AI Is Cutting Your Own Hand: What the Data Shows About Human Skills, Thinking and Trust

Here is the thing people get backwards: it is not AI cutting off its own hand. It is you cutting off yours, and handing AI the knife. AI loses nothing by doing your thinking for you. It is not the one that forgets how to write, how to debug, how to argue a point or how to spot a wrong answer. You are. Every time a person lets AI do the hard part instead of doing it themselves, a human skill quietly weakens — not a machine one.

And there is a second half to the same problem. If everyone starts using AI to write things, and more and more of the internet's traffic is bots reading and clicking instead of people, who is actually left in the room to notice, question or care? That is not just a worried feeling any more. There is real data behind it now, and it is specifically about what happens to human beings — not machines — when they hand over too much of their own thinking. Some of it is genuinely worrying. Some of it is less scary than the headlines make it sound. This piece goes through what has actually been measured, and says plainly where a number is a guess.

Everything below rests on named studies and surveys from 2025 and 2026: a Swiss business school, MIT (a preprint, not yet peer-reviewed), Cloudflare, Stanford with Imperial College London and the Internet Archive, Ahrefs, and a consumer survey by Clear Spark Digital. Figures that circulate online but that I could not trace to a source have been left out, and I say so where that happened. This is general commentary, not advice.

Yes, your brain really does get lazier when AI does the thinking

Start with the clearest evidence.

Researchers at MIT's Media Lab had people write essays while measuring their brain activity with EEG caps. People who used an AI chatbot to help write showed weaker brain activity than people who used a regular search engine, or no tool at all. It is a small, early study and has not been through peer review yet. But it is one of the first times this has actually been measured in the brain rather than guessed at.

A bigger study from SBS Swiss Business School, published in January 2025, looked at 666 people of different ages. Here is what it found.

Students say the same thing about themselves. Survey after survey has a majority of students reporting that they feel AI is weakening their own critical thinking and that they are becoming too dependent on it. The exact percentages vary by survey, so I will not quote one as if it were the number.

Here is the fair part, though. The researchers say this is not automatic. The author of the Swiss study put it plainly: it mostly comes down to how people use the tool, not the tool itself. If AI removes all the struggle from a task, your brain never gets the workout. That is the real mechanism. It is not that AI is harmful — it is that thinking, like a muscle, gets weaker when you stop using it.

Creativity takes a hit too

This is not just about facts and logic. The same line of research on people who lean heavily on chatbots finds less creativity, more dependence, and a weaker ability to judge whether something AI wrote is actually good or true.

That last part is the uncomfortable bit. It is not only that people create less on their own. They also get worse at noticing when AI hands them something wrong, dull or biased — which is exactly the skill you need most when you are using AI.

Now the bigger problem: who is even reading any of this?

This is where the question — who is going to distribute and consume all this content? — gets a genuinely surprising answer.

In June 2026, bots overtook humans on the internet for the first time. Cloudflare, which sees roughly a fifth of all websites, reported that 57.5% of web page traffic was automated and 42.5% was people. Its chief executive said he had expected the crossover at the end of 2027; it came eighteen months early, driven by "agent" bots — the kind that can click buttons, fill in forms and buy things on their own.

The content being posted tells the same story. Researchers from Stanford, Imperial College London and the Internet Archive checked 33 months of new websites in the Wayback Machine, from ChatGPT's launch to May 2025. By the end, 35% of new sites had been made with AI help, and 17.6% were written entirely by AI. A separate Ahrefs check of 900,000 newly published pages found 74% carried at least some AI-written text.

One honest warning. You will see people claiming that "90% of the internet will be AI by the end of the year". That is a forecast, not a measurement, so do not treat it as fact. But the careful, checked numbers — the bot-traffic data, the new-website study — all point the same way: more and more of what gets written online, and of what reads it, is not human.

People trust AI more than they check it

This is the part that answers the worry directly — that whatever AI says simply gets accepted as the truth.

Nearly four in five people had acted on an AI answer. Fewer than one in five always verify one. And the gap is widest exactly where it matters most: daily users, the group the Swiss study found losing critical-thinking skill fastest, are the least likely to check. That is not a coincidence. It is the same habit seen from two sides.

It is not all bad news. Other surveys find real scepticism too — plenty of people still trust a search engine over a chatbot, and many say they go back to the original source after reading an AI summary. The figures for these vary so much between surveys that I will not pin one down. The pattern is fair to state, though: most people are not blindly believing everything AI says. A real chunk of people are. And it is the heavy daily users.

So what should you actually do about this?

The good news buried in all this research is that none of it is automatic. It depends entirely on how you use the tool.

1. Think first, ask AI second.Write your own first draft, or your own answer, before you check it with AI. Your brain does the actual work, and AI catches your blind spots instead of replacing your thinking.
2. Check anything that actually matters.People already do this naturally for big decisions — nobody blindly trusts AI about their health or their money. Extend the same habit to anything you will publish, share or act on professionally.
3. Notice which skills you have stopped using.No more first drafts? No more sitting with a hard problem before asking? That is the exact muscle getting weaker. Practise it on purpose sometimes, the way you would keep any skill sharp.
4. If you create for others, your audience is changing.More than half of web traffic is now bots. Original thinking, real expertise and honest sourcing matter more now, not less — because a flood of AI-written content cannot easily fake them.

What this means for a finance professional

This section is my own reading, not part of any of the studies.

Accountants, analysts and anyone who signs off numbers should read the "check it" finding twice. A tax computation, a valuation or a compliance filing is exactly the kind of "real decision" that 78.5% of people now make with AI in the loop — and exactly the kind where a confident wrong answer costs money. The rule this site follows is the one the research points to: let the tool draft, but make the working visible and check every figure against the section it rests on. That is also why the Tax & Money Lab tools show their arithmetic and name their sources rather than just giving a number.

Bottom line

"AI is cutting your own hand" is not just a catchy line. It is a real, measurable thing, and it is happening to people, not to the AI. It has a name — cognitive offloading — it shows up in brain scans, and it hits some people, young daily users especially, harder than others. At the same time, more of the internet is being written by AI and read by bots than ever before, which means fewer humans on either end to notice when something has gone wrong.

None of this means you should stop using AI. It means you should notice what you are quietly letting it do for you, and make sure you are not the one losing the skills, judgment and independent thinking that made you valuable in the first place. AI is not going to miss the hand. You will.

Frequently Asked Questions

Does using AI really make you worse at thinking? It is your skill at stake, not the machine's. The evidence so far says heavy use is linked with lower critical-thinking scores — a 666-person study found it, and an MIT preprint saw weaker brain activity in chatbot-assisted writers. The researchers stress it depends on how you use the tool: offloading the struggle is what does the damage.

Are bots really more than half of internet traffic? Yes, as of June 2026, per Cloudflare Radar: 57.5% automated, 42.5% human, across the roughly one fifth of websites Cloudflare sees. Its CEO had expected that crossover only at the end of 2027.

How much of the web is written by AI? Measured, not guessed: about 35% of new websites by mid-2025 were made with AI help and 17.6% entirely by AI (Stanford, Imperial College, Internet Archive); 74% of 900,000 new pages carried some AI text (Ahrefs). Claims of "90% of the internet" are forecasts.

How many people check what AI tells them? In an April 2026 survey of 2,127 US adults, 78.5% had made a real decision based on AI and only 17.4% always verify its answers — 14% among daily users.

What is the single most useful habit? Think first, ask AI second: produce your own draft or answer before you check it with the tool, so it corrects you rather than replaces you.

Sources: Gerlich, "AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking", SBS Swiss Business School, January 2025 (666 participants); Kosmyna et al., "Your Brain on ChatGPT", MIT Media Lab, June 2025 (preprint); Cloudflare Radar data published by Matthew Prince, June 2026; "The Impact of AI-Generated Text on the Internet", Stanford, Imperial College London and the Internet Archive, 2026; Ahrefs, April 2025; Clear Spark Digital AI search survey, April 2026. Figures on AI-written content vary with the detector used; treat them as a trend rather than an exact count.