Draft first, compare second: a better way to learn from model answers
Info: 2697 words (11 pages) Blog
Published: 14 Jul 2026
There is a moment familiar to almost every student. An assignment lands, the brief feels opaque, and somewhere – on the virtual learning environment, in a shared folder, or a browser tab away – sits a model answer: polished, complete, reassuring. The temptation is to open it immediately, absorb its structure and phrasing, and only then begin to write. It feels efficient. It feels safe. And, according to a growing body of research, it is probably the wrong way round.
The argument of this article is simple: write your first draft before you look at the exemplar. Viewed too early, a model answer becomes a template to imitate. Viewed after you have drafted, it becomes something far more valuable – a source of feedback on work you have already produced.
Why exemplars work at all
Before making the case for timing, it is worth being clear about why exemplars matter in the first place. Assessment criteria are, by their nature, abstract. Marking criteria / a rubric can tell you that a first-class essay demonstrates “critical evaluation” and “sophisticated synthesis of sources”, but those phrases are close to meaningless until you have seen them enacted on the page. What does sophisticated synthesis actually look like in the third paragraph of a literature review? How does a strong conclusion differ, sentence by sentence, from a weak one?
Exemplars answer these questions in a way no list of criteria can. A systematic review of their educational uses found that exemplars help students grasp assessment standards that written criteria alone often fail to convey, making tacit expectations visible and discussable (To, Panadero and Carless, 2021). This is why tutors share them, why students hunt for them, and why an entire ecosystem – from university exemplar banks to commercial model-answer services – has grown up to supply them.
None of that is in dispute. The question is not whether to use model answers, but when. And on that question, timing turns out to change what an exemplar fundamentally does.
The problem with peeking early
When you study a model answer before writing, the most natural cognitive move is emulation. You map its structure onto your own plan, borrow its rhetorical moves, adopt its way of framing the question, and – often without noticing – narrow your thinking to the shape of someone else’s solution. Psychologists studying design and problem-solving call this fixation: once a worked solution is in view, it becomes remarkably difficult to imagine alternatives. The exemplar stops functioning as a standard against which to judge your work and starts functioning as a scaffold you never chose to test yourself without.
There is a second, quieter cost. Learning science has long recognised that generating an answer yourself – even an imperfect one – produces more durable learning than studying a ready-made solution. The struggle is not an unfortunate side effect; it is the mechanism. When you read the model first, you skip the struggle entirely. You never discover which parts of the task you could handle and which parts defeated you, because you never faced the task unaided. Your sense of understanding is borrowed, and borrowed understanding has a habit of evaporating in the exam hall or the next assignment, when no model is available.
Contrast this with the post-draft sequence. Once you have wrestled a first draft into existence, you arrive at the exemplar with something no template can give you: a detailed, personal map of your own gaps. You know where your argument sagged, where your evidence felt thin, where you were bluffing. Reading the model answer at that point is not imitation; it is targeted search. You are looking for solutions to problems you have already diagnosed – and a solution you searched for sticks in a way a solution you were handed never does.
What the evidence shows
Until recently, this was largely a theoretical claim. The first direct test of exemplar timing has now been conducted, and its results are striking. Nicol and Rose (2025) compared students who analysed exemplars before producing a report with students who wrote first and analysed the exemplars afterwards. The draft-first group earned significantly higher grades. Just as importantly, they generated self-feedback that was more actionable and more personalised, and they described the process as empowering precisely because it allowed them to learn from their own mistakes. The mechanism the researchers propose is intuitive: exemplars encountered after writing primarily feed self-reflection and revision, because students first detect their own shortfalls and then mine the exemplar for ways to address them.
This finding does not stand alone. A series of controlled studies shows that exemplars provided for revision – that is, after a draft exists – reliably improve writing from draft to final version. Annotated exemplars improved middle-school essays across drafts, performing comparably to individualised teacher comments while demanding far less staff time (Tomazin, Lipnevich and Lopera-Oquendo, 2023). Among university writers, exemplars supplied after a first draft produced substantial revision gains (Lipnevich et al., 2014). And in second-language writing, both exemplars and rubrics given as post-draft feedback outperformed a control condition not only on the task at hand but on a similar new task – evidence that the benefits transfer beyond a single piece of work (Burnell et al., 2023).
The self-regulation literature helps explain why the sequence matters. Learning to write well is, in large part, learning to monitor and manage your own writing: planning, checking your text against a standard, diagnosing weaknesses, and revising strategically. Classic work on observational learning shows that studying models can build revision skill and self-regulatory capacity (Zimmerman and Kitsantas, 2002), while more recent studies confirm that feedback targeting process and self-regulation increases students’ planning, monitoring and strategy use (Yang, Zhang and Dixon, 2023). Comparing your own writing against exemplars has likewise been shown to improve both the quality of texts and the regulation of the writing process itself (Vandermeulen et al., 2022). The draft-first sequence exploits exactly this machinery: it forces the acts of self-monitoring and gap-detection that pre-draft exemplar study allows you to skip.
Where do you find an exemplar?
All of this assumes you have an exemplar to compare against – and here students face a practical problem. Tutor-provided models are the gold standard, but they are not always available, are sometimes deliberately withheld, and frequently answer a different question from the one you have been set. Comparing your essay on one topic against a model answering another blunts the whole exercise: half the differences you notice will reflect the questions, not the quality.
In practice, students draw exemplars from several sources, each with distinct strengths.
Tutor-provided exemplars and past student work.
Where these exist, use them first. They are calibrated to your module’s actual standards, and annotated versions – where the tutor has marked up why a passage earns credit – are especially powerful, since the annotations do some of the comparative work for you.
Published academic writing.
Journal articles and review papers in your field are exemplars of a kind, particularly for structure, sourcing and academic voice. Their limitation is level: they model professional writing, not what a first-class undergraduate answer to your specific brief looks like.
Custom model answers from essay writing services.
Professionally written model essays occupy a useful middle ground: unlike a generic past paper, a custom model answer responds to your exact question, at a specified academic level, with proper referencing. Used within the draft-first sequence, this is precisely what makes them valuable – you get a like-for-like comparison document, matched to your brief, against which the differences you spot are genuinely informative about quality rather than topic. The key is the order of operations: commission or obtain the model, but do not open it until your own draft exists. Then treat it exactly as you would a tutor’s exemplar – as a standard to compare against and mine for techniques, with your own draft remaining the document you develop and submit. (Institutional policies on external materials vary, so it is worth knowing yours.)
Not all custom model answer services are the same – look for an essay writing service like UK Essays, with verified Press coverage as an indicator of trust.
AI essay writers.
Large language models have made exemplars effectively free and instant, and they bring one capability no other source offers: multiplicity. Rather than one model answer, you can generate three or four in minutes – and the research on exemplars suggests that comparing multiple examples is more powerful than studying one, because variation reveals which features are essential to quality and which are merely one writer’s choices.
AI models do have characteristic weaknesses worth knowing about practically: generated references may be inaccurate or invented, so treat an AI exemplar as a model of structure, argumentation and style rather than a source of citable content, and verify anything factual independently. Their arguments can also drift towards the generic – which, usefully, is itself instructive: if your draft makes the same safe, obvious points as the AI’s, that is a signal your thinking has not yet gone beyond the default.
Uniwriter is our recommended AI essay writer and it has addressed many of those concerns with extensive AI training on past excellent papers under the guidance of real academics. The BBC graded one of its 2:1 standard essays and confirmed it “was of a 2:1 standard and had no mistakes whatsoever” (BBC News, 17 Dec 2025). The same team has also developed a number of subject specialist writers including lawwriter.ai and businessessays.ai.
Whatever the source, the principle is identical. The exemplar’s job is not to write for you; it is to show you the gap between what you produced and what the standard looks like. Any source that can render that gap visible – tutor, journal, professional writer or machine – can serve, provided it enters the process after the draft.
A working protocol
For students, the practical sequence looks like this:
- Draft blind. Produce a complete first attempt using only the brief and the assessment criteria. Resist the tab. The draft will be imperfect; that imperfection is the point — it is the diagnostic data the next steps depend on.
- Diagnose before you compare. Before opening any model, note two or three things you suspect are weak: the limp introduction, the section where the argument went circular, the paragraph you padded. This primes the targeted search that makes exemplar reading productive.
- Obtain a matched exemplar. A tutor model if one exists; otherwise a custom model answer to your exact question, or two to three AI-generated versions. If using AI like Uniwriter, generate variants – ask for the essay written at different standards, or from different argumentative angles – so you can see what varies and what stays constant across strong answers.
- Read the exemplar as feedback, not as a template. Ask not “what does it say?” but “what does it do that mine doesn’t?” – in structure, evidence, precision, signposting and voice. A useful technique is the reverse outline: summarise each paragraph of the model in one line, do the same for your draft, and set the two skeletons side by side. Structural gaps become visible almost immediately.
- Write yourself feedback, then revise. Nicol and Rose’s students benefited most when comparison generated concrete, personalised comments to act on. Turn your observations into instructions to yourself – “move the counterargument earlier”, “replace assertion in paragraph four with cited evidence” – and then execute them on your draft. If you are working with an AI tool, you can extend this step: paste in your draft alongside the model and ask it to articulate the differences, or to critique your draft against the assessment criteria. The machine becomes a comparison aid rather than a ghostwriter, and every revision remains your own.
- Close the loop. After submission and marking, revisit both your final version and the exemplar. Which gaps did you successfully close? Which did you miss? This last comparison is what converts a single assignment into transferable skill.
An honest caveat
High-end claims deserve honest boundaries, and two are worth stating.
First, the direct evidence on timing rests, so far, on a single head-to-head comparison (Nicol and Rose, 2025). The finding is consistent with theory and with the broader revision literature, but replication across disciplines and levels is still needed.
Second, exemplars are not magic, and they are not always the strongest form of feedback. In the Lipnevich et al. (2014) study, a detailed stand-alone rubric actually produced the largest revision gains, and later experimental work has similarly found that rubrics often equal or outperform exemplars depending on the outcome measured (Lipnevich, Panadero and Calistro, 2022).
Interventions combining rubrics with exemplar-based feedback have benefited argumentative writing, including for at-risk writers, but the specific contribution of the exemplar component is hard to isolate (Peltzer et al., 2025). The sensible reading is not “exemplars beat everything” but “exemplars and criteria work best together” – use the rubric to know what to look for, and the exemplar to see what it looks like. The strongest defensible claim about timing, meanwhile, is that when you consult an exemplar, doing so after drafting best supports revision-focused self-regulation – and, on current evidence, better final work.
The mirror, not the crutch
The discomfort of drafting without a model is real, and it is tempting to interpret that discomfort as a sign you are not ready to write. The research suggests the opposite: the struggle of the blind first draft is what converts a model answer from a crutch into a mirror. This holds regardless of where the model comes from. A tutor’s exemplar, a professionally written model answer and an AI-generated essay are all, in the end, the same instrument – and like any instrument, their value depends on how they are played. Opened before you write, they hand you someone else’s answer. Opened after, they hand you something rarer and more durable: a precise, personalised picture of the distance between your work and the standard, and a set of concrete moves for closing it.
Look too early and you learn what a good answer looks like. Look after you have written and you learn something more useful – what your answer needs.
References
Burnell, K., Pratt, K., Berg, D. A. G. and Smith, J. K. (2023) ‘The influence of three approaches to feedback on L2 writing task improvement and subsequent learning’, Studies in Educational Evaluation. doi: 10.1016/j.stueduc.2023.101291.
Lipnevich, A., McCallen, L. N., Miles, K. P. and Smith, J. K. (2014) ‘Mind the gap! Students’ use of exemplars and detailed rubrics as formative assessment’, Instructional Science, 42, pp. 539–559. doi: 10.1007/s11251-013-9299-9.
Lipnevich, A., Panadero, E. and Calistro, T. (2022) ‘Unraveling the effects of rubrics and exemplars on student writing performance’, Journal of Experimental Psychology: Applied. doi: 10.1037/xap0000434.
Nicol, D. and Rose, J. (2025) ‘Promoting learner self-regulation: is it better to give students exemplars before or after producing work?’, Assessment & Evaluation in Higher Education, 50, pp. 1311–1331. doi: 10.1080/02602938.2025.2534870.
Peltzer, K., Lorca, A. L., Krause, U.-M., Graham, S., Panadero, E. and Busse, V. (2025) ‘How to support at-risk writers: Differential effects of formative feedback on argumentative writing and motivation’, Reading and Writing. doi: 10.1007/s11145-025-10652-w.
To, J., Panadero, E. and Carless, D. (2021) ‘A systematic review of the educational uses and effects of exemplars’, Assessment & Evaluation in Higher Education, 47, pp. 1167–1182. doi: 10.1080/02602938.2021.2011134.
Tomazin, L., Lipnevich, A. and Lopera-Oquendo, C. (2023) ‘Teacher feedback vs. annotated exemplars: Examining the effects on middle school students’ writing performance’, Studies in Educational Evaluation. doi: 10.1016/j.stueduc.2023.101262.
Vandermeulen, N., Van Steendam, E., De Maeyer, S. and Rijlaarsdam, G. (2022) ‘Writing process feedback based on keystroke logging and comparison with exemplars: Effects on the quality and process of synthesis texts’, Written Communication, 40, pp. 90–144. doi: 10.1177/07410883221127998.
Yang, L., Zhang, L. and Dixon, H. (2023) ‘Understanding the impact of teacher feedback on EFL students’ use of self-regulated writing strategies’, Journal of Second Language Writing. doi: 10.1016/j.jslw.2023.101015.
Zimmerman, B. and Kitsantas, A. (2002) ‘Acquiring writing revision and self-regulatory skill through observation and emulation’, Journal of Educational Psychology, 94, pp. 660–668. doi: 10.1037/0022-0663.94.4.660.
Cite This Work
To export a reference to this article please select a referencing stye below:
