Opinion | I'm Begging You: Never Write With A.I.
By: Bret Stephens (The New York Times)
An opinion motivated by naivety? Ignorance? Conceit?
S E E D E D C O N T E N T
Aug. 4, 2026 -- This column isn't so much an argument as it is a plea: Don't use artificial intelligence to help you write. Never let A.I. do your writing for you.
Don't use it for school papers, work briefs, letters to your in-laws, speeches at your company gathering or emails (however perfunctory) to your colleagues or friends. Don't let it organize your notes. Don't let it suggest an opening sentence, a segue or a closing paragraph. Don't ask it to write a first draft and pretend that editing that draft somehow makes it your own. It doesn't.
Do none of these things not because they are unethical. Writing with A.I. is unethical when it's a deception: when you pass off words, ideas and information as your own when they aren't. An acknowledgment can largely address the problem. Do none of these things, either, because you might be able to learn to write better than an A.I. can. Pretty soon, if not already, you won't, just as you can't outrun a car or outplay a chess app.
The problem with writing with A.I. is that it's mentally enfeebling — an escalator toward a result when you really need to make a daily habit of taking the stairs. As it becomes ubiquitous, it undermines not only our individual ability to write but also a society's collective ability to reason, a culture's inner capacity to create and everyone's reason to care. We're already reckoning with the well-documented decline of reading; A.I. is accelerating the decline of writing, ushering us further into what The Atlantic's Rose Horowitch calls our "postliterate age."
What, uniquely, does writing do? It compels thought. It compels thinking in ways that silent contemplation or spoken language rarely can. It compels us to subject our thinking to the effort of articulation, the rigor of grammar, the tests of intelligibility and coherence, the inspection of others. In doing so, it also enjoins us to be clear, logical, accurate — and accountable. People may easily forgive a word said in anger but not so easily one written in anger, precisely because the fact that it was written tells us that it was considered.
The considered word matters. It has staying power and moral weight. We use the expression "I want that in writing" not only so there's no mistaking the exact words that have been agreed upon but also as proof that the words are meaningful in the first place.
But what happens when the words aren’t considered because they were written by (or with the help of) a machine?
Words become hollow. Imagine the father of the bride turning to A.I. to help him write his daughter’s wedding toast: What the A.I. can supply in eloquence it subtracts in effort and authenticity. Now imagine that the daughter learns that the speech was written by A.I.: Wouldn’t she be right to be disappointed, if not insulted?
Many would argue there’s a big difference between using A.I. for something as significant and personal as a wedding speech and the boring and frequently exhausting tasks of writing, say, routine office memos. But the intellectual muscle that’s exercised in everyday writing is the same one used for the writing that matters. We become better writers by the constant effort that mundane writing demands. We become better thinkers by investing into that writing a degree of attention we’d never give it if we relinquished the task to A.I. We become more creative when we learn how to put a personal stamp even on routine things, making them unmistakably our own.
These skills are now evaporating. A recent examination using A.I.-detection technology of 100 doctoral dissertations found that 56 percent of them contained “more than negligible” A.I.-generated writing and 19 percent had “more than 50 percent A.I.-generated text.” Anecdotal stories from college professors have a finger-in-the-dike feeling with regard to their efforts to hold back the flood of A.I.-generated student papers. It’s easy to treat the problem as one of mere cheating. But cheating doesn’t capture the real scope of the A.I. threat, which isn’t merely to academic integrity or personal betterment but also to democratic self-governance.
The link between broad-based literacy and the rise of democracy is well known: People who can read and write for themselves have, historically, also been good at choosing and advocating for themselves. What happens when the process goes into reverse? When people lack the patience to absorb arguments more than a few pages or sentences long? When they lack the ability to contradict those arguments because they’ve never honed, through daily practice, the most essential skill for clear thinking?
We’re already getting a taste of what this world looks like in President Trump’s social media posts and the sway they clearly exert over tens of millions of Americans. But that’s a reflection of where the country was a decade ago, in the pre-A.I. age. How much farther down the road are we now, with our even more diminished capacity to read, write and think?
The next time you’re tempted to just adopt an automated response for a text message or an email, don’t. Type out a response that’s your own. Think of it as an act of free will. Or a vote against Trump. What takes you a few additional seconds is an act of resistance against an ever-dumber world.
Bret Stephens is an Opinion columnist for The Times, writing about foreign policy, domestic politics and cultural issues.
Be a good human. Grossly derogatory comments, blatantly inflammatory comments, excessive trolling, mean-spirited taunting and insults, scamming, spamming, and off topic comments may be removed in compliance with the site's Terms of Service and the Code of Conduct. The moderators have the final word.
Tags
Who is online
92 visitors
Does Bret Stephens similarly reject editorial mark up and peer review? Does he condemn ghostwriting? Or script rewrites? Does Stephens abhor quotes, common catchphrases, historical rhetoric, or an interesting comment heard over lunch?
Not all writing is created in an intellectual vacuum - nor should it be. Stephens’ rigid concept of 'intellectual purity' ignores the reality that writing has always been a collaborative, iterative process.
Can artificial intelligence be misused? Of course. Can people be lazy and take shortcuts using AI? Naturally. Will people use AI in unintended and clumsy ways out of ignorance? Undoubtedly. But it's the ignorance that needs addressing. Pretending we can avoid AI out of naivety, ignorance, or conceit won't solve anything. The AI genie is already out of the bottle.
AI is a tool. A lot of people are overreacting. Many times out of ignorance. Other times out of fear of change.
Regardless, my comment is that people need to deal with the change. It is inevitable. Make good use of it rather than cower and complain.
Like any new technology, AI will be used for good and for bad. Human nature always winds up doing that. People need to take a breath and focus on how we can best use this amazing breakthrough and how we can best avoid its misuse.
AI is a plagiarist.
Actually, that is oddly somewhat true at the current state of the technology. Not true in the literal sense, but true in that the results we see from AI are essentially a synthesized version of the vast corpus of human language used to train the AI.
As long as the input can be made to manipulate the response to one the writer wants it is suspect. I think that is one of the reasons people rarely tell you what the exact question was when they tell you what AI had to say on a subject. As far as I am concerned (as of now) AI is just another opinion to support the authors opinion.
When AI results are posted with links, you can always try to refute what it presented.
So if you think the AI results were skewed into falsehood by the prompt, you can illustrate this with a rebuttal.
Merely whining that it was AI illustrates that the AI showed you to be wrong and you have no response other than to complain.
If I wanted to have a discussion with AI then I would. Not interested in taking the time to fact check AI.
Are you suggesting the prompt can not skew the results? Not to mention unless the exact prompt is given there is no way to rebut it even if you wanted to.
There is a difference between questioning the credibility and whining.
This is not about you having a discussion with AI. Another strawman.
I noted that if you think content provided by AI is wrong, then you have the option to verify the accuracy. Links are provided.
Merely claiming it is wrong is simply bullshit.
Question the credibility by showing evidence that it is wrong. If you do nothing like that, it is just whining.
So you either didn't read my comment or just ignored it and restated the same thing.
Either way I stand by my correct comments, both of them
You are not going to fact-check AI so you have no credible foundation to claim it is wrong. You offer nothing but bullshit.
So you either didn't read my comment or just ignored it and restated the same incorrect thing.
Either way I stand by my correct comments, both of them
You complain about and dismiss AI results but refuse to explain why they are wrong. That is whining.
You still don't have what I said and what I do correct.
Good strawman argument.
Incoherent nonsense. You have no argument. No contribution other than platitudes and bullshit.
AI routinely counters your nonsense but instead of putting forth a persuasive rebuttal, you dismiss the source. That is a tacit admission that you cannot defend your position.
You lose when the only thing you have is strawman arguments and making it personal.
You continue to just make shit up. Seemingly posting any allegation that comes to mind.
Provide some value, offer something thoughtful.
Sticking with the strawman i see.
Offer something about the article or at least respond to what I actually said and not a made up comment
Best example of projection I have seen in a long time
But without knowing the prompt, it's difficult to judge whether biases in the response are a result of instructed bias or an inherent flaw in the AI model. We do know that AI models can be instructed to include biases in responses. The individual facts may be indisputable but the conclusions from the assemblage of facts will always be suspect.
I could prompt an AI model to "provide a list of statements and actions made by Kamala Harris during her political career that could be perceived as motivated by nefarious intent" then the model will attempt to provide facts that satisfy that request. If I only provide the AI model's response then my bias in the prompt won't be known. The only thing that the reader will see are indisputable facts.
An even more biased prompt would be "provide a factual summation of Kamal Harris' political career from the viewpoint of a Republican opponent". The AI model will attempt to fulfill that request by provide a list of indisputable facts, too. See why knowing the prompt is as important as the AI response?
A good explanation of what I was saying in comments 4 and 4.1.1.
Not when supporting links are provided. And even without links, one can fact check the claims.
If there is bias skewing the facts then you can illustrate this with a counterpoint argument.
Just as one would do with a human source.
Or a person using Chatgpt as support for their position could always supply the prompt used.
Or a person using Chatgpt as support for their position could always supply the prompt used.
Or a person using Chatgpt as support for their position could always supply the prompt used.
Seems kind of interesting someone would want to hide it. I wonder why.
When claims are stated with supporting links, that is the core information. The claims are either supported or not.
You are making feeble excuses because you dislike the fact that AI fact-checking counters your bullshit.
Too bad.
Why are you so against supplying the prompt when discussing? That is also important information to see how credible the answers are and if the poster is trying to manipulate the answer.
Wrong...............Again. Trying to tell others what they're thinking or doing is not your forte. Maybe you should stop it, it does not make you look good
So sad.
They can't refute the facts, so they're forced to attack the messenger. It's a tactic as old as politics.
I am not against it.
You are just making excuses.
The AI delivers fact-checks with links. If you have a problem with a fact check then do some research and see if the AI is wrong.
You are trolling. You offer no valuable information, just witless, cliche responses.
Then why don't you do it?
Because I am not here to cater to your pointless, stupid demands.
If you do not think an AI response (especially with supporting links) is correct, then rebut it instead of whining. Rebut it as a thoughtful individual would do on a claim by a human author.
I see the problem here. You're assuming there is some thinking going on and that the objective is to make a rational argument for or against the topic when a troll's only objective is to whine which requires no thought.
Uh huh. Sure
Maybe the problem ain't the selected facts but rather the biased opinion masked as a rational conclusion. Throwing around labels like 'troll' is not a defense of a biased opinion. Picking five facts out of ten to state an opinion is not a rational conclusion -- it's still just an opinion. Using an AI model as some sort of authority to state an opinion based on selected facts doesn't mean the opinion cannot be challenged with another opinion.
He's not giving an opinion; he's simply complaining about the delivery vehicle the facts arrived on.
Read again, I gave my opinion way up above
One hundred percent correct
An AI response is more credible than your average human source.
So if one does not think the AI provided a correct response then fact check as you would a human source.
Constantly whining about AI bias, etc. on and on, is just trolling.
A thoughtful response would be to make an argument as to why a particular response is wrong. Put forth a thoughtful, clear, sound argument.
I have to 'fact check' Google Gemini Flash quite regularly. Gemini has a tendency to move goalposts in responding. I frequently have to point out that the response is to something that I did not request. And I have had to refute Gemini responses far too often.
My experience is that AI is more fallible than humans. I've never had any of the models ask for clarification of a prompt.
Well then at least you are trying to fact check. (Taking you at your word.)
The key is that if you doubt something then investigate. Categorical dismissal is a bullshit move.
AI tools in general often over explain. They tend to cover their butts so to speak. That is not a false response, it is just an over explanation. It is sometimes annoying when it explains something that you find to be obvious and did not ask for. But it is not wrong.
I used Google Gemini Flash as a sounding board and editor to seed this article. Here's the session for anyone interested: (NOTE: It's necessary to expand the prompts using the arrow in the lower right corner.)
This is how I normally use Gemini. Even before Gemini was available, I tended to go through 4 or 5 revisions (or more) before publishing seeds to the wild. Gemini has helped with spelling and grammatical errors.
Has any of you ever tried to blow out the sun? That is what it is to deny the FUTURE its means of arriving. Ultimately, you will fail. . . . The 'next' generations must learn to meet the test of time and space as their 'thresholds' open up under them and in front of them.
Being a former High School English teacher I can say I'm glad AI did not exist when I was teaching.