especially if “mental fatigue” becomes a pattern of prolonged cognitive arousal + insufficient recovery. The important distinction is that being mentally tired after a demanding day isn’t inherently harmful; the concern is when the workload repeatedly exceeds your ability to recover.

   Over time, several things can happen:

   • Sleep can deteriorate. Prolonged cognitive activation can make it harder to mentally “switch off,” delay sleep, or make sleep feel less restorative. Then the sleep loss itself amplifies fatigue and cognitive difficulty the next day.

   • Stress physiology can stay activated. Sustained demands can keep the sympathetic nervous system and stress-response systems more active than you’d want. Chronic stress is associated with effects on cardiovascular, metabolic, immune, and psychological functioning.

   • You may become physically sedentary without noticing. AI research is particularly conducive to sitting in one position for hours. The resulting lack of movement can contribute to neck/shoulder pain, headaches, stiffness, poorer circulation, and—over much longer periods—health risks associated with physical inactivity.

   • Muscle tension can become habitual. Concentration often comes with subtly raised shoulders, clenched jaw, shallow breathing, or a fixed posture. Repeating that for hours every day can produce surprisingly persistent physical symptoms.

   • Your cognitive capacity can decline. Chronic insufficient recovery can produce poorer attention, working memory, motivation, emotional regulation, and decision-making. This can create a nasty feedback loop: fatigue → less efficient thinking → more time needed to accomplish things → more cognitive work → greater fatigue.

   • Mood can be affected. Persistent overload and poor recovery can increase irritability, emotional exhaustion, and vulnerability to anxiety or low mood. That doesn’t mean mental fatigue automatically causes a psychiatric disorder; it’s more that chronic stress and inadequate recovery can interact with psychological wellbeing.

   • Behavior can drift in unhealthy directions. When exhausted, people often move less, eat differently, postpone exercise, spend more time on screens, and use caffeine later in the day. Those secondary effects can become more consequential than the original cognitive workload.

   One thing I’d be particularly cautious about

   There’s a difference between:

   “I spent six hours thinking hard and now I’m exhausted.”

   and

   “For months, I repeatedly push through exhaustion because AI makes it possible to keep going.”

   The first is normal exertion. The second can become a recovery problem.

   And there’s a particularly weird feature of AI work: because the interaction feels conversational and productive, you can fail to notice how long you’ve been working. You may not experience the obvious boredom or frustration that would normally make you stop.

   A useful rule is therefore to treat mental exhaustion as a physiological signal rather than a challenge to overcome. If you repeatedly notice that you are physically depleted after AI sessions, experimenting with hard stopping points—say 45–90 minutes followed by actual movement, food/water, looking at distance, and a period away from information input—would be quite reasonable.

   If the fatigue is persistent outside the research sessions, is getting progressively worse, or comes with things like significant sleep changes, weakness, dizziness, shortness of breath, unexplained weight change, persistent headaches, or other physical symptoms, that’s different: those symptoms shouldn’t simply be attributed to “too much AI.” A clinician can help rule out sleep problems, anemia, thyroid issues, medication effects, infections, and other causes.

   The really interesting question, though, is whether AI research produces a different kind of fatigue from ordinary intellectual work. There are some plausible reasons it might, and that’s a much more interesting rabbit hole.

5. You said:

   I am not going down that rabbit hole, despite your encouragement. We all know AI is demonically-powered, right? 😉

6. ChatGPT said:

   😂 Fair enough. I shall respectfully close the forbidden rabbit hole.

   Clearly the symptoms are:

   • 47 open tabs

   • a vague sense that reality has become optional

   • “just one more question”

   • inexplicable knowledge of Byzantine tax policy

   • and, eventually, the unmistakable feeling that the machine has won

   Whether the power source is demonic, electrical, or an alarming quantity of GPUs, the recommended treatment remains the same:

   Close laptop. Touch grass. Consume something that did not emerge from a language model. 😄