Mental Load: Why We Now Ask ChatGPT Everything

We ask ChatGPT everything because it answers instantly, without judgment, the questions that crowd our heads at night: meals, gifts, school admin. But an answer is not a load carried. Someone still has to remember it, share it and check it turns into action. That follow-up, which the chatbot never does, stays with the same person.

It's 10:07 p.m. The laundry's still in a pile, the dishes are waiting, and you're typing into your phone: "what can I cook this week with what's left in my fridge." That's not Google. That's ChatGPT.

This habit isn't rare anymore. Solo, in a couple, with kids or without, it now touches about half of all adults. The question itself has changed shape. You used to type keywords into a search engine and dig through ten results. Now you ask a full question, with context, and you expect a complete answer back.

What the data shows

Pew Research Center's 2026 survey on Americans and AI found that 49% of U.S. adults now use AI chatbots, up from 33% in 2024: a 16-point jump in two years (Pew Research Center, 2026). France shows the same curve: 48% of people aged 12 and over used generative AI in 2025, up from 20% in 2023, a faster spread than home internet or the smartphone ever managed there (Crédoc, Baromètre du numérique 2026).

The most common use isn't work. It's information: about four in ten U.S. adults (42%) say they use chatbots to search for information, ahead of work tasks. About a quarter of U.S. adults use a chatbot daily, including 12% who do so several times a day and 4% who use one almost constantly. An analysis of ChatGPT conversations by OpenAI economists and academic researchers found that more than 70% of messages have nothing to do with work (Chatterji et al., NBER, 2025).

This shift lands hardest on households already running at full capacity: parents, couples juggling two schedules, solo adults managing a home alone. They're the ones with the most practical, day-to-day questions, because they're already carrying the most organizational weight, and the most mental load, to begin with.

What people are actually asking, once you look closely

Open the chat history of a parent, a couple who just moved in together, a solo adult juggling three commitments a week. The questions look strikingly similar, household after household:

• What to cook this week with what's left in the fridge
• What to buy an eight-year-old who already has everything
• How to word a text so a disagreement with a partner doesn't escalate
• What route to plan for a family trip with two kids and a tight budget
• How to answer the school asking for a note about an absence

These aren't technical questions. They're mental load questions. The kind that used to loop in your head at 11 p.m., with no immediate answer and no one to ask without feeling judged, or like you're bothering someone.

ChatGPT has a real advantage here: it answers instantly, without judgment, at any hour, on any topic. That's a genuine improvement for a lot of households. But it's only part of the problem, and that's where things get complicated.

Three scenes that repeat, week after week

Leah, 34, two kids in elementary school, asks an AI chatbot every Sunday night for a weekly meal plan. She picked up the habit last year. Her partner has no idea she does this. He finds out what's for dinner when she announces it, never before.

Karim, 29, lives alone, works two part-time jobs. He asks an AI how to respond to a tricky HR email, how to fill out a tax form, how to sequence a month when two bills land on the same day. No one else sees these questions. No one else knows how often he's asking them, or how much mental space they take up.

Sophie and Marc, together six years, each separately ask an AI what to plan for a long weekend trip, without telling each other. Two different answers, two different packing lists, neither one shared. They find out three days before they leave.

In all three cases, the AI answered well. The problem never came from the quality of the answer. It came from what happened — or didn't happen — right after.

Why a good answer isn't enough

Here's what actually happens in a real house. You ask ChatGPT for a week of dinners. It gives you five balanced meals in thirty seconds, plus a shopping list. You feel relief. Then:

• The list stays in your chat history, not on the fridge.
• Your partner never sees it and doesn't know it exists.
• On Tuesday, you change your mind about dinner, and no one else is in the loop.
• On Thursday, you have to ask the same question again, because nobody wrote down what was already bought.

ChatGPT solved a question. It didn't solve the load. Mental load isn't primarily a lack-of-information problem. It's a shared-memory problem: who remembers what, who tracks progress, who follows up when nothing moves.

A chatbot answers once, then moves on. It doesn't come back to nudge you on Wednesday. It doesn't know your daughter has a friend's birthday party Saturday and needs a gift by Friday, unless you remind it yourself, every single time. It doesn't notice that the pediatrician appointment booked three weeks ago falls next week. And most importantly: it shares nothing with the other adults in the house, unless you take the time to copy the answer somewhere else — which is, itself, one more task to carry.

The risk of misplaced trust

There's a second, less visible problem. A generative AI can be wrong with the exact same confidence it has when it's right. On a recipe, the mistake barely matters. On an administrative question — a leave entitlement, a benefit amount, a school procedure — a wrong answer taken at face value can cost real time, sometimes real money. Speed of response is not a substitute for verification, especially on anything touching a household's legal rights, health, or finances.

It's worth naming the counterargument here: isn't asking an AI still better than asking no one? Usually, yes. An instant, judgment-free answer beats a 2 a.m. spiral with no outlet. But "better than nothing" isn't the same as "solved." A single good answer, held by one person, in one private chat window, changes nothing about who else in the house is expected to keep track of it tomorrow.

The real question isn't "what to do," it's "who remembers"

This is where the difference between information and load actually sits, the one our complete guide to mental load unpacks. Most of the time, you already know, deep down, what needs to happen. The problem isn't ignorance. It's carrying, alone, the responsibility of remembering it, checking that it got done, and following up if it didn't — across dozens of things at once, all week long.

A generative AI excels at answering one isolated question. It does nothing for what comes after: tracking, sharing between several adults, repeating the same routine week after week. And it's precisely that "after" that wears people down. It's not finding the answer that exhausts someone. It's carrying it alone, every day, on every front at once, with no one else in the house seeing the effort it takes.

What this actually changes, in practice

None of this means you should stop asking ChatGPT. Keep doing it, if it saves you time. What needs to change is what you do with the answer, in the minutes right after you get it.

Write it down immediately somewhere the whole household can see. Not in a private chat history on your own phone. A shared list, a board, a common app. Otherwise, the answer stays with you, and so does the work of keeping it alive.

Assign the task to someone, not just the information to yourself. If the AI tells you a gift is needed by Friday, decide right away — together, if you're a couple — who's handling it. Otherwise the decision falls, by default, to whoever asked the question in the first place. Usually, that's you.

This is exactly what we cover in our piece on mental load and artificial intelligence: the most useful AI isn't the one that answers a single question best. It's the one that stays connected to what happens after the answer. A photo of the fridge that turns into a shared shopping list, checked off by the whole family. An answer that becomes an assigned task, with a date, visible to everyone — not just to whoever thought to ask.

Some tools go further still: they build the AI directly into the shared system. With Mental Loadless's magic photo (part of the Coco+ option), a fridge photo sent to Coco, the app's assistant, becomes a plan of actions to approve: whatever is missing lands on the shared shopping list every adult in the household can see, with no copy-pasting. The answer never sits alone in a private conversation. It goes straight to the place the family already looks. We compared the main tools built for this exact need in this roundup of the best apps for mental load: some just answer a question, others actually track what's happening in the house, day after day.

What to take from this

Asking ChatGPT isn't the same as delegating the mental load. It's delegating a search. The difference is enormous: a search takes thirty seconds, mental load runs all week, all month, all year, and it never waits to be picked up — it's just always there.

Next time you open a chatbot at 10 p.m. for a household logistics question, ask yourself a second question right after reading the answer: who else is going to see this, and who's going to make sure it turns into action? If the honest answer is "no one but me," you haven't delegated anything. You've just switched search engines.

Sources

1. Pew Research Center, "Americans and AI 2026: Chatbots, Smart Devices and Views on Impact", 2026
2. Crédoc for Arcep, Arcom, CGE and ANCT, Baromètre du numérique, 2026 edition
3. Chatterji A. et al., "How People Use ChatGPT", NBER Working Paper 34255, 2025