Semantics.rs
Semantics.rs/SEO/Clear writing
Content

Clear SEO writing: answer, condition, evidence.

Good writing makes reading easier and reduces ambiguity. There is no documented rule that BERT or MUM require short subject–predicate–object sentences, so we fit the syntax to the meaning and the audience.

Author: Precise Search SEO · Published and checked:
On this page
  1. How a useful section opens
  2. What we cut
  3. Before and after
  4. Writing the first sentence
  5. Headings as questions
  6. Paragraphs that stand alone
  7. The order of sections
  8. Fact, estimate, recommendation
  9. Terminology consistency
  10. Editorial checklist

How does a useful section open?

When an H2 poses a question, the first sentence should give a self-contained answer. Conditions, exceptions, examples and the source follow after it.

Good writing makes reading easier and reduces ambiguity. There is no documented rule that BERT or MUM require short subject–predicate–object sentences, so we fit the syntax to the meaning and the audience rather than to an assumed model.

What do we cut from a text?

Short sentences do not automatically win. Brevity is not a goal in itself: one clear claim per sentence is a useful rule for critical data, while a more complex sentence can express cause, condition or exception more precisely.

Before and after: same information, different structure

The most common problem is not style but order. The answer sits in the third paragraph, behind an introduction that asserts nothing.

Before:

These days, the question of taxi pricing is increasingly on people’s minds. Many wonder how much a ride actually costs and how the price is formed. In what follows, we will attempt to answer some of the most frequently asked questions on this topic.

Three sentences, not one claim. A reader who came for a number still does not have it.

After:

An 8 km ride in Belgrade costs about 1,160 dinars during the day — a 320-dinar base fare plus 105 dinars per kilometre. At night the same ride is about 1,400 dinars, because the 135-dinar rate applies. The amounts are set by the city and are identical for every registered service.

The first sentence contains the answer with a number. The second introduces the condition that changes it. The third gives the basis for the claim. That is the answer–condition–evidence pattern.

Writing the first sentence of a section

Patterns we avoid and what replaces them
Instead ofWrite
“There are several factors that affect the price.”“Three things set the price: the base fare, the per-kilometre rate and the tariff in force at that moment.”
“It is important to note that prices vary.”“Prices vary by city — the base fare runs from 80 dinars in Valjevo to 320 in Belgrade.”
“The answer to this question is not simple.”“The short answer is no, with one exception explained below.”
“Many experts agree that…”“Google’s documentation states that…”, with a link to the source.

What all the left-hand examples share is that they announce an answer instead of giving one. Such a sentence can always be deleted without losing information — which is also the fastest way to test for it.

Headings as questions, not labels

A heading phrased as a question obliges the first sentence beneath it to be an answer. A heading used as a label obliges nothing, so the section often drifts.

Label versus question
Heading as a labelHeading as a question
PricingWhich tariffs exist and when do they apply?
Our methodologyHow do we verify every figure before publishing?
BenefitsWhat does this approach solve that the previous one did not?
ConclusionWhat follows from all this for your own site?

Not every heading has to be a question. For sections that carry a procedure or a set, the imperative works better: “Check it in five steps”. The rule is that a heading should announce the type of content that follows, not merely the subject.

Writing a paragraph that can be extracted

A paragraph is extractable when it makes sense on its own, away from the page. That means no pronouns pointing at the previous paragraph, no “as we saw above”, and a named subject.

This is not a trick for AI systems but a consequence of how people read — by scanning, entering the text in the middle. A paragraph that needs the two before it loses both the reader and any chance of being quoted.

Not self-contained: “Because of that it is considerably higher, particularly during that period.” Three vague references in one sentence.

Self-contained: “The night tariff is higher than the day rate — 135 against 105 dinars per kilometre in Belgrade — and applies between 10pm and 6am.”

What breaks a paragraph’s independence
PatternReplacement
“this”, “that”, “it” with no named subjectRepeat the noun, even at the cost of repetition
“as we saw”, “the above”Delete it; if the connection matters, name it
“in the previous table”“in the table of fares by city”
A sentence continuing from the previous paragraphMerge the paragraphs or start afresh

The order of sections within a document

Sections follow the order in which questions are actually asked, not the order that was easiest to write. That is almost always: answer, then conditions, then process, then limitations.

The usual order and its reasoning
#SectionWhy here
1A direct answer to the page titleMost visitors want only this
2Conditions and exceptionsThey change the answer, so they must follow it immediately
3How the mechanism worksFor those who want to understand, not only to know
4Procedure or applicationThe shift from understanding to doing
5Limitations and sourcesCloses the text and separates fact from estimate

The most common deviation is putting the mechanism first, because it feels logical to the author to start from fundamentals. The reader who came for an answer gives up before reaching it.

Marking the difference between fact, estimate and recommendation

Three types of claim need three different formulations. When all three are written in the same tone, the reader cannot tell what is verifiable.

Claim type and how it is written
TypeHow it is writtenExample
FactA claim with a source and a date“The base fare in Belgrade is 320 dinars, per the city’s official gazette.”
MeasurementA number, a period, the tool and who owns the data“The AI impression share was 14.6% between 12 June and 14 August, per the owner’s Search Console.”
EstimateAn explicit caveat and its basis“Probably because the pattern has more variations, but we did not measure it.”
RecommendationAn imperative with the condition under which it holds“Remove any URL that fails the demand threshold.”

The test is simple: for every claim in the text, ask what evidence would show it to be false. If no answer exists, the claim is either empty or an estimate presented as a fact.

Terminology consistency

Use one name per concept. Consistency does not mean a poor vocabulary — it means a change in wording is reserved for a genuine change in meaning.

If one paragraph says “query”, the next “search” and a third “search term”, the reader has to hold three labels for one thing. On a project with several writers this drift begins after about a dozen pages, and reconciling it afterwards takes longer than writing did.

Terminology decisions on this site
ConceptWe useWe do not mix with
What the user typesquerysearch, search term, keyword
A page appearing in resultsimpressionview, appearance
The pattern behind many queriesquery templatequery pattern, template query
The primary content of a pagemain contentbody, central section
Data describing the page for machinesstructured dataschema data, markup data

A table like this is worth keeping on any project with more than one author. It is also the fastest way to review a translated text, where terminology drift is the most common defect.

Editorial checklist before publishing

  1. Does the first sentence under each H2 answer the question in that H2?
  2. Can the opening paragraph be deleted without losing information? If so, delete it.
  3. Is there any number, date or comparison without a stated source and period?
  4. Is there a claim that would sound the same if it were reversed? That is an empty claim.
  5. Can the reader see the difference between a fact, an estimate and a recommendation?
  6. Is the same claim repeated in two sections in different words?
  7. Does every paragraph make sense if read on its own?
Documented basis: Google recommends helpful, reliable, people-first content. The way a text is phrased here is our editorial methodology, not a hidden NLP optimisation. Primary source.

Frequently asked questions

Do BERT or MUM require short sentences?

No documented rule of that kind exists. Clarity helps the reader, and we fit syntax to meaning and audience rather than to an assumed model.

How long should the answer under a heading be?

Long enough for the claim to be self-contained and verifiable. For most questions that is one to three sentences, followed by conditions and the source.

Does longer content rank better?

Length is not the goal. A page without its own intent consumes crawl budget regardless of word count, and a complete shorter text often outperforms a padded longer one.

How do you keep terminology consistent across writers?

Keep a short table of decisions — one agreed name per concept — and review against it. Drift usually starts after a dozen pages, and fixing it later costs more than writing did.