WordPress word count and translation volume calculator.

How are Chinese, Japanese, Korean and Thai words counted?

Chinese and Japanese are counted one character as one word, and Thai, Lao, Khmer and Burmese are split into words with a dictionary. Splitting on spaces fails for all of them, because they are written without spaces between words. Korean, Arabic, Hebrew, Hindi and Russian put spaces between words and are counted like English.

“reliable detection of word boundaries in languages such as Thai, Lao, Chinese, or Japanese requires the use of dictionary lookup or other mechanisms”

• Chinese and Japanese: each Han character, hiragana and katakana counts as 1. A page that mixes Japanese with English product names gets both rules at once.
• Thai, Lao, Khmer, Burmese: words come from the browser and server's built-in Intl.Segmenter, which MDN documents for text that “does not use whitespaces between words”.
• Character count: shown beside every word count, because machine translation APIs bill by the character.

WordPress's own editor counts Chinese and Japanese by the character too. Its count type is a setting inside each language pack, read by wp_get_word_count_type(), and the packs on translate.wordpress.org set it like this:

Translation vendors bill these scripts by the character too. Gengo prices Japanese, Chinese and Korean sources per character, and WPML converts with total characters x 0.55, so 1,000,000 Chinese characters are billed as 550,000 words.

Benchmarks used, with sources

A professional translator delivers 2,000 to 3,000 words a day, post-editing machine translation runs at 5,600 to 7,000 words a day, and adults read 238 words a minute in silence. Those are the benchmarks behind the calculator. Every time and cost figure in it is a published number, opened and checked on October 8, 2026. Where a language has no published figure, the calculator shows none for it.

Machine translation prices are the pay-as-you-go list rates in US dollars, before the monthly allowances each vendor gives (Google's first 500,000 characters a month, for example). Google also bills the markup: it counts <p>こんにちは</p> as 12 characters, not 5. This tool counts visible characters only.

What if the WordPress REST API is closed?

If a site hides its REST API, you can still get the number four ways: a sitemap sample, an export file, a WP-CLI command or one SQL query. Security plugins and some hosts close the API to visitors, and the WordPress handbook documents how: sites “can require authentication for all REST API requests” with the rest_authentication_errors filter. The counter detects this and tells you which case it found.

1. Check whether the API is open

Run this from any terminal. A line such as x-wp-total: 137 means the API is open and the site has 137 published posts.

curl -sI "https://yoursite.com/wp-json/wp/v2/posts?per_page=1" | grep -i "x-wp-total"

2. Estimate from the sitemap

The counter reads /wp-sitemap.xml (in core since WordPress 5.5) or a Yoast-style /sitemap_index.xml, opens a spread sample of 30 pages per sitemap, and counts each page's main content area. It is an estimate, labelled as one, and it only reads pages that robots.txt allows.

3. Upload a WordPress export file

In WP admin, go to Tools, then Export, choose All content and download the file. Drop it on the Export file tab above. The file is parsed in your browser, drafts and private items are left out, and nothing is uploaded.

4. Count on the server with WP-CLI or SQL

With shell access, this WP-CLI command counts published posts and pages with WordPress's own tag and shortcode stripping:

wp eval '
$total = 0;
$ids = get_posts(["post_type" => ["post", "page"], "post_status" => "publish", "numberposts" => -1, "fields" => "ids"]);
foreach ($ids as $id) {
  $post = get_post($id);
  $text = wp_strip_all_tags(strip_shortcodes($post->post_title . " " . $post->post_content));
  $total += count(preg_split("/\\s+/u", trim($text), -1, PREG_SPLIT_NO_EMPTY));
}
echo count($ids) . " items, " . $total . " words" . PHP_EOL;
'

With database access only, this query gives the count per post type. It needs MySQL 8.0 or MariaDB 10.0.5 or later for REGEXP_REPLACE. Change wp_posts if your table prefix is different.

SELECT post_type,
       COUNT(*) AS items,
       SUM(IF(txt = '', 0, LENGTH(txt) - LENGTH(REPLACE(txt, ' ', '')) + 1)) AS words
FROM (
  SELECT post_type,
         TRIM(REGEXP_REPLACE(
           REGEXP_REPLACE(CONCAT(post_title, ' ', post_content), '<[^>]+>', ' '),
           '[[:space:]]+', ' ')) AS txt
  FROM wp_posts
  WHERE post_status = 'publish'
    AND post_type IN ('post', 'page')
) t
GROUP BY post_type;

Both read the raw post content, so block comments are stripped as tags but shortcode output and page-builder text stored outside the post body are not included. Expect them to land a little under the REST figure on sites that lean on shortcodes.

How the counter treats the sites it reads

The counter makes plain, identified requests for public pages and nothing else. A WordPress site does not let other websites read its API from a browser, so the requests go through our server, which follows these rules:

• It says who it is. Every request carries the user agent HydoriWordCount/1.0 (+https://hydori.io/tools/wordpress-word-count; counts public words at a visitor's request).
• It follows robots.txt. A rule that names the tool is always obeyed. A general rule against the API path stops the count until you confirm the site is yours or you have permission. The sitemap route never overrides robots.txt.
• It goes slowly. At most 2 requests at a time from your browser, 2,400 requests an hour to any one site across all visitors, and 1,500 an hour per visitor. If a site answers that it is getting too many requests (HTTP 429), the count pauses for as long as the site asks and then carries on.
• It says what it could not check. If a content type does not answer, or the site has more than 14 of them, the result is labelled a partial count and names what is missing.
• It keeps nothing. Each batch is counted and discarded. Only the counts, public URLs and titles come back to your browser.
• It only reads the public web. Private, local and internal network addresses are refused, on the first request and on every redirect.

Site owners who want it kept out can add User-agent: HydoriWordCount and Disallow: / to robots.txt.

Sources

1. WordPress REST API handbook: Pagination (per_page cap, X-WP-Total and X-WP-TotalPages headers)
2. WordPress REST API reference: Post Types (the /wp/v2/types route)
3. WordPress REST API handbook: FAQ (requiring authentication for all requests)
4. wp_get_word_count_type() (count type per language pack)
5. WordPress 5.5 XML sitemaps (/wp-sitemap.xml)
6. Unicode Standard Annex #29: Text Segmentation
7. MDN: Intl.Segmenter
8. Brysbaert, M. (2019). How many words do we read per minute? Journal of Memory and Language, 109, 104047. Open preprint
9. Slator: How Fast Can You Post-Edit Machine Translation? (December 12, 2022)
10. memoQ: machine translation vs post-editing
11. Translated: translation rates
12. Toral, Wieling and Way (2018). Post-editing effort of a novel with statistical and neural machine translation. Frontiers in Digital Humanities
13. Gengo: pricing by language
14. Google Cloud Translation pricing, Amazon Translate pricing, Azure AI Translator pricing
15. Weglot help: total number of translated words and Weglot pricing
16. WinningWP: Weglot Review by Kevin Muldoon (updated July 13, 2024)

Written and tested by

Junaid Khalid

Multilingual SEO Specialist. Runs SEO for 14 sites, with 3,600+ articles under his byline and 50M+ organic impressions. He ran this counter against 15 live sites, his own blog included, before publishing it. Last updated October 8, 2026.