Disclosure: I built the plugin reviewed here. I am writing this as the developer, not as a neutral third party. Everything below is testable on your own site in about two minutes, and I have been deliberate about listing what the plugin does not do yet.
The problem, in one number
I run a Hindi personal-finance site. A few days ago I opened a published article — 956 words, properly structured, images, headings, internal links — and looked at the Rank Math panel in the editor.
It said 0 / 100.
Not 40. Not 65. Zero. The same article, translated into English, would have scored somewhere in the 70s. Nothing was wrong with the article. Something was wrong with the tool measuring it.
If you write in Hindi, Marathi, Nepali, or any language written in Devanagari, you have almost certainly seen this and quietly assumed your content was the problem. It usually isn’t.
Why English SEO plugins fail on Devanagari
This is not any one plugin being lazy — the major English SEO tools share the same blind spot. It comes down to three assumptions baked into how they count text.
1. Word counting breaks
PHP’s built-in str_word_count() — the function most word counters lean on — only recognises the Latin alphabet. Hand it a Devanagari sentence and it returns nonsense, because it has no idea that क, ख, or ज्ञ are letters. Some plugins fall back to splitting on spaces, which is closer but still wrong once you hit conjuncts and matras.
2. The sentence never ends
Hindi does not end sentences with a full stop. It ends them with the danda — the vertical bar । (Unicode U+0964). English SEO plugins look for ., ? and !. They find none, conclude the entire article is one enormous run-on sentence, and hand you a catastrophic readability score.
This single character is responsible for most of the damage.
3. Keyword matching is too literal
Hindi is an inflected language. Search “निवेश” and the honest answer should include “निवेश करना”, “निवेश करने”, “निवेशक”. A plain string match catches none of them, so your keyword density reads as far lower than it actually is.
Put all three together and you get a score of zero on a perfectly good article.
What Hindi SEO Assistant does
Rather than fight the existing plugins, I wrote a small one that sits alongside them and measures the things they get wrong. Version 0.1.0 adds a panel to the WordPress editor showing five numbers:
- Word count — a counter written specifically for Devanagari, handling matras and conjuncts correctly rather than guessing.
- Sentence count — recognises the danda (।), double danda (॥), plus
?and!. - Average sentence length — the single most useful readability signal in Hindi. Anything above roughly 22 words per sentence gets hard to read.
- Reading time — calculated at 200 words per minute, which is closer to actual Hindi reading speed than the 225 wpm most English plugins assume.
- Language mix — what percentage of the article is actually Devanagari versus Latin script. Useful if you write Hinglish and want to know how far you have drifted.
Everything updates live as you type. It works in both Gutenberg and the Classic Editor, with no build step and no external requests.
Testing it on a real article
I ran it on the same 956-word article that had scored zero. The panel returned:
- 956 words
- 75 sentences
- 12.7 words per sentence
- 5 minutes reading time
- 73% Devanagari
Check the arithmetic yourself: 956 ÷ 75 = 12.7, and 956 ÷ 200 = 5 minutes. The numbers hold up. And 12.7 words per sentence is genuinely good — comfortably inside the readable range for Hindi prose.
So the article was fine all along. The measurement was broken.
A few things I made sure it handles
Text analysis is full of small traps, and most of the work went into these rather than the headline features:
- Shortcodes like
[my_calculator]are stripped before counting, so widgets don’t inflate your word count. - Anything inside
<script>and<style>tags is excluded. - Gutenberg block comments are removed rather than counted as content.
- Decimals survive —
7.1%stays one token and does not get split into a new sentence. - Abbreviations like
U.S.A.no longer break a sentence in three. - Hyphenated compounds such as
देश-विदेशcount as one word, which is what a reader would say.
What it does not do yet
This is version 0.1.0, released after one day of work, and I would rather be straight about the gaps:
- No keyword density analysis. This is the big one and it is next. Doing it properly means stripping Hindi suffixes so that “निवेश” matches “निवेश करने” — real work, not a regex.
- No overall score. It reports measurements, not a verdict. A score without a trustworthy readability model would just be a prettier version of the problem I am complaining about.
- No title or meta description analysis. Planned.
- No integration with existing SEO plugins. Right now it sits beside them rather than correcting them.
- Not yet on the WordPress.org repository. Submission is in progress.
If you need a full SEO suite, keep the one you have. This is a narrow instrument that measures five things English tools measure wrongly.
Who this is for
Useful if you publish in Hindi or another Devanagari language on WordPress and want word counts and readability figures you can actually trust. If you write only in English, there is nothing here for you — the existing plugins already do this well, which is rather the point.
Verdict
An SEO score of 0/100 on a well-written 956-word article is not feedback. It is a tool admitting it cannot read the language.
Hindi SEO Assistant 0.1.0 does not fix your SEO. It gives you five numbers that are correct, which is more than the alternatives currently manage for Devanagari. The keyword analyser lands next, and that is where it starts becoming genuinely useful rather than merely honest.
If you write in Hindi on WordPress, try it on your worst-scoring article. You may find the article was never the problem.
Rank Math and Yoast SEO are trademarks of their respective owners. This site and the plugin described here are independent projects, and are not affiliated with, endorsed by, or sponsored by either company. Product names are used only to describe and compare real, observed behaviour.