The first draft you are shown becomes the piece you write

You cannot go back to not having read it

Ask for a draft before you have thought about the problem, and the draft becomes the ground you work from. You revise it, disagree with parts of it, improve its examples — and the structure it proposed, the framing it chose and the things it did not think to mention are now the shape of your piece.

This is ordinary human behaviour rather than anything about the tool. A starting point exerts a pull; editing is easier than originating; and having a plausible version in front of you removes the discomfort that would otherwise have driven you to find a better one.

What makes it hard to manage is that it does not feel like influence. It feels like efficiency, because you are working quickly and making changes, and the changes are real. The options you never considered leave no trace, so nothing signals that the field was narrowed before you started.

The middle of the range is where the anchor sits

A generated first draft is generally a competent version of the most conventional treatment of the subject. As a floor that is useful, and as a starting point it is a particular problem, because the conventional treatment is precisely what a piece worth reading has to get past.

The effect is visible at scale. A set of pieces produced this way tends to share a structure — the same order of sections, the same three points, the same balanced conclusion — not because anyone copied anything but because everyone started from the same central tendency.

It also suppresses the awkward angle. The observation that does not fit the standard structure is the one that makes a piece distinctive, and it is exactly the thing an initial draft will not contain, since it is by definition not the usual way of covering the subject.

Write badly first, then ask

Ten minutes of your own notes before requesting anything is the whole intervention. They do not have to be good. A rough argument, the three things you actually think, the example you want to use, and the point you are unsure about — that is enough to give you a position to compare against.

Comparison is where the value appears. Once you have your own version, the generated one becomes useful in a different way: it shows you what you left out, where your argument is unclear, and which of your points are so standard that they can be assumed rather than explained.

The order matters more than the effort. The same ten minutes spent after reading a draft produces refinements to that draft, while the same ten minutes spent before produces an independent view that the draft can then be measured against. Sequence is doing all the work here.

Use it for the parts that are not the idea

Anchoring is dangerous where the shape of the thinking matters and harmless where it does not. Formatting, transitions, a summary of established background, the boring middle section that has to exist — none of these benefit from your originality, and starting from a draft there costs nothing.

The parts to protect are the framing, the argument and the specific material only you have. If a piece has a thesis, write the thesis yourself. If it has a structure that carries the argument, decide the structure yourself. The rest can be assembled from wherever is convenient.

A workable arrangement is to write the outline unaided and generate within it. You keep the decisions that determine whether the piece is worth reading and hand over the labour that determines how long it takes, which is close to the division most experienced writers already run with human collaborators.

Not every task deserves a fresh mind

Most writing is not distinctive and does not need to be. A routine notice, a summary of a meeting, a standard section of a report — for these, anchoring on the conventional version is a positive, because conventional is what the reader wants and originality would be an obstacle.

The cost is concentrated in work where being unlike the standard version is the entire point: opinion, analysis, anything competing for attention, anything where your judgement is what is being purchased. Those are a minority of what most people write, and they are the pieces that matter for reputation.

So the practical rule is not to abstain but to notice which kind of task you are on. If a hundred other people could write the same document and it would be fine, start wherever you like. If the value depends on it being yours, spend the ten minutes first.