Home › SEO Guide › How to write a page that answers a question
How to write a page that answers a question
Texts on websites are written, almost always, like school essays: introduction, body, conclusion. That is exactly the wrong order. The reader on the internet does not read to the conclusion, and an engine that extracts a passage takes the first one that looks complete.
The rule that changes everything: the answer first
“Choosing an insurance policy is an important decision that depends on many factors. In recent years, the insurance market in Romania has seen significant development, and the offer is increasingly varied. In what follows we will analyze...”
“The difference is simple: RCA covers the damage you cause to someone else and is mandatory, while CASCO covers the damage to your own car and is optional. If the car is worth under 5,000 euros, many owners choose only RCA.”
The second text starts with the answer. Whoever reads it finds out what they wanted to know in ten seconds. And an engine looking for a passage to quote has it ready, at the top of the page.
A good answer block has between 40 and 60 words and makes sense taken out of context.
The test: cover the rest of the page and read only that paragraph. If it makes sense on its own, it can be quoted. If it starts with “that is why” or “as we saw above”, it cannot.
Subheadings are questions, not labels
A subheading that says “General aspects” helps no one. One that says “How much does an RCA policy cost for a driver under 25” matches a real search and becomes, on its own, an entry point into the page.
| Instead of | Write |
|---|---|
| General considerations | What the policy covers and what it does not |
| Financial aspects | How much it costs and what the price depends on |
| The procedure | What to do in the first 24 hours after an accident |
| Conclusions | Which option suits you |
On our articles we went one step further: subheadings phrased as questions also automatically become structured data of the frequently asked questions type. In other words, the same work produces both a text that is better to read and information the engine can display directly.
The table of contents, which is not decorative
On articles over a thousand words we put a table of contents generated automatically from the subheadings. It has three effects, in order of importance:
- The reader sees right away whether the page contains what they are looking for and jumps straight there.
- Each section gets its own address, so it can be sent to someone as a link.
- The search engine sees the complete structure of the page before crawling it.
A detail that matters: we generate it on the server, not in the browser. A table of contents built from code is seen only by whoever runs the code, and in chapter 3 we showed what happens then. And if the article has fewer than three subheadings, we do not put a table of contents at all; a table of contents with two entries is noise.
How long it should be
There is no magic number of words, despite what you will read in articles. There is, however, a useful question: have you fully answered the question in the title?
If the question is “what documents do I need for a Schengen visa”, the complete answer has a list and a few clarifications, so about 600 words. If it is “how do I choose home insurance”, the complete answer requires comparisons, so about 1,500.
Text stretched to hit a word target can be felt, and the reader abandons it. Text that is too short leaves the question open, and the person goes back to the search, which is the clearest signal that the page did not do its job.
What turns a good text into a quotable one
Here we have, for the first time, measurements. A paper presented at the KDD conference in 2024, signed by researchers from Princeton and three other institutions, tested nine ways of modifying a text and measured how often it ends up cited by a generative engine.
| What was tested | The effect |
|---|---|
| Adding quotations from sources or experts | The best result for visibility, 22% above the unmodified text |
| Adding statistics with concrete figures | Up to 37% improvement in how the text is perceived |
| Citing sources | Improvement, and best in combination with the others |
| Repeating keywords | Worse than doing nothing. 21.9 versus 24.1 for the unmodified text |
The last row deserves emphasis: the most widespread practice in “optimized” texts produces results below the level of not intervening at all. The figures above are those measured on Perplexity, a real engine, not in a laboratory.
Translated into writing instructions: put in concrete figures with their source, quote someone who knows, and do not repeat the target phrase. It is almost exactly what you would do if you were writing for a demanding person.
The complete structure of a page
- A title that contains the question, phrased like a person, not like a catalog.
- The answer, in the first 40 to 60 words.
- The table of contents, if the text is long.
- Sections with question subheadings, each with its answer at the beginning.
- Figures, examples, a table where things are compared.
- What the reader should do next, concretely.
- The sources, in their own section, with links that actually work.
- Who wrote it and when it was updated.
The quick test, on any existing page
- Read only the first paragraph. Did you find out the answer?
- Read only the subheadings. Is it clear what the page is about?
- Cover the page and leave a single paragraph visible. Does it make sense on its own?
- Count the concrete figures in the text. Zero is a bad sign.
- Search for the target phrase. If it appears more than a few times, take it out.
- Is there a sources section, with links that work?