AI;DR: Nobody Read the Article, But Everyone Has an Opinion

When was the last time you read a long article from top to bottom? Be honest. Somebody drops a link, you paste it into a chatbot, you ask for three bullet points, you move on. That reflex now has a name: AI;DR — AI; Didn’t Read.

From TL;DR to AI;DR

The old internet had TL;DR — Too Long; Didn’t Read. It came with an unwritten courtesy. Either the author wrote their own summary up top, or someone who’d actually slogged through the piece posted a recap for everyone else. Either way, the summary came from a human who had read the thing.

AI;DR breaks that. The summary exists, but nobody read anything. Not the person sharing it, not the person reading the share, not the person quoting it in a thread three hops downstream. The act of reading has been surgically removed from the chain, and the chain keeps working anyway.

Rick Manelius wrote an essay under that exact title, and it lands on the sharpest version of the point: summarization doesn’t compress information, it deletes context. And deleted context is invisible. You never see the hole where it used to be.

What summaries strip out isn’t length — it’s structure

Here’s the common misconception. Most people think of a summary as wringing water out of a sponge. The substance stays, the filler drains away. That’s not what happens.

In good writing, length isn’t waste. The sequence of the argument is the length. The paragraph where the author anticipates your objection and heads it off. The sentence conceding an exception. The three detours taken before arriving at the conclusion. All of that registers as word count.

A summary strips it and keeps the verdict. Which is why summaries make authors sound far more certain than they were. A carefully hedged “this may suggest” in the original becomes a flat “this is” in the recap. The qualifiers evaporate first, because qualifiers are the cheapest thing to cut.

One line from the Hacker News thread on the essay stuck with me: a summary tells you what the author concluded, but not why they got there. And someone who only holds the conclusion can neither defend it nor revise it. They’ve memorized a position, not reached one.

The dangerous part is feeling like you understood

Reading researchers have known this for decades. People who read a summary rate their own comprehension higher than people who read the full text. Their actual comprehension is lower.

The mechanism isn’t mysterious. Summaries are smooth. Fluent prose triggers a sense of “got it” in your brain — psychologists call it the fluency illusion, and it’s remarkably durable. Original text does the opposite. Unfamiliar terms surface. A passage doesn’t parse and you back up. You start arguing with the author in your head. That friction is uncomfortable.

That friction is also where learning happens.

AI summarizers are extraordinarily good at eliminating friction. That’s the product. It’s also the problem. The feeling of understanding is now available at a steep discount relative to understanding itself, and most people cannot tell the two apart from the inside.

The writer–reader contract just got broken on one side

Writing has always run on a quiet contract. The writer spends time getting it right. The reader spends time going through it. Both sides pay in the same currency.

What happens when one side stops paying? Writers notice first. Why agonize over a transition if a model is going to flatten it into a bullet? So the writing adapts. Bullet points. Aggressive subheads. A topic sentence bolted onto the front of every paragraph. Prose stops being written to be read and starts being written to be parsed.

The trap is that this kind of writing really is easier to summarize. So it gets summarized more. So more of it gets written for parsing. The loop closes. Nuance, humor, digression, voice — none of it survives that loop, because those are precisely the things a summarizer cuts first.

The infrastructure is already shipping. Publications posting an AI recap link alongside the article. News apps defaulting to summary-first views. Browsers offering to “catch you up” before you’ve scrolled. Every one of these started as convenience, and every one of them demotes the original from default to effortful opt-in. The moment reading the source becomes the deliberate choice rather than the obvious one, the contract is already dead. Nobody announced it.

So don’t use summaries?

No. Use them. Just sort your inputs first.

Most content genuinely is summary-grade — honestly, most of it doesn’t deserve even that. SEO sludge engineered to rank. A recipe blog with 900 words of childhood memoir before the ingredient list. A news piece that buries one fact under twenty paragraphs of throat-clearing. Compressing that is a public service.

The whole game is discrimination. Reading to extract information and reading to change your mind are different activities. The first survives summarization fine. The second does not. Looking up how an API handles pagination? Take the bullets. But an essay carrying someone’s decade of thinking, the reasoning behind a hard judgment call, an argument you disagree with — those only become yours if you go through them.

Practically, one rule holds up well: treat the summary as a filter, never as the conclusion. Read the recap, decide whether the thing is worth real attention, and if it is, go back and actually read it. The summary becomes a table of contents for your reading, not a replacement for it.

And a corollary. If you’re going to quote something or argue against it, read the original. Debating from a summary means you’re fighting a paraphrase, not a person. The author may never have said the thing you’re attacking — and increasingly, they didn’t.

The bill arrives in a few years

When a time-saving tool takes something from you, the charge rarely posts immediately. It shows up later, as a slow decline: the attention span for a long argument, the ability to organize a complicated case in your own head, the patience to follow someone else’s thinking all the way to the end. None of that disappears in a dramatic way. It just gets a little harder each year, and you attribute it to being busy.

Look at the links you saved today. How many are actually worth reading? And of those, how many will you read start to finish rather than in three bullets? That number is a decent proxy for the range of thoughts you’ll still be able to have in ten years.