Closing pattern in writing

By Jack Stovell · published 2026-09-24

Closing pattern is how a piece typically ends, recorded in a ScriptGrain voice profile as a short description of up to 15 words rather than a number. It sits in the Rhetoric group. The extraction model judges it from your samples, and it never moves the voice match score.

Analyse your closing pattern free

Your free voice profile measures closing pattern among all 45 attributes, from a sample of your own writing. No card needed.

Try the six-signal check first

At a glance

What closing pattern measures

Closing pattern measures the shape of your last few sentences, not their content. Some writers end on a call to act. Some reframe the point they just made. Some ask a question and leave it hanging. Some just stop, flat, the moment the argument's finished. The attribute captures which of those habits you actually default to, written up as a short description (max 15 words) rather than scored on a scale.

It lives in the Rhetoric group alongside opening style, the other bookend, and argument structure, which makes sense: how you end a piece is really an extension of how you argue through it. A writer who builds conclusion-first often closes on the sharpest version of their point.

How closing pattern is scored

Closing pattern does not move your voice match score. The voice-match engine has no comparator for this attribute, so it never enters the calculation and polish cannot target it. It's recorded and sent with every draft ScriptGrain writes, but it doesn't feed the maths.

That matters if you're chasing a higher voice match percentage. Twenty-eight of the 45 attributes in a profile can shift that score; closing pattern isn't one of them. It's descriptive, not comparative. If you want the full mechanics of how scoring actually works, the voice measurement framework covers it properly.

How to analyse closing pattern in your own writing

You can do this by hand in ten minutes with five or six of your own pieces open in front of you.

  1. Pull the last two to three sentences from each piece, nothing more.
  2. Read them without the rest of the piece attached. Do they stand alone, or do they need the build-up to land?
  3. Sort each ending into a rough type: call to act, reframe, question, flat stop, summary.
  4. Count which type shows up most. That's your default, not your only mode.
  5. Check whether your endings are the most concrete sentence in the piece, or the vaguest.
  6. Ask yourself the calibration question a ScriptGrain profile also asks: how do your pieces usually end? Write the honest one-line answer, not the aspirational one.

ScriptGrain's free voice profile reads closing pattern from a writing sample as one of the 45 attributes it extracts.

Examples of closing pattern in real writing

Example: "So that's the whole argument. Fair enough if you don't buy it, but the numbers don't lie." That's a flat stop dressed up as a shrug, conclusion already delivered earlier in the piece.

Example: "What would you do differently on Monday morning?" That's a closing question, no answer supplied, the reader left holding it.

Example: "Book the call. Or don't. But stop reading and start deciding." That's a call to act, three fragments doing the work three clauses would've done more slowly.

None of these are "typical" in any measurable sense; there's no corpus to compare them against. They're just three shapes an ending can take.

How to make AI end the way you do

Getting an AI assistant to end like you do takes description, not vibes.

  1. Decide which shape you actually use: a call to act, a reframe, a question, a flat stop or something of your own. Pick the honest one, not the one that sounds most writerly.
  2. Give the assistant a real example of your own closing sentence, quoted exactly, so it has something concrete to match rather than a category label.
  3. On a ScriptGrain profile, answer the calibration question "How do your pieces usually end?" directly; that answer feeds Pass 2 when the profile is re-extracted.
  4. Tell it explicitly to stop once the last point is made. Don't leave "wrap it up" as an instruction; it invites a bolted-on summary.

When ScriptGrain writes from your profile, closing pattern is sent inside the voice profile JSON block of the system prompt, under an instruction to treat every attribute as a hard constraint, not a suggestion. That profile reaches ChatGPT, Claude and other tools through the API and the MCP server, so the same closing habit travels wherever you're writing.

How to stop AI writing generic conclusions

Stopping the generic AI conclusion starts with banning the phrase that causes it.

  1. Ban "In conclusion" outright, along with any summary paragraph that just restates what the piece already said.
  2. Cut hedged, abstract last lines: phrases like "perhaps enough after all" or "something approaching" a point. They're padding wearing a suit.
  3. End on the most concrete sentence in the piece, not the vaguest. If your sharpest, most specific claim is in paragraph four, your ending is in the wrong place.
  4. Reread the last line alone. If it could sit at the end of any piece on any topic, it's generic. Replace it with one that couldn't.

That's the whole fix: specificity at the very end, nothing summarised, nothing hedged. ScriptGrain's own drafts ban "In conclusion" and hedged abstract closes such as "its own form of grace" unless they appear in your profile's own lists.

Questions

How do I end a piece of writing?

Pick a shape and commit: a call to act, a reframe of the point, a direct question, or a flat stop the moment your argument's done. A common mistake is drifting into summary by default, restating what was already said. Look at your last two sentences in isolation. If they need the rest of the piece to make sense, tighten them until they don't.

How is closing pattern measured?

It isn't measured on a numeric scale. A ScriptGrain profile stores it as a short free-text description, judged by the extraction model (Claude Sonnet 5) across two passes, and informed by your answer to "How do your pieces usually end?" on re-extraction. There's no reference corpus and no code count behind it; it's a judged reading of your habits, not a tally.

How do I make AI stop writing "in conclusion"?

Ban the phrase directly in your instructions, along with any paragraph that restates the piece rather than adding to it. Give the model a real closing sentence of yours as an example instead of a description. When ScriptGrain writes, "In conclusion" is banned unless it appears in your profile's own lists, and your closing pattern goes into the prompt as a hard constraint.

How do I make ChatGPT write better endings?

Tell it exactly how you end things: reframe, question, call to act or flat stop, with an example quoted in full. Vague instructions like "make it punchy" don't transfer; concrete examples do. A ScriptGrain profile does this by sending your actual closing pattern as a stored attribute, reaching ChatGPT through the API or MCP server rather than a one-off prompt.

How do I stop AI summarising at the end?

Explicitly forbid the summary paragraph and tell the model to stop the moment the last point lands. A request to "wrap up" often reads to a model as an instruction to repeat itself. Replace that instruction with a concrete ending type from your own writing, and end on your most specific sentence rather than your vaguest one.

Related

Sources