Text typed in the Preeti font (प्रीति फन्ट) only looks like Nepali. What is actually stored in the file is ordinary Latin characters, which the font draws as Devanagari shapes. To Google, a screen reader, a translator or an AI model, the Preeti word for "Nepal" is not नेपाल. It is g]kfn. Unicode (युनिकोड) stores the real Devanagari characters, so the same word can be searched, read aloud, copied and quoted. If you want Nepali text on the web to be found, it has to be Unicode.
Most Nepali websites have already made that move. Many Nepali documents have not. On 5 November 2026 I checked the fonts inside notice PDFs linked from Nepali government homepages. Among the PDFs that contain Nepali text in a font at all, 15 of 21 still use Preeti or another legacy Nepali font. The details are below.
What exactly is Preeti?
Preeti is a "font-encoded" or ASCII Nepali font, one of a family that includes Kantipur, Himali, Ganesh and Sagarmatha. It was made for typing Nepali before operating systems handled Devanagari properly. It is commonly credited to Ajay Mishra around 1993, though other sources name a different designer and year, and I have not found a primary record that settles it.
The trick it uses is simple. Each Latin key is redrawn as a Devanagari shape. The mapping below is taken from two independent open-source converters, npttf2utf and Shuvayatra's preeti, which agree on every entry:
| Key stored in the file | Shape Preeti draws | Unicode character |
|---|---|---|
g | न | U+0928 |
s | क | U+0915 |
k | प | U+092A |
d | म | U+092E |
; | स | U+0938 |
/ | र | U+0930 |
c | अ | U+0905 |
f | ा | U+093E |
Whole words follow the same rule. In the converters' test files, g]kfn becomes नेपाल and lxdfn becomes हिमाल. There is also a structural difference: in Preeti the vowel sign ि is typed before the consonant, as it appears visually, while Unicode stores it after. That is why converting Preeti to Unicode is not a simple letter swap, and why cheap converters sometimes mangle words.
Unicode, by contrast, gives every Devanagari letter its own number in the block U+0900 to U+097F, defined by the Unicode Consortium. Any font that supports Devanagari, such as Kalimati, Mangal or Noto Sans Devanagari, can display those numbers, and every program that handles text knows what they mean.
Why Preeti text is invisible to search engines
Search engines index characters, not pictures of characters. When someone in Kathmandu searches for नेपाल, the search engine looks for the Unicode characters न, े, प, ा, ल. A page or PDF written in Preeti contains g]kfn, so it does not match. To the index it is a short run of Latin letters and punctuation with no meaning in any language.
Google has seen this problem at scale before. In a 2018 talk at the Unicode Conference, Google engineers described how Myanmar's Zawgyi font encoding made content hard "to type, search, sort", because what users typed did not match what Google indexed. Zawgyi worked differently from Preeti, reusing Myanmar code points rather than Latin ones, but the lesson is the same: a font that fakes its characters breaks search. A W3C staff member made the same point about Indian and other font hacks as early as 2006.
Three other things break for the same reason:
- Screen readers read the Latin letters aloud, so a blind reader hears nonsense. The Kathmandu Post reported in April 2026 that a ministry published a draft disability-law amendment in Preeti files that screen-reader software could not interpret. The people the law concerned could not read the draft.
- Copying and quoting fails. Paste a Preeti sentence into a message or a search box and it turns into Latin gibberish.
- AI systems receive the same Latin string. No AI company documents special handling for Preeti, and there is no reason to expect a language model to recognise
g]kfnas Nepal. A notice an assistant cannot read is a notice it cannot cite or summarise.
How common is Preeti in 2026?
On websites: mostly gone
In the 34 Nepali websites I have audited since August, no homepage declared Preeti or any other legacy Nepali font in its HTML or stylesheets. The three Nepali-language news sites in the sample serve between 10,000 and 18,000 Devanagari characters on their homepages, all of it Unicode. Nepali news publishing has moved to Unicode.
In government PDFs: still common
Documents are a different story, so I checked them. In the 3 November census of Nepali government websites, 376 organisations' homepages answered from outside Nepal. On 5 November I fetched each of those homepages again. Where the homepage linked a PDF on the same site, I downloaded the first one and read the names of the fonts embedded in it.
- 78homepages linked a PDF; 66 downloaded as valid PDFs
- 22(a third) contain no fonts at all: scanned images with no text layer
- 21contain Nepali text set in a font
- 15of those 21 use Preeti or another legacy font
- 6use only Unicode Devanagari fonts
Of the 15, four use legacy fonts only. Eleven mix legacy and Unicode fonts in the same document: a Kalimati heading above a Preeti body, for example, which is typical of a letter assembled from older templates. The legacy fonts I found were Preeti, Kantipur, Fontasy Himali, Himal, Himchuli, Ganesh, Aakriti, Nagarik and CV Sadhana. The Unicode ones were Kalimati, Mangal, Kokila, Nirmala UI, Utsaah and Adobe Devanagari. The remaining 23 PDFs used only Latin fonts, which usually means an English document or a scan with an English text layer.
Put plainly: of the 44 notices that carry any text at all, only six are Nepali text a search engine can fully read. The other Nepali-language notices are either pictures of text or text in a font that stores Latin letters.
This is one PDF per organisation, the first linked from its homepage, so it says nothing about each organisation's full archive. It is mostly federal bodies, because local government homepages rarely link PDFs directly: only two local governments are in it. A font appearing in a PDF means some of its text uses that font, not all. I measured from a server outside Nepal, and 704 organisations did not answer at all. No organisation or document is named; the totals are published as JSON.
Does the government require Unicode?
According to the Kathmandu Post, yes. Its April 2026 report says the government's 2021 directive on building and managing government office websites (सरकारी कार्यालयको वेबसाईट निर्माण तथा व्यवस्थापन सम्बन्धी निर्देशिका, २०७८) requires websites to use Unicode text, and that the rule is applied inconsistently. I could not open the directive's own text to confirm the clause, so treat that as reported rather than verified. The directive is about websites. The notices those websites carry, typed in Word and exported as PDF, are where Preeti survives.
Where did Unicode Nepali come from?
The groundwork was laid two decades ago. A 2003 Nepali Times report describes a font standardisation project run by Madan Puraskar Pustakalaya with UNDP and the Ministry of Science and Technology, which produced Unicode Nepali keyboard input in two versions, a traditional layout and a romanised one. The Kalimati font, developed by Sanir Karmacharya, is distributed free through the same Nepali Unicode programme. The romanised keyboard is still available through tools such as Keyman.
The reason Preeti survives is muscle memory, not technology. People who learnt to type on the Preeti layout type quickly on it, and office templates were built in it. Both are solvable: Unicode keyboards exist in romanised and traditional layouts, and a template only has to be converted once.
How to check whether your Nepali text is Preeti or Unicode
- Copy a sentence and paste it into a search box or a plain-text editor such as Notepad. If it stays Nepali, it is Unicode. If it turns into Latin letters and symbols, it is a legacy font.
- For a web page, search Google for an exact phrase from it in quotation marks. If the page is indexed but the Nepali phrase finds nothing, suspect a legacy font or text inside an image.
- For a PDF, open its document properties (Ctrl+D in most readers) and look at the Fonts tab. Preeti, Kantipur, Himali, Ganesh and Sagarmatha are legacy. Kalimati, Mangal, Kokila, Nirmala UI and Noto Sans Devanagari are Unicode.
- Try selecting text in the PDF. If nothing selects, it is a scanned image with no text layer.
How to fix it
- For new documents: type in Unicode from the start. Install a Nepali Unicode keyboard (Windows includes a Nepali layout, and the Madan Puraskar romanised and traditional layouts are free) and use Kalimati, Mangal or Noto Sans Devanagari.
- For existing Preeti documents: convert with a Preeti-to-Unicode converter, then proofread the ि placement and conjuncts, which is where converters usually go wrong. Convert the template, not every letter.
- For scanned notices: publish the text as well as the scan. A short Unicode summary on the web page, saying what the notice is, who it is for and the deadline, makes it findable even if the PDF is an image.
- For websites: check that the Nepali text is real text and not an image, and that the page declares
lang="ne". Only one of 34 sites in my August audit declared Nepali.
This matters beyond search. A notice about scholarships, tenders or a public hearing that cannot be found, read aloud or quoted reaches fewer of the people it is for. That gap is widest for Nepalis who are blind or partially sighted, and for those working abroad, who search rather than visit a notice board. I have written separately about how Nepali-language search is underserved and how 23 of 34 Nepali websites contain no Nepali at all. Preeti is the same problem in a different form.
Questions people ask
Is Preeti the same as Nepali Unicode?
No. Preeti stores Latin characters and draws them as Devanagari. Unicode stores the Devanagari characters themselves. They can look identical on screen and be completely different underneath.
Can Google read Preeti text?
It can read the Latin characters, which is the problem. It cannot know that g]kfn means नेपाल, so the text does not match anyone's Nepali search.
Which font should I use for Nepali on a website?
Any Unicode Devanagari font. Kalimati and Mangal are common on Windows. Noto Sans Devanagari and Mukta are free web fonts. The font matters less than the encoding: if the text is Unicode, it is searchable whatever font displays it.
Does converting Preeti to Unicode change how the text looks?
Only the typeface, which you choose. The words stay the same if the conversion is proofread.
Is Preeti still used for government exams?
I found conflicting, unofficial claims about Public Service Commission typing tests and no official source, so I am not stating either way.
Sources
- Unicode Consortium, Blocks.txt (Devanagari, U+0900 to U+097F).
- casualsnek, npttf2utf character map, and Shuvayatra, preeti converter test vectors, GitHub.
- Carr, Hoffman and Swartz (Google), Fixing Burmese: Dealing with Zawgyi, Unicode Conference, 12 September 2018.
- Chris Lilley (W3C), www-style mailing list, 28 April 2006.
- Sajana Baral, Government invites feedback on disability law online, but target group is locked out, The Kathmandu Post, 30 April 2026.
- Vibek Raj Maurya, Fontastic Nepali, Nepali Times, 10 October 2003.
- Keyman, Nepali Romanized keyboard (Madan Puraskar Pustakalaya).
- Law Commission of Nepal, government office website directive, 2078 (listing).
Method: on 5 November 2026, one homepage request to each of the 376 government organisations that answered in the 3 November census, then one request for the first same-site PDF linked, if any. Fonts were read from the PDF's font dictionaries. Nothing else was requested.
Featured photo: Naaranga, Wikimedia Commons, CC0. Cropped to 16:9.
.png){kind=link}