Analogies in writing

By Jack Stovell · published 2026-09-24

Analogy preference is a yes or no reading of whether a writer reaches for analogies as a go-to way to explain things. ScriptGrain stores it as analogy_preference; no code counts it. The extraction model judges it from your samples, Claude Haiku judges a draft the same way, and it moves the voice match score at weight 0.5.

Analyse your analogy preference free

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

Try the six-signal check first

At a glance

What analogy preference measures

Analogy preference measures one thing: does this writer explain hard ideas by comparing them to something the reader already knows? That's it. No sliding scale, no "moderate use of comparison", just yes or no. It sits in the content patterns group alongside specificity level, anecdote usage and claim density, and the app labels it plainly as "Analogy Preference". The API key is analogy_preference, boolean field type.

A yes/no attribute forces a clean read. A writer either treats analogy as a default tool for tricky explanations, or they don't. Someone who writes "hiring the wrong developer is like handing your car keys to someone who's only driven a go-kart" gets a yes. Someone who just states the risk directly gets a no. There's no "sometimes" option: the extraction model forms a judgement about habit rather than counting instances.

How analogy preference is scored

Analogy preference is scored, but it's a small piece of the puzzle. It carries a weight of 0.5 in the voice match score, and it's judged yes/no from the draft by Claude Haiku. Same answer as the profile: full marks for that attribute. Different answer: 0.2. So getting it wrong doesn't wreck the score, but it doesn't help either.

Worth being precise here, because it's easy to assume attributes like this get tallied up somewhere. They don't. There's no analogy-counter running over your draft. Claude Haiku reads the text, decides whether analogy is a go-to device in it, and compares that call against what the profile says about you. Match, and the attribute contributes cleanly. Miss, and it drags a little. Twenty-eight of the 45 attributes in a voice profile can move the voice match score at all; this is one of them, just not a heavily weighted one. For the mechanics of how the full score gets built (code-counted features plus Haiku's judgement plus a fixed calibration), see the voice measurement framework.

How to analyse analogy preference in your own writing

You can work this out by hand in a few minutes, no tool required.

  1. Pull five or six pieces you've written where you explained something moderately complex.
  2. Read each one and mark every spot where you compared the thing you're explaining to something unrelated but familiar (a process to a journey, a system to a machine, a risk to a bet).
  3. Count how often that happens relative to how often you just explained the thing directly, with no comparison at all.
  4. Ask whether the comparisons were doing real explanatory work, or just decorating a sentence that would have worked fine on its own.
  5. If comparison shows up as a repeated habit across most of the samples, you're a yes. If it shows up rarely or only for effect, you're a no.

That's the whole exercise. ScriptGrain's free voice profile reads this attribute, along with all 44 others, from a writing sample you upload.

Examples of analogy preference in real writing

A quick set of examples, each one doing a different job.

Example: "Onboarding a new hire without documentation is like handing someone a car with no dashboard: they can drive it, eventually, but they'll stall a lot first." That's analogy carrying the explanation.

Example: "Our pricing model is tiered, with three levels based on usage volume." No comparison at all, just the fact. That's a no-analogy explanation of a similar kind of idea.

Example: "Debugging this code felt like untangling headphone wire in the dark." Here the comparison conveys a feeling rather than explaining a system, so on its own it says less about analogy as an explanatory habit.

None of these are typical values; there's no reference corpus for this attribute, so there's nothing to benchmark against. They're just illustrations of the yes/no line.

How to make AI use analogies

If your AI writing feels flat and analogy-free when you want the opposite, try this:

  1. Ask for one analogy specifically where a concept is hard to grasp, and specify that it should be drawn from your reader's world (a marketer's world, a builder's world, whatever fits).
  2. Show the model two or three analogies you've used before, so it has a pattern to match rather than guessing at your taste.
  3. Ask it to keep the analogy short and return to the literal point straight after; don't let the comparison run on and become the explanation itself.

A ScriptGrain voice profile handles this differently: the whole profile gets sent in the voice profile JSON block of the system prompt, under an instruction to treat every attribute as a hard constraint. There's no separate rule bolted on for analogy; it's one line in that JSON, treated the same as the other 44. That profile reaches ChatGPT, Claude and other tools through the API and the MCP server.

How to make AI stop using analogies

If the model keeps reaching for comparisons you didn't ask for, the fix is mostly subtraction.

  1. Ask directly for explanation with no analogies, stated plainly as a constraint rather than a preference.
  2. Go through the draft and remove any comparison that needs its own explanation to make sense, since those are usually the weakest ones anyway.
  3. Where an analogy was standing in for evidence, replace it with a real example instead, something concrete and specific rather than a comparison to something else entirely.

A ScriptGrain profile whose analogy preference reads no sends that answer as a hard constraint in every draft it writes, and at scoring time a draft Claude Haiku reads as analogy-led scores 0.2 on this feature instead of 1.

Questions

What is an analogy in writing?

An analogy in writing is a comparison used to explain one thing by relating it to something else the reader already understands. It's mainly a tool for clarity, though it can decorate too. In a voice profile, analogy preference just asks whether this is a habit for the writer or not: a straightforward yes or no.

What is the difference between analogy and metaphor?

An analogy compares two things directly to explain something, often spelling out the comparison ("it's like X because Y"). A metaphor asserts one thing is another, without the explanatory scaffolding ("time is a thief"). Analogies tend to teach; metaphors tend to evoke. A writer can favour one, both, or neither, and ScriptGrain reads metaphor use separately from analogy preference.

How do you use analogies in writing effectively?

Use analogies when a concept is genuinely hard to grasp, not as a stylistic habit. Draw the comparison from something your reader already knows well, keep it short, then return to the literal explanation straight away. If the analogy needs its own footnote to make sense, cut it.

How is analogy use scored in a voice profile?

Analogy preference is judged yes or no by an extraction model reading your samples, then it enters the voice match score at a weight of 0.5. Claude Haiku judges the draft the same way, yes or no, and compares that against your profile. A matching answer scores fully; a mismatch scores 0.2. It's one of 28 attributes that can move the score at all.

How do you make ChatGPT stop using analogies?

Tell it plainly to explain directly with no analogies, since that's a clearer instruction than asking it to "tone down" comparisons. Then check the draft for any comparison that needs its own explanation and cut it. Where an analogy was doing the work of evidence, swap in a real example instead: something specific, not another comparison.

Related

Sources