To estimate an article’s reading time before it goes out, count the words in the body text only, divide by a reading speed that suits your audience, and round the result up to a whole minute for the label. A 1,850-word feature at 238 words per minute comes to 7 min 46 sec, labelled “8 min read”. The steps below turn that into a routine you can repeat on every piece, using one draft as a running example.

Step 1: Get a clean word count

The estimate is only as good as the count, and the commonest error is counting text the reader will not read as part of the article.

Take the count from your draft or editor, not from the published page. Selecting everything on a live page drags in the menu, share buttons, related links, the author bio and the comment thread, and can inflate a count by hundreds of words.

Then decide what belongs. A workable default:

  • Include: body paragraphs, headings, subheadings, and list items.
  • Exclude: navigation, bylines and bios, pull quotes (they repeat body text, so counting them counts the same words twice), related-article blocks, comments, and alt text.
  • Decide once and stick to it: image captions, footnotes, sidebars and code samples. These are covered in step 4.

Our example draft has 1,850 words of body text and headings, 140 words of captions across six images, and 260 words of footnotes. The count we will use is 1,850.

You can paste the body text straight into the reading time calculator, which counts the words for you. If its figure differs from your editor’s by a handful, the cause is nearly always a difference in how dashes or hyphenated compounds are treated; how reading time is calculated lists the exact rules. A gap of five words in 1,850 changes the estimate by about a second.

Step 2: Choose a reading speed for the audience

Pick the speed for the people who will read the piece, not for yourself. You wrote it and know what it says, so timing your own read-through may underestimate the time a new reader needs.

Three starting points:

  • 238 WPM for a general adult audience and ordinary English non-fiction. This is the average from the meta-analysis by Brysbaert (2019) and the most defensible default.
  • 180 WPM as a planning figure when the piece is technical, heavy with new terms, or meant to be studied. Just & Carpenter (1980) found that readers slow down on infrequent words and where ideas have to be connected, which is exactly what specialist writing asks of them. See how text difficulty changes reading speed for more on judging this.
  • 260 WPM for fiction, the corresponding average in the same meta-analysis.

These figures describe English-speaking adults. If a large share of your audience reads English as a second language, a slower assumption is the cautious choice, though that is editorial judgement and not something the listed research quantifies.

For a publication, the speed should be a standing decision and not chosen per article. Changing it piece by piece makes labels impossible to compare.

Step 3: Calculate the exact time and the label

Divide words by speed: minutes = words ÷ WPM.

For the example at 238 WPM, 1,850 ÷ 238 = 7.77 minutes, which is 7 min 46 sec. Round up to the next whole minute for the label: “8 min read”.

Keep both numbers. The exact time is for you, because it responds to small edits. The label is for readers, and it moves only when the exact time crosses a minute boundary.

It is worth running the careful speed as well, to see the spread. At 180 WPM the same draft takes 10 min 17 sec, an “11 min read”. If your audience could plausibly be either, that three-minute gap tells you how much weight the single label can bear.

Step 4: Allow for what is not prose

A word count registers words. It sees nothing of the time spent studying a chart or working through a code sample, and it treats a footnote like any other sentence. There is no research-backed conversion for these elements among this site’s sources, so every adjustment here is a rule of thumb. Choose one, label it as a house rule, and apply it the same way every time.

Element Reasonable options Things to consider
Images and charts Ignore them, or add a fixed allowance per image A decorative photo costs a glance; a dense chart can cost a minute. One flat figure will be wrong for both, but consistently so
Captions Include in the word count, or exclude Include if captions carry information the body does not
Code blocks Exclude and say so, or count them at your careful speed Readers either skip code or study it; neither matches prose speed
Tables Count the cell text as words Comparing values across rows takes longer than the count implies
Footnotes and endnotes Exclude unless they are essential to the argument Most readers do not follow every note
Embedded video or audio State the runtime separately Folding a six-minute video into a “min read” label hides what the time is for

Applied to the example: suppose the house rule is ten seconds per image, an illustrative figure and not a measured one. Six images add one minute, taking the estimate from 7 min 46 sec to 8 min 46 sec and the label to “9 min read”.

Compare the alternative of counting everything. Body, captions and footnotes together are 2,250 words, which is 9 min 27 sec and a “10 min read”. The two approaches differ by a minute on the label. Neither is wrong, but mixing them across articles would be.

Step 5: Work backwards to set a length target

The same formula in reverse turns a time budget into a word count: words = minutes × WPM. This is useful at the commissioning stage, before anything is written.

How many words fit into a reading session
Time available180 WPM238 WPM300 WPM
5 min900 words1,190 words1,500 words
10 min1,800 words2,380 words3,000 words
15 min2,700 words3,570 words4,500 words
20 min3,600 words4,760 words6,000 words
30 min5,400 words7,140 words9,000 words
45 min8,100 words10,710 words13,500 words
60 min10,800 words14,280 words18,000 words

If a newsletter slot is meant to be a ten-minute read for a general audience, the table gives 2,380 words at 238 WPM. That is a ceiling. Subtract whatever allowance you make for images, and brief the writer for a little less.

Labels have exact limits you can calculate the same way. A label of N minutes at a given speed holds at most N × WPM words. At 238 WPM, a “7 min read” tops out at 1,666 words and an “8 min read” at 1,904. Our 1,850-word example sits comfortably inside the eight-minute band, 54 words below its upper edge.

If the piece will also be read aloud, as a podcast segment or a recorded version, set a separate target. Brysbaert (2019) gives 183 WPM as the adult average for oral reading, so the same 1,850 words runs to about ten minutes spoken. Reading time versus speaking time covers that conversion.

Step 6: Record the method in your style guide

A method that lives in one editor’s head will not survive a staff change or a new content system. Write it down. An entry can be as short as this:

Reading time. Calculated from body text and headings only. Captions, footnotes, pull quotes and code blocks are excluded. Speed: 238 WPM. Add ten seconds per image as a house allowance. Round up to the next whole minute; minimum one minute. No label on pieces under 300 words. Recalculate after any edit that changes the length.

Every choice in that entry could reasonably be different. What matters is that each one is stated, so two people working on two articles produce comparable labels, and so a reader who asks how the number was reached can be given a straight answer. The broader question of whether to display the label at all, and on which kinds of page, is taken up in should your blog show a “min read” label.

Step 7: Re-check after edits

Reading time is a property of the final text, not the first draft. Recalculate at the last stage before publishing, after cuts, additions and the sub-editor’s pass.

Small edits often leave the label alone. Cut 150 words from the example and it becomes 1,700 words: 7 min 9 sec, still an “8 min read”. To bring the label down to seven you would need to reach 1,666 words, a cut of 184.

Additions cross boundaries more easily than you might expect. Add a 300-word sidebar and count it, and the piece is 2,150 words: 9 min 2 sec. Those two seconds over nine minutes make it a “10 min read”.

If your content system calculates the label automatically, the re-check is a matter of confirming that it used the text you intended. Paste the final body into the calculator at your house speed and compare. A mismatch usually means the system is counting something it should not, such as an embedded widget or a block of related links.

What this estimate can and cannot tell you

The result is the time a reader at your assumed speed would need to read every word once. Real readers differ from the assumption, skip sections, re-read others and get interrupted. The label will be too long for some and too short for others.

That does not make it useless. Applied consistently, it lets readers compare one piece with another and plan accordingly, and it gives writers and editors a shared, checkable way to talk about length. Treat it as a planning figure with stated assumptions, and say “about” when you quote it.