A summary you accepted is a document you have not read

Omission leaves no mark

A wrong sentence in a summary can be spotted by anyone who knows the material. A missing one cannot be spotted by anybody, because there is no gap on the page where it used to be. The text reads as complete, and completeness is exactly what a summary cannot demonstrate about itself.

This makes summarising unusual among these tasks. Most output announces its own weak points — a hedge, an awkward passage, a claim that sounds too strong. A summary of something you have not read presents a smooth surface, and the smoothness is a property of the format rather than evidence about the content.

The consequence is a particular kind of confident ignorance. You can discuss a document intelligently, quote its main argument and answer questions about it, while being unaware of the paragraph that would have changed your view entirely. Nothing in the experience distinguishes that state from having read the thing.

What gets dropped is not random

Summarising favours what is repeated, prominent and central. That is usually the right choice and it systematically loses a specific category of material: the single-sentence exception, the condition attached to an offer, the caveat in a footnote, the dissenting paragraph, the small figure that undermines the headline claim.

Those are frequently the parts that mattered. A contract is mostly boilerplate and a few clauses that determine what happens when things go wrong, and a fair summary of the whole document weights the boilerplate correctly and the clauses barely at all, because they occupy little space.

Structure suffers too. A summary tends to present a document as a set of points, and some documents are arguments where the order and the dependencies carry the meaning. Reduced to points, an argument that only works in sequence becomes a list of assertions, several of which are unconvincing in isolation.

Ask for the exceptions, not the gist

The useful requests are the ones that target what a general summary suppresses. List every condition, deadline and exception. Identify anything the document says is required of me. Extract each figure with the sentence around it. Find any statement that limits or contradicts the main claim.

These questions have checkable answers, which is the other advantage. If the response says a clause on page eleven imposes a notice period, you can go to page eleven, and the verification takes a minute. A general summary offers nothing similarly specific to confirm.

It is also worth asking what the document does not address, though the answer needs treating with care. It sometimes surfaces a genuine absence — no mention of what happens on cancellation — and it is a weaker question than the ones about presence, because absence is harder to establish than to assert.

Match the depth of reading to what is at stake

Not everything deserves a full read, and pretending otherwise is how people end up reading nothing carefully. Most incoming material is genuinely disposable, and a summary is an appropriate way to decide whether something belongs in that category.

The distinction worth holding is between deciding and committing. Summaries are good for triage: is this relevant, does it concern me, should I read it. They are a poor basis for signing, quoting publicly, making a financial commitment or telling other people what a document requires of them.

A middle position covers most professional reading. Take the summary, then read the sections it identifies as important plus the ones it barely mentions, particularly anything about obligations, money, termination or liability. That is fifteen minutes rather than an hour and it removes most of the exposure.

Chains of summaries are the worst case

A summary of a summary compounds the omissions and loses the ability to check, because the original is now two steps away. Each stage looks reasonable and the final version can be confidently wrong about material that was accurately represented at the first step.

This happens organisationally without anyone deciding to do it. A report is summarised for a meeting, the meeting produces notes, the notes are summarised in an update, and a decision is eventually made on the basis of a paragraph whose relationship to the underlying evidence nobody can reconstruct.

The practical defence is to keep the link to the source and to require quotation at the first step. If the summary carries the sentences it rests on, the second-hand reader has something to check, and the chain has at least one point where the material can be recovered.