SEO Copywriting: Writing for Search Engines and People at Once

Why do so many SEO texts read so badly?

Almost every business knows both kinds of text. Some are clearly written for the search engine: the keyword appears in every other sentence, the introduction repeats the heading, and at the end the reader has learned nothing new. Others are carefully worded, but nobody finds them, because they do not contain the customers' words and the answer is hidden in the fourth paragraph.

That raises one question: How do you write a text that search engines find and people enjoy reading, without serving one reader at the expense of the other? To answer it, we first look at how people read online and introduce a model from reading research. Then we weigh up two views on whether search engines and people demand opposite things. At the end there is a test procedure and a checklist for your next page.

Which words belong in a text is covered in the article researching keywords, and which kind of page fits which query in the article understanding search intent. This guide starts after that: with the writing itself.

A good search engine text is a text a person in a hurry understands. The machine reads in the same places.

How do people read online? The principle of good enough

An SEO text is a text written and structured so that search engines recognize which question it answers, and so that it gives a usable answer to people who arrive with that question. SEO stands for search engine optimization. The term sounds technical, but at heart it is a writing craft.

The thinking model of this article comes from reading research. In the 1950s, Herbert Simon coined the term satisficing, a blend of satisfy and suffice: people do not look for the best solution but for the first one that is good enough. Duggan and Payne (2011) showed that people read exactly this way when they skim texts under time pressure. They read a section as long as it delivers enough new information. When the yield drops below a threshold, they jump to the next section. The readers in their study were able to direct their attention to the most important passages and therefore retained more of the meaning of the text.

How much skimming differs from careful reading is summarized by Fitzsimmons, Jayes, Weal and Drieghe (2020). Early eye-tracking studies by Just and Carpenter already found that skimmers read about two and a half times as fast, looked at fewer words and dwelt more briefly on each. In their own two experiments with web pages, Fitzsimmons and colleagues showed that skimmers fully processed linked words but only fleetingly processed unlinked ones. Readers therefore use links as markers for what matters. The authors conclude that readers online try to lose as little understanding as possible while staying fast.

For a business, this means that the visitor to your service page does not read from top to bottom. They check headings, first sentences and highlighted words and decide again at every point whether reading on is worth it. A text that takes three paragraphs to get to its point loses them before that.

Search engine or person: two readings

The first reading sees a real conflict of goals. Search engines, in this view, need repetition, length and the exact keyword in many places, while people need brevity and variety. Those who think this way write two texts in one: one for the reader at the top and a long block for the machine below. This view once had some basis, when search engines mainly counted words.

The second reading sees no conflict but the same task. Google describes its expectation explicitly in its guidance on helpful content: content should be created primarily for people and not to manipulate rankings. To the question of whether you should write to a particular word count because Google supposedly prefers one, Google itself answers: no. Repeating the same words until it sounds unnatural is listed in Google's spam policies as an example of keyword stuffing, that is, filling a page with keywords.

I consider the second reading correct, with one qualification. A remnant of the conflict remains, and it lies in the words. Furnas, Landauer, Gomez and Dumais (1987) found that two people spontaneously choose the same word for the same thing with a probability of less than 20 percent. A text that uses only the in-house jargon is easy to read for colleagues and invisible to searchers. The task is therefore not to write for the machine but in the customer's language. Then the machine understands it too.

Where search engines and readers look in the same places

The principle of good enough leads to a practical observation: the places where a hurried person decides are the same places where a search engine recognizes what the page is about. If you fill them deliberately, you do not need a second text for the machine.

The meta description does not appear on the page, but search results often show it. It is the advertisement for the click: one sentence on what the reader gets and one on whom it is for.

Link text deserves special care because, according to the findings of Fitzsimmons and colleagues (2020), it is read even when skimming. How pages strengthen each other through links is described in the article internal linking.

The skimmer test: checking a text in three minutes

From the model and the findings I derive a test procedure that I call the skimmer test. It simulates the hurried reader. A text that passes it has, as a rule, also filled the places where the search engine reads.

1. Write down the question

   Write down in one sentence which question the page is supposed to answer. If you cannot, the problem is not the text but the page.

2. Read only the frame

   Read only the title, main heading, subheadings, the first sentence of each section and the link texts. Everything else stays covered.

3. Check the answer

   After reading the frame, can you answer the question? If not, the answer is missing from a load-bearing place.

4. Check the words

   Does the word the customer searches with appear? Is it in the title and the main heading? Or only the in-house word?

5. Check the next step

   Does the skimmer know at the end what they can do: call, inquire, read on? A text without a next step ends in nothing for both readers.

The test takes about three minutes per page. It does not replace proofreading, but it finds the most common mistakes: the answer in the fourth paragraph, headings without a message and link texts that reveal nothing.

How long should an SEO text be?

The question of the right length is the most common and the least fruitful. Google names no preferred word count. A text is as long as the question requires. A knowledge question such as “How long does a flat roof last?” needs an answer in the first sentence, followed by influencing factors, examples and limits. A service page needs scope, process, service area, evidence and the inquiry path. A text does not get better because it is longer, but because it answers more of the reader's questions.

If you want to check length, you therefore check behavior, not words. The following worked example shows a service page over one month, measured with Search Console and your own inquiry list.

How to read these figures is described in the article understanding Google Search Console. What matters is the order of the questions: is the page found, is it clicked, is it read, does anyone inquire? Each stage has its own lever in the text.

What follows beyond the individual page

If search engines and people read in the same places, the distinction between an SEO text and a good text is essentially outdated. What remains is a craft with two rules: write in the customer's language, and put the answer where a hurried reader looks. This also applies to AI assistants that compile answers from websites. The article visibility in AI answers describes why clearly structured pages with an answer in the first sentence have an advantage there.

The limits of the findings are clear. The studies on skimming come from the laboratory, with students, expository texts and Wikipedia-like pages, not with buyers checking a service page. Google does not disclose exactly how it evaluates texts; its guidance describes goals, not formulas. Whether one phrasing ranks better than another can hardly be proven for a single business. What can be proven is whether visitors inquire. Every text should be measured by that.

If you hand the writing over, for example to quik-services, a service run by a business partner of Owners Council for websites and search engine optimization, still check every delivered text with the skimmer test.

Conclusion and next steps

The guiding question was how a text can be found by search engines and enjoyed by people at the same time. The answer: by being written for the hurried person. Online, people read by the principle of good enough and decide on the basis of headings, first sentences and links whether to stay. These are exactly the places where the search engine recognizes what a page is about. If you put the customer's words and the answer to their question there, you do not need a second text for the machine. The remaining conflict is not between person and machine but between the language of the business and the language of its customers.

• Pick your most important service page and write down the question it should answer.
• Run the skimmer test: read only the title, headings, first sentences and links.
• Rewrite every subheading without a message as a question or statement.
• Rearrange the first sentence of each section so that it contains the answer.
• Check whether the customer's word appears in the title and main heading.
• Replace every “learn more” or “here” link text with the link's destination.
• After one month, check in Search Console whether clicks and inquiries have changed.

For years, search engines have been learning to read like people. It is sensible not to stand in their way.