the real question · the answer that travels · where good and findable part company
Two camps, both wrong. Writing for machines produces the flattened, hedged prose this course exists to prevent; refusing to think about it produces good work nobody reads. Almost everything that makes a piece findable is something that makes it better, and the short list of exceptions is worth knowing exactly.
A person searching and a machine indexing want the same four things, for different reasons that happen to agree.
Four things help you be found and cost you something as a writer, and the honest move is to know which four rather than to pretend there is no tension.
People look for: content discoverability, search intent, standalone answers, article metadata, and internal linking.
Read the last column. Where it says both, do it without thinking. Where it says costs you something, that is a decision rather than a rule.
| The move | Why a reader wants it | Verdict |
|---|---|---|
| Name the question in the reader’s own words | They recognise their problem in your first line | Both |
| Answer so the paragraph survives being lifted out | They can quote it to a colleague | Both |
| Make a claim rather than name a topic | There is something to agree or disagree with | Both |
| Honest headings, in the order things happen | They can skip to the part they need | Both |
| Repeating the key phrase more than you would naturally | Nothing. It reads as a tic | Costs you something |
| Answering the obvious question you would rather skip | Something, if they are new. Nothing if they are not | Costs you something |
| A summary at the top of a piece built as an argument | They get the conclusion without the reasoning | Costs you something |
| Splitting one strong piece into three thin ones | Nothing at all | Do not |
Take any paragraph out of the piece, put it on its own, and read it. If it still says something complete to somebody who has not read the rest, it is both findable and good. If it collapses without the page around it, it was never travelling anywhere. The same test day 17 sets for a question, applied to an answer.
Findability begins with the reader's real question and a clear answer; metadata, structure, and links help systems carry that meaning accurately.
Not your term for the problem — theirs. People with the problem do not know your vocabulary, which is why they have the problem. This is the one place where finding out what strangers type is genuinely useful, and it improves the piece whether or not anybody ever searches for it.
Quoted in a review, pasted into a message, read aloud in a meeting, or shown as an answer with no page attached. All four are the same requirement, and only the last one is about machines. Write for the first three and the fourth arrives free.
Two things, both narrow. Listing the ways a stranger might phrase your subject, which you cannot do from inside your own vocabulary. And reading your headings on their own and telling you what it thinks the piece covers — if that list is wrong, your structure is lying to somebody who skims.
A useful guide used the team's internal name for a problem customers described differently. Adding the reader's wording to the title and opening connected the same content to the real question.
A frequently quoted paragraph began with 'this approach' and lost its subject outside the article. Naming the approach made the answer portable without adding keywords.
Three thin pages competed for variations of one question. Combining them around a single claim produced a stronger answer and clearer internal paths to the narrower companion topics.
Discoverable writing names a real question in the reader's language, makes a clear claim, uses headings that describe the subject, and provides answers that still make sense outside the page. Metadata and links help systems interpret that work, but they cannot rescue an article whose purpose is unclear.
It does when keywords are repeated mechanically or when imagined ranking rules replace the reader's need. It does not when search research reveals the words people genuinely use, missing questions, or confusing terminology. Use that evidence to clarify the work, then write for the person who arrives.
Collect support questions, search suggestions, forum titles, internal site searches, and the phrases non-specialists use in conversation. Group expressions that seek the same outcome, then choose the clearest natural wording. Do not replace a precise expert term when the distinction itself matters.
Answers travel into search results, messages, notes, meetings, and AI summaries. A self-contained answer states its subject, gives the conclusion, and includes the condition that limits it. This helps readers while reducing the chance that a detached excerpt changes the original meaning.
Titles, descriptions, canonical URLs, structured data, and internal links identify the page and its relationships. They should be generated from the same article data the reader sees. Their job is accurate interpretation and navigation, not inserting claims or questions absent from the visible work.
A paragraph that needs its page has not gone anywhere, and never will.