EXTRACTIVE SUMMARY
A question type is the class of answer a question demands, independent of its subject matter. The type determines the answer shape, meaning the construction the first sentence beneath a heading must take for the answer to be extractable. Ten question types cover nearly every query a subject generates: boolean, definitive, distinctive, grouping, comparative, procedural, causal, quantitative, temporal, and locational. The types map onto contextual vectors, with each vector leading on two or three types and treating the others as supporting. Three types are systematically under-answered across published content, and those are quantitative, distinctive, and boolean, each for a different and predictable reason. A question inventory is ordered by dependency rather than by importance, because some types cannot be answered before others have been. A wrong answer shape produces a page that covers its question and still fails extraction.
What a question type is
A question type is the class of answer a question demands, independent of its subject matter.
“How much does a topical map cost” and “how long does a topical map take” concern different attributes and belong to the same type, because both demand a number with a unit. “What is a topical map” and “how do you build one” concern the same attribute and belong to different types, because one demands a category with a distinguishing property and the other demands a first action.
The type is a property of the question rather than of the domain, which is what makes it useful. A classification learned once applies to every subject a property ever covers.
Why the type determines the answer shape
The type determines the answer shape because extraction succeeds when the answer’s construction matches what the question asked for.
A heading asking a quantitative question is followed by a sentence. If the sentence states a number with a unit, one clean triple is available. If the sentence explains the factors that influence the number, a parser holds a passage about influences and no answer, and the page has covered the question without answering it.
The mismatch is invisible to the writer because the section is usually correct. The influences paragraph is accurate, useful, and belongs on the page. It belongs second.
This is the difference between coverage and answerability. A page can hold every fact a question requires, arranged so that none of them is the answer.
The ten question types
Ten types cover nearly every query a subject generates. Each demands one answer shape in the first sentence beneath its heading.
| Type | What it asks | Required answer shape | Failure mode |
| Boolean | Can, does, is, should | Yes or no as the first word, then the condition | Hedging into “it depends” before the verdict |
| Definitive | What is X | X is a [category] that [distinguishing property] | Describing what X does before saying what X is |
| Distinctive | How does X differ from Y | The single axis of difference, named | Listing attributes of both without naming the axis |
| Grouping | What are the types of X | The count, then the enumeration | Enumerating without stating how many |
| Comparative | Which is better, X or Y | The criterion, then the verdict under it | A verdict with no criterion, or a criterion with no verdict |
| Procedural | How do you X | The first action, or the step count | Explaining why the procedure matters before starting it |
| Causal | Why does X happen | The mechanism, stated as a because | Restating that X happens with more detail |
| Quantitative | How much, how many, how long | The number with its unit | Stating the factors that determine the number |
| Temporal | When does X | The time, date, duration, or trigger condition | A relative answer where an absolute one exists |
| Locational | Where does X | The place, position, or containing structure | Describing the surroundings instead of the position |
Two notes on the set. Distinctive and comparative are frequently conflated and are different: distinctive asks how two things differ, comparative asks which to choose. A distinctive question answered with a preference has answered a question nobody asked.
Grouping requires the count first because the count is what a reader needs to decide whether to keep reading, and because a parser holding “four types” can verify that four were delivered.
How question types map onto contextual vectors
Question types and contextual vectors are different classifications and they correspond predictably. The vector determines which types lead, and the remaining types appear as supporting questions.
| Contextual vector | Leading types | Supporting types |
| Definitional | Definitive, distinctive | Grouping, boolean |
| Procedural | Procedural, grouping | Quantitative, temporal |
| Comparative | Comparative, distinctive | Definitive, quantitative |
| Quantitative | Quantitative, temporal | Causal, boolean |
| Causal | Causal, definitive | Boolean, distinctive |
| Evaluative | Boolean, comparative | Quantitative, causal |
| Temporal | Temporal, procedural | Quantitative, causal |
| Problem-solution | Causal, procedural | Definitive, boolean |
The correspondence is a useful check on both. A page assigned a quantitative vector whose question inventory contains no quantitative questions has the wrong vector or the wrong inventory, and the mismatch is visible at brief stage before anything is written.
→ How to write a content brief: field provenance and the limits of writer discretion
The three systematically under-answered types
Three types are under-answered across published content, each for a predictable reason.
Quantitative. Writers avoid committing to numbers, because a number can be wrong and a range cannot. The resulting sentence states what the number depends on, which is the influences paragraph occupying the answer position. The remedy is to state a number with its conditions attached rather than to state the conditions instead of the number.
Distinctive. Distinguishing a subject from its neighbour requires knowing the neighbour, and a writer briefed on one subject usually does not. The resulting section lists attributes of both and leaves the reader to locate the difference. The remedy is at brief stage: a distinctive question in the inventory should name the axis of difference, not just the comparison subject.
Boolean. A boolean question has a yes or no answer and most subjects worth writing about have conditions attached, so the writer leads with the conditions. “It depends” is true and it is not an answer. The remedy is the verdict first and the condition second, which is a reordering rather than a change of content.
All three failures produce accurate sections. That is why they survive editing: nothing in them is wrong.
→ Content gap analysis: five gap types, and why competitor gap analysis finds only one
How to order the question inventory
A question inventory is ordered by dependency rather than by importance.
Three dependencies govern most orderings. Definitive precedes comparative and distinctive, because a reader cannot evaluate a difference between things they cannot yet identify. Grouping precedes procedural where the procedure operates on the group, since steps referring to types the reader has not met require forward reference. And causal precedes procedural where the procedure is non-obvious, because a reader who understands why a step exists follows it and a reader who does not substitutes their own.
Boolean questions sit early when they gate the rest of the page. A page answering “can you build a topical map without a query network” should answer it before the procedure, because a no changes what the reader does with everything after it.
Quantitative questions sit wherever their attribute sits. Cost, duration, and count are not a separate section in most articles, they are the answer to a heading that happens to ask for a number.
The ordering fixed in the inventory becomes the ordering declared in the extractive summary, which becomes the ordering of the sections. All three inherit from one decision made at brief stage.
→ The extractive summary: why declaring order first changes what gets extracted
What a wrong answer shape produces
A wrong answer shape produces a page that covers its question and fails extraction, which is the hardest failure to detect because nothing on the page is incorrect.
Three consequences follow. The passage is not liftable, because the first sentence beneath the heading states something other than the answer and a system selecting a passage gets the wrong one. The heading becomes a promise the section does not immediately keep, and readers scanning for the answer scroll past it. And a question gap is created on a page that appears to have no gaps, because absence of the answer is invisible to any check that confirms the topic is addressed.
The detection method is mechanical. Read each heading, read only the sentence beneath it, and ask whether that one sentence answers that one heading. Sections failing this are usually fixed by promoting a sentence that already exists further down.
BRIDGE
The answer shape standard applies at every heading on every page, which is hundreds of applications across a network and none of them individually consequential. That is precisely the kind of standard that holds for three articles and degrades by article forty, because it depends on a writer recalling which type each heading asked and which construction that type requires.
Specification is what makes it survive. The question inventory in a content brief names the type of every question the page answers, in order, which means the writer receives the required shape per heading rather than reconstructing it. A content brief system derives that inventory from the annotated query network by rule, so that building topical authority to rank produces pages whose answers sit in the position the question put them in.
→ Content brief systems: provenance enforced at scale
SUPPLEMENTARY CONTENT
How many question types one page should answer
A page answers as many types as its attribute generates, which in practice runs between four and eight. Fewer than four usually means the attribute was over-fragmented across URLs. More than eight usually means the page has absorbed an adjacent attribute.
The count is a useful scope check at brief stage and a poor one afterward, since by then the scope is already set.
Whether question types map onto query classes
Question types and query classes are independent. Query class records commercial stage as informational, commercial investigation, or transactional. Question type records the answer shape required.
A quantitative question can be informational, as in how long topical authority takes in general, or transactional, as in what an engagement costs. The two classifications answer different questions about the same query and both belong in the annotation.
Whether this taxonomy is standard
The set is drawn from question classification work in information retrieval, where taxonomies of this kind have existed for decades, and the specific ten used here are the ones that recur across the query networks of commercial subjects.
Treat the set as an operating taxonomy rather than as a standard anyone else recognises. What matters is that one property uses one classification consistently, not that it matches a published one.
What to do with multi-part questions
Split them. A query asking what something is and how much it costs is two questions of two types, and answering both in one sentence produces a construction that satisfies neither shape.
Multi-part queries are common in voice and conversational search and they decompose cleanly, because each part carries its own type.