A slide gets about three seconds of reading from a person who is also listening to you talk, which in plain prose is fifteen to twenty words. Everything past that is text competing with your own voice, and your voice loses, because reading is silent and faster. The fix is not trimming a finished slide. It is writing the slide after you know what you are going to say, so that it carries only the part your voice cannot.
How many words should actually be on a slide?
The real limit is a reading-time budget, not a word count. Three seconds buys roughly fifteen to twenty words of ordinary English, fewer if the words are unfamiliar, and fewer still if any of them are numbers, because a figure like 41.7% costs about as much attention as a short sentence. The rules that circulate get the arithmetic close and the reasoning wrong. Six words per bullet, six bullets per slide, treats the slide as a document with a quota. The constraint is not the rectangle. It is a person whose attention you are already spending.
The budget is also measured against you, not against the slide on its own. A slide you leave up for ninety seconds while you work through a hard claim can carry more than one you click past in eight. The same forty words are right in one position and wrong in another, which is why any rule that ignores pacing gets ignored back. Ask how long the slide is on screen with you talking over it, then subtract the talking.
Why is the bullet almost always too long?
A bullet is usually a compressed paragraph. Somebody wrote a full thought, felt the slide sag under it, and squeezed out the connective tissue until it fitted the box: Churn improved due to onboarding changes rolled out in Q2 across the SMB segment. That is not a bullet. It is a sentence with its verbs taken out for parts, and the reader has to put them back before it means anything. Decompressing is reading, and reading is the one thing they cannot spare.
That line had two better forms. As a sentence: onboarding cut SMB churn. As a picture: the churn line with the Q2 release marked on it, which shows the causal claim instead of asserting it and costs almost no reading at all. Slideable ships eight chart types, and the honest reason most decks use two of them is that a bullet is faster to type, not clearer to read.
- It has two verbs. Two verbs is two claims, and two claims is either two slides or one sentence.
- It describes a movement over time. A movement is a shape, and a shape belongs in a chart.
- It only makes sense with the bullet above it. Bullets that depend on each other are a paragraph in a costume.
What is the fastest test for a slide that says too much?
Read the slide aloud, then say the thing you were going to say about it. If the two overlap, one of them is redundant, and it is not you. Cut the slide back to what your sentence cannot carry — the number you must not misquote, the name, the shape of the curve — and let your voice do the rest.
The same test catches a second failure that nobody flags in review: the slide you have nothing to add to. If reading it aloud is your commentary, the slide is doing your job and the meeting could have been the file. Sometimes that is the right answer: send the document. More often it means you have not yet decided what you think, and the slide is standing in for the decision.
Should a headline state the topic or the finding?
The finding, every time. "Q3 revenue" is a filing label: it says which drawer the slide came from, and makes everyone read the chart to find out why they are looking at it. "Revenue grew 12%, and two accounts did it" is a claim. A claim can be agreed with, argued with, or remembered on the way out, which is the whole job of the line.
Finding-first headlines also make the deck legible as an outline. Slides are reordered by dragging them in the rail, and the rail is the outline — so when every headline states a finding, dragging slides around is visibly rearranging an argument, and a step that no longer follows is obvious at a glance. When every headline states a topic, you are shuffling a table of contents and cannot see whether the reasoning survived the move.
Where does everything you cut actually go?
"We cut it for space" is the wrong frame, and it is why cut material creeps back. Space is negotiable: someone senior will always find a bit more, in a smaller size, in the corner. The right sentence is that we cut it because the audience cannot read and listen at once. That version survives the meeting, because it names a fact about people rather than a fact about a rectangle.
Speaker notes are where the cut material goes, and they are not a graveyard. Notes hold the things you will say once, if asked, and would rather not improvise.
- The source and date of every number on the slide, so you can answer where it came from without guessing.
- The objection you expect, with its two-sentence answer, written down before you are being asked in front of people.
- The one phrase you must not paraphrase: a legal wording, a definition, a customer sentence in their own words.
- The second example, which exists only for the person who asks whether the first one was a fluke.
Why is a character budget kinder than a colleague?
Every layout slot in Slideable carries a maxChars value: the length that box was actually drawn for, decided when the layout was designed rather than guessed at while you type. The editor audits a deck against those budgets and reports which slides overflow. That moves "too long" from an opinion somebody holds in a review to a fact the tool reports, with a slide number attached to it.
This is the kinder version of the feedback. "This feels wordy" is unanswerable — you cannot tell how much to remove, or whether removing it will satisfy the person who said it. A budget names a number and leaves the choice of what goes to the person who knows the material. There are 94 layout archetypes, so a slot that will not take the copy is also a prompt to pick a different slot rather than a verdict on the sentence. The MCP server exposes 14 tools, including a deck review that flags the problems, so an agent can read the deck, rewrite the offending slide and check the result — a loop that ends in a measurement instead of in taste.
Which leaves the trick that makes all of the above easy: write the slide last. Draft what you are going to say first — out loud, in a document, on a walk — and only then build the slide as the caption for it. Slides written first become the script, and a script on screen is read faster than you can speak it, so the room reaches your point before you do and then waits, politely, for you to catch up.
A slide written before the talk becomes the talk. A slide written after it becomes a caption, which is all it was ever meant to be.
Common questions
- How many words should be on a presentation slide?
- Between fifteen and twenty words of plain prose, because that is roughly what someone can read in the three seconds they can spare while also listening to a speaker. The real limit is reading time rather than word count, so a slide held on screen for ninety seconds can carry more than one you click past in eight. Numbers and unfamiliar names cost more than ordinary words, so budget fewer of them.
- Should slide headlines state the topic or the conclusion?
- State the conclusion. A topic headline like "Q3 revenue" only says which drawer the slide came from and forces the audience to read the chart to find out why it matters, whereas "Revenue grew 12%, and two accounts did it" is a claim that can be agreed with, argued with, or remembered. Finding-first headlines also let you read the deck outline as an argument rather than as a table of contents.
- What should go in speaker notes instead of on the slide?
- Anything you will say at most once: the source and date behind each number, the objection you expect with its answer already written, wording you must not paraphrase, and the extra example that only comes out if someone asks. The test is whether the audience needs to see it or only to hear it, since they cannot read and listen at the same time. Cutting material for that reason rather than for space is what stops it creeping back onto the slide.