This is the number the methodology page promised to publish once it existed. It exists: 300 UK homepages attempted on 24 October 2026, 72 of which returned enough readable text to score, measured with the same algorithm as the readability checker.
The result: the median homepage scores 48 on Flesch Reading Ease. Plain-English guidance puts business writing at 60–70. 79% of the pages we could measure sit below 60, and the main cause is sentence length, not vocabulary.
The numbers
- Homepages scored
- 72
- Median score
- 48
- Median sentence
- 18.5 words
- Below 60
- 79%
| Band | What it reads like | Share |
|---|---|---|
| Below 50 | Difficult — university-level reading effort | 58% |
| 50–59 | Fairly difficult — needs a motivated reader | 21% |
| 60–69 | Plain English — the recommended band | 18% |
| 70 and above | Easy — plain, occasionally thin | 3% |
79% below 60 is the headline. What it does not mean: that the writing is bad. A homepage at 45 is usually written for someone who already knows what the company does. The problem is that most visitors do not.
Sentence length is the lever
The median page carries 18.5 words per sentence, and 51% of pages have more than a fifth of their sentences over 25 words. Twenty-five words is about two lines on a phone; the reader has to hold the start of the sentence in memory while they reach the end.
Splitting a 25-word sentence into a 12 and a 13 lifts the score by roughly four points without changing a single idea. Vocabulary is the smaller lever: most homepages already use ordinary words, then stack them into long noun phrases that cost syllables and add nothing.
Method, so the number can be checked
- Frame. The Tranco top-1M domain list, downloaded 24 October 2026 — a public aggregate of Alexa, Cisco Umbrella, Majestic and Quantcast rankings.
- Filter.
.co.ukand.ukdomains ranked below 10,000, one domain per registrable name. That left 24,329 domains. - Sample. 300 drawn with
random.Random(20261024), so the draw repeats exactly. - Crawl. One request per domain, homepage only, 8-second timeout, identifying user agent.
- Text. Paragraph text where the page has at least 120 words of it; otherwise the
mainelement with navigation, header, footer, forms and scripts removed. This is the rule that decides which pages can be scored at all. - Sentences. Split on terminal punctuation, fragments under three words discarded. Pages whose average sentence fell outside 8–38 words were excluded as extraction failures rather than scored, because that range is not human prose — it is a menu, a cookie notice or a table.
- Syllables and formula. Rule-based syllable counting (the public tool uses the same), Flesch Reading Ease = 206.835 − 1.015 × (words ÷ sentences) − 84.6 × (syllables ÷ words).
What happened to the other 228 pages
76% of the sample produced no usable score. Publishing that is the difference between a study and a screenshot:
implausible sentence length — 86; fewer than 5 sentences — 60; fetch failed — 52; under 150 words of text (JS-rendered or thin) — 29; score outside plausible range — 1
Most of the failures are javascript-rendered homepages: the HTML we receive contains a framework shell and no prose, so there is nothing to measure without a headless browser. Others are parked domains, bot walls and network timeouts. A bolder study would render every page and score 100% of the sample; that costs a headless browser per page, which is why most published numbers of this kind quietly skip the question.
Limits
- 72 pages is a small sample. The direction is clear; the second decimal place is not. The dataset is published so the draw can be repeated or extended.
- Homepages, not articles. Navigation and legal text are removed, but service names and calls to action remain.
- Popularity bias. Tranco ranks popularity, so larger sites are over-represented against one-person businesses.
- Readability is not comprehension. A short sentence that says nothing scores well.
What to do with it
Find the longest sentences on your most-visited page and cut them in half. The readability checker lists the longest sentences rather than only scoring the page, which is the part that leads to an edit. If you write for both markets, the spelling problem is separate: 166 British and American differences.
Citing this
The dataset is a CSV: one row per domain, with status, word count, sentence count, Flesch score, Flesch–Kincaid grade, average sentence length, the share of sentences over 25 words, and the reason any page was excluded.
Suggested citation: BrandPromotions, “UK homepage readability: 300 crawled, 72 scored”, 24 October 2026. https://brandpromotions.shop/articles/uk-website-readability-study-2026/
