Nepal Website Health Report 2026: what we found in 2,339 Nepali business websites

I audited 2,339 Nepali business websites to answer one question: can Google and AI assistants like ChatGPT actually read them? The average site scored 73 out of 100, and the shape of the result is the story. Almost no sites block AI crawlers (1%), yet 63% give those crawlers no structured data and 92% have no llms.txt. Nepali sites aren't shutting AI out. They are giving it very little to work with. This is what I found, how I found it, and what to do about it.

The headline numbers

How Nepali business sites scored

Each site gets a score out of 100 from the same checks as the free AI visibility checker. 54% scored 70 or above (grade B or better), while 13% scored below 55, the range where AI assistants are likely to struggle with the site.

• Grade A (85 to 100)32%
• Grade B (70 to 84)23%
• Grade C (55 to 69)33%
• Grade D (40 to 54)8%
• Grade F (below 40)5%

Where the points are lost

Averaged across sites, here is how well they do in each group of checks. AI crawler access scores 99%, because almost nobody blocks the crawlers. Structured data scores 33%, because most sites never describe themselves to a machine. In plain words: the door is open and the room is empty.

• AI crawler access99%
• Readable content77%
• Structured data33%
• Search basics72%
• Discovery files66%

The most common problems

Share of audited sites with each problem. Several of these are cheap fixes that take an afternoon.

• No llms.txt92%
• No profile links or contact details in structured data90%
• Missing or badly sized meta description67%
• No business or person named in structured data65%
• No structured data (JSON-LD)63%
• Missing or repeated main heading (h1)58%
• No share preview (Open Graph)54%
• No canonical address40%
• No sitemap found32%
• Not fully on HTTPS24%

The AI crawler question

Good news first. Only 1% of sites block the crawlers that find pages for AI answers (OpenAI's search crawler, ChatGPT's page reader, Perplexity, Claude's search crawler), 1% block a training-only crawler, which is a legitimate choice, and 0% block Google or Bing. For the large majority, robots.txt is not the problem.

One group is different: 37 sites (1% of those that answered) refused our checker outright at the firewall, with an error or a bot-protection page. Some firewalls and CDN modes turn away every crawler that isn't a normal browser, and that can include AI assistants. If that is you, it is worth checking. This article explains how it happens.

What Nepali sites are built on

WordPress is the most common setup (39% of sites, average score 79). Among the named platforms, Blogger sites averaged the highest score (82) and React or Vue app the lowest (56). That is a pattern in this sample, not a verdict on the platform: how a site is set up matters more than what it is built on.

What this means if you own a Nepali business

• 01Describe yourself to machines. 63% have no structured data and 90% don't put their profile links or contact details in it. This is the biggest gap, and the cheapest to close.
• 02Fix the search basics. 44% have no meta description, 40% have no main heading and 40% have no canonical address.
• 03Add the cheap files. 32% have no sitemap and 92% have no llms.txt.
• 04Make sure the text is in the page. 13% show almost nothing without JavaScript, and a further 5% show very little. A crawler that doesn't run scripts sees an almost empty page.
• 05Check your firewall once. It only takes a minute, and 37 sites in this sample were turning crawlers away.

I can fix these for you, or you can run the checker on your own site first and see your score. The average to beat is 73.

How I did it, and what it can't tell you

• 01The sample. 2,648 sites under the .com.np domain that Common Crawl, a free public archive of the web, fetched successfully in its September 2026 crawl. Sites on plain .com, .org or .net are not included. This first edition covers domain names from 0-9 and A to P, because the archive's index stopped answering before the later part of the alphabet could be collected; I will add the rest in an update. Names are not related to quality, so the results should still be representative, but treat the figures as a large sample, not a full census.
• 02The audit. In October 2026 each homepage was requested once, along with robots.txt, llms.txt and the sitemap, and read the way an AI crawler reads it: the raw HTML, without running JavaScript.
• 03The result. 2,518 sites answered. 2,339 gave a readable business homepage; 72 were parked, default-server or coming-soon pages and are left out; 107 refused or returned an error; 130 did not answer at all (offline or gone).
• 04The margin. For the shares quoted, about plus or minus 2 percentage points at 95% confidence, for .com.np sites like these.
• 05The limits. The score measures readiness, not rankings. A site that blocks crawlers at the firewall may still work in a normal browser. Crawler rules change often. And this is a sample of .com.np sites, not every Nepali business.
• 06Privacy. No website is named in the report or the data. Only totals are published.

Use these numbers

Frequently asked questions

Related

• Free tool: can ChatGPT read your website?
• Your Nepali business is probably invisible to ChatGPT
• Most Nepali websites have no schema markup
• SEO services in Nepal

Anurag Jha

SEO expert and AI SEO & GEO consultant in Kathmandu, and Head of Marketing at Daewoo Nepal, where he built the brand's organic search from zero. Works with businesses in Nepal and, remotely, in Australia and the US.

About me Free AI checker