Show, Don't Tell Is Bad Advice

A draft lands in my inbox. The writer has clearly been told the rule, and has obeyed it with real discipline:

She set the mug down. Her fingers stayed curled around the handle for a moment after it touched the table. She looked toward the window. Outside, a car alarm started and stopped. She pulled her sleeve down over her thumb.

Five sentences. Zero abstraction. Pure showing, exactly as instructed. And I have no idea what is happening, who this person is, or why I should keep reading. The writer followed the most repeated instruction in all of writing advice and produced something nobody can read.

I have been editing drafts like this for a while now — my own first, then other people's, then thousands more while building a tool that scores drafts on specificity. The paragraph above is not a rare failure. It is what happens when a good instinct gets compressed into a slogan and the slogan gets taught as a law.

"Show, don't tell" is not wrong. It is one rung of a ladder, handed to you as if it were the whole ladder.

The rule isn't what you think it is

Nearly every version of this advice arrives attached to Chekhov: don't tell me the moon is shining, show me the glint of light on broken glass. It is a wonderful line. Chekhov didn't write it.

The sentence is a twentieth-century distillation that was eventually re-attributed to the man it was distilled from — the researchers at Quote Investigator traced the chain. What Chekhov actually wrote, in a letter to his brother Alexander in May 1886, was advice about selection: pick a few small details and arrange them so that a reader who closes their eyes assembles the scene. The example he reached for involved a fragment of bottle glass catching light on a mill dam.

Notice what the slogan dropped. Chekhov's advice is about grouping and arrangement — which details, in what order, so the picture forms. The slogan turned a craft problem into a prohibition.

The formal distinction has a similar history. It was popularized by the critic Percy Lubbock in The Craft of Fiction (1921), where the pair isn't "showing versus telling" but picture versus drama — two modes a novelist moves between. A question of proportion. Not a prohibition on one of them.

A slogan that survives a century gets compressed a little every time it is repeated. What got squeezed out of this one is the word when.

The ladder, and the failure at both ends

The most useful map I know for this comes from outside writing advice entirely. In Language in Thought and Action(1939), the semanticist S. I. Hayakawa drew what he called the abstraction ladder. At the bottom: a particular cow, Bessie, standing in a particular field. Climb a rung and she becomes "cow." Another: "livestock." Another: "farm asset." At the top: "wealth." Each rung is true. Each rung has thrown away detail the rung below it had.

Hayakawa's warning is the part writing advice never quotes. He called the failure dead-level abstracting — getting stuck on a single rung and staying there. And crucially, he described it as a failure in both directions. Stuck at the top, you produce vague generality. Stuck at the bottom, you produce endless detail that never adds up to a point.

Writing advice warns you about one of those. Only one. So we have a generation of writers who can spot this:

The team's communication challenges impacted delivery outcomes across several initiatives.

— and no vocabulary at all for diagnosing the mug paragraph, which is the identical disease at the other end of the ladder. Both leave the reader doing work the writer refused to do. One withholds the evidence. The other withholds the point.

A rule that actually works

Here is the version I use, and the one I would teach instead:

Every abstraction owes a concrete rung below it. Every pile of detail owes an abstraction above it.

Telling is a claim. Showing is the evidence. State it that way and the whole thing becomes obvious:

• A claim with no evidence is an assertion the reader has to take on faith. Most readers won't.
• Evidence with no claim is a pile the reader has to sort. Most readers won't do that either.

Good prose pays both debts, usually within a few sentences of incurring them. Watch what one sentence does to the mug paragraph:

She had decided, somewhere between the kitchen and the table, that she was going to lie to him. She set the mug down. Her fingers stayed curled around the handle a moment after it touched the wood.

I added a claim and cut two details, and the remaining details became legible — the curled fingers now mean something, because there is something for them to be evidence of. That is the trade. The claim doesn't weaken the showing. It is what makes the showing land.

When telling is simply correct

The slogan's worst damage is the writers it has made afraid to state anything. Four cases where telling is the right call, and dramatizing would be a mistake:

1. Pace

Not every stretch of time deserves scene-level attention. "The next three years were the worst of her life" can be better than three years of scenes — because it spends four seconds on something that earns four seconds, and leaves the budget for what matters. Compression is a decision, not a failure.

2. Scale

Some claims are too large to dramatize honestly. A system, a trend, a century. You can open with a single family to make it concrete — but at some point you have to say the thing the single family is evidence for, or you have written an anecdote and implied a conclusion you never defended.

3. Orientation

Readers need to know where they are standing before detail means anything. A detail delivered to a disoriented reader is noise. This is the same working-memory cost I wrote about in why correct sentences can still be hard to read — unplaced specifics are expensive to hold.

4. Claims the reader already grants

You do not need evidence for what nobody disputes. Showing me that the office was busy, in four sensory sentences, when I was going to believe "the office was busy" for free, is a bill you sent me for nothing.

That last one generalizes into the actual selector, and it is not an aesthetic judgment at all:

Tell what the reader will grant you. Show what they would doubt.

Specificity is not a virtue you sprinkle evenly. It is a currency you spend where resistance is. This is why two drafts with identical detail density can read completely differently: one spent its specifics on the contested claims, the other spent them on the mug.

Why AI drafts fail this one so reliably

A language model lives near the top of the ladder, and for a structural reason. High-rung sentences are cheap to generate: there are a million ways to say "the initiative improved outcomes," and all of them are available to anyone. A specific is expensive, because specifics come from having been somewhere. A number, a name, a Tuesday — these are the residue of experience, and a model has no experience to leave residue.

This is the mechanism underneath the fifth tell in the anatomy of AI voice: the absence of specifics isn't a stylistic quirk, it is dead-level abstracting at the top rung, produced by a system that has nothing to descend to.

But notice the trap in the obvious fix. "Add sensory detail" produces the mug paragraph — dead-level abstracting at the bottom, which reads as human but says nothing. Writers who over-correct against AI voice land here constantly. The fix was never "more detail." It was paying the debt in whichever direction it is owed.

The altitude audit

Fifteen minutes, on a draft you have already finished. No AI, and no rewriting until the marking pass is done — the point is to see the shape before you have opinions about it.

1. Mark every sentence H or L. H if it is a claim, summary, or generalization. L if it is a concrete particular — a number, a name, a thing you could photograph. Mark in the margin, fast, no deliberating. Sentences that do both get both letters.
2. Read only the letters. Ignore the prose entirely and look at the string: H H H L H H H H. Any run of five or more identical letters is a dead-level stretch. You have just found your problem paragraphs without reading a word.
3. Pay the H debts. For each unpaired H, either add one L beneath it — a number, a name, a place, a moment in time — or cut the H. Cutting wins more often than writers expect.
4. Pay the L debts. For each long L run, add the sentence that says what it all means. If you cannot write that sentence, that is the finding: you do not yet know what the passage is for.
5. Run the resistance pass. Go back to every remaining H and ask: would a skeptical reader grant me this for free? If yes, delete the evidence under it — you are paying for something nobody charged you. If no, that is where the specifics belong.

Step 5 is the one that changes how people write. Steps 1 through 4 improve a draft. Step 5 teaches you where specificity is worth anything, which is the thing you actually want to internalize.

Where this doesn't apply

Some honest limits, because a rule presented without its edges is how we ended up with "show, don't tell" in the first place.

Plenty of good writing withholds the claim deliberately. Poetry does it constantly; so does a strand of literary fiction where the reader's act of supplying the abstraction is the experience. That works when the reader has been trained by the form to do it — which is a real contract, and a different one from the contract an essay or a cover letter signs. If you are writing in a form where the reader expects to do that work, the debt is transferred, not cancelled.

The H/L marking is also deliberately crude. Many sentences operate at two rungs at once, and a careful taxonomy would have five levels rather than two. I use two because a crude pass you actually run beats a precise one you don't.

And to be clear about what this is: an account of what I have found useful editing my own drafts and designing feedback for other people's. It is a craft heuristic with a reasonable theory behind it, not a controlled finding. Treat it the way you would any editor's rule of thumb — test it on your own pages and keep it only if it earns its place.

The point

"Show, don't tell" survived a hundred years because the failure it names is real and common. It just named half of one. Writing well is not a preference for the bottom of the ladder. It is the ability to know which rung a given sentence belongs on, and to move deliberately between them — paying each debt where the reader would otherwise be left holding it.

That ability is also the reason WriteFit scores specificity separately from clarity, personal voice, and AI-likeness, and why one of its dimensions is reader resistance — because a specific that lands on an uncontested claim is wasted, and no single "quality" score can tell you that. (WriteFit is the tool I build; this article is as much an argument for how I think it should work as it is a pitch.)

If you want to see your own ladder, run one session, no account needed. Bring a draft you think is vivid. Those are the ones where the mug paragraph is usually hiding.