Two people drafting the same document converge on a voice neither of them has

Shared documents used to sound like an argument

A document written by two people has always had seams. One writer is terser, another explains more, and the joins are visible in the finished text. Editors have spent careers smoothing those joins, and the roughness was a signal: you could see who wrote what, and therefore who to ask about it.

When both authors are drafting with assistance, the seams disappear early. Two sections produced from different notes by different people arrive in a similar register, and the document reads as unified from the first version onwards.

That sounds like a gain and it removes information. A consistent surface is now the default rather than the product of editing, so consistency has stopped telling you anything about whether the document was thought through as a whole.

Uniform prose can hide two incompatible positions

The real cost is not aesthetic. Where two authors disagree, the disagreement used to show up as a tonal collision, a section that hedged where the previous one asserted, a definition used differently on page two and page five. Those are the things a careful reader catches.

Smoothed prose absorbs them. The same term is used confidently in two sections meaning two different things, and nothing in the writing signals the conflict, so it survives into the version that goes out.

This is a specific check to add rather than a general worry. Before circulating, read the document once for definitions and once for commitments, asking whether every use of a key term means the same thing and whether any two sections promise different things to the reader.

Decide the argument before the drafting is split

The structural remedy is to make the agreement explicit and early. Two authors should settle what the document claims, what it will not say, and what each key term means, before either of them produces prose. This has always been good practice and it used to be optional, because the draft itself would eventually expose the gaps.

Write that agreement down in a short shared note rather than reaching it in conversation. A note can be checked against; a conversation gets remembered differently by each person, which is how two sections come to be written against two versions of the same decision.

Then split by section with the note in hand. What comes back will still need reconciling, and now there is a fixed reference to reconcile it against.

It is also worth agreeing how much drafting help each author intends to use, not as a rule but as information. Knowing that one section began as rough notes and another as a generated outline tells the reviewer where to read hardest, and that is a practical fact about the document rather than a judgement about anybody’s method.

Keep the authorship visible while the document is in progress

It is worth preserving the information the smooth surface removes. Comments naming who is responsible for a section, a working file where each part is attributed, or simply a note at the top of the draft listing who wrote what: any of these mean a question about a paragraph can be directed to a person.

This matters most for the parts nobody wants to own. In a document where every section reads the same, an assertion with no author attached tends to survive review, because each reader assumes the other one put it there deliberately.

Strip the attributions at the end if the finished document should be seamless. The point is to have them during the period when the questions are being asked.

Someone has to read the whole thing as a reader would

The final pass on a jointly drafted document is not a copy-edit. One person, ideally not the author of the largest section, reads the whole document from the top with no knowledge of who wrote what and asks whether it makes one argument.

That reading finds the things the process hides: the point made twice in different words, the section that answers a question the document never raised, the conclusion that does not follow from the middle. None of these are visible while working on a part.

For anything consequential, that reader should be a person rather than a checking step, because the failure being looked for is coherence across the whole, which is exactly the property that a document smoothed section by section can lack while appearing to have it.