Sample essays should be used to reverse-check diagnostic evidence.

Common mistake

Copying sentences from samples does not teach you to develop your own idea on a new prompt.

Diagnostic evidence

If you only remember sentences but cannot explain why the paragraph works, analysis is shallow.

What to confirm before writing

Mark thesis, topic sentences, examples, and linking-back sentences in the sample.

What to revise in the next version

Transfer the structure to a new prompt while replacing the ideas and examples.

• Whether structure is visible
• Whether development is understood
• Whether copying is avoided

Analysis vs memorisation: one digs in, one copies out

Memorisation asks "what makes this sentence good"; analysis asks "why did the writer make this decision here, and what would a different decision cost". The first produces an excerpt notebook; the second produces transferable judgement. The test: close the sample and try to state how the position was chosen, what job each body paragraph does, and where the concession sits and why. If you cannot, you were memorising.

Worked example: three decisions in a Band 7 introduction

Introduction: "As cities grow denser, whether governments should prioritise public transport over road building has become a genuine policy dilemma. I side largely with public transport investment, though roads in rural regions remain a different matter." Decision one: the paraphrase copies no full phrase from the prompt — comprehension, not synonym swapping. Decision two: the position is qualified with "largely" — an interface reserved for the concession paragraph, not vagueness. Decision three: "rural regions" draws the boundary in one stroke, freeing the body paragraphs from that exception so they can go narrower and deeper. All three decisions transfer directly into your own next opening.

Two steps to convert analysis into your own writing

Step one: compress the sample's decisions into a card of at most five entries (not sentence excerpts), e.g. "qualify the position, one mechanism per paragraph, always link back after the example". Step two: take a new prompt on the same topic and write using only the card, then check each entry afterwards. The entry you failed to execute is your real gap from that sample — more precise than any overall impression.

• Decision card ≤ 5 entries, all actions, no sentences
• New same-topic prompt, without rereading the sample
• The unexecuted decision = the next practice target

Choosing samples worth analysing

Prefer official samples with examiner comments, then anything with a verified band label: analysing a real 6.5 beats analysing an unsourced "perfect essay", because the 6.5's problems are your problems, while viral model essays are often native-speaker showcases whose decisions do not survive exam conditions. A 6 and an 8 on the same prompt, read side by side, is the highest-yield material there is.