Type "digital marketing Nepal" into a keyword tool & it will tell you the volume is 320 a month. Type "digital marketing agency ma kaam garne tarika" & it will tell you the volume is zero.
One of those numbers is roughly true. The other is a limitation of the tool, not a fact about Nepal. & if you build your entire content plan on what the tools can see, you will spend a year competing for a handful of broad terms while ignoring the specific, high-intent searches that actually turn into enquiries.
Keyword research in a small market works differently. Here is how to do it properly.
What keyword research actually is
It is demand mapping. You are trying to establish three things: what people want, in what words, & how badly. Everything else is admin.
The common failure is treating it as list production. Somebody exports 2,000 keywords, sorts by volume, hands over a spreadsheet & calls it a strategy. Nobody ever builds anything from it, because a list of phrases is not a plan. A plan is a set of pages, each one answering a specific thing a specific person wants, in the order you are going to build them.
So the output you are working towards is not a keyword list. It is a page list.
The five things that make Nepal different
1. The volume data is mostly missing
Global tools round small numbers down. Anything under roughly 50 monthly searches shows as zero, "10" or a range. In a market of Nepal's size, that describes the majority of commercial searches. A term showing zero volume can still bring you three enquiries a month, & three enquiries a month is a business.
Use volume data for relative comparison — is A bigger than B — & never as an absolute count.
2. People search in romanised Nepali
Most Nepalis type on an English keyboard. So the search is not किन्ने or "buy", it is a hybrid: "sasto laptop kathmandu", "ghar jagga sale lalitpur", "best momo near me". Keyword tools handle this badly, & most competitors ignore it entirely, which is precisely why it is worth attention.
Check Search Console for romanised queries you are already receiving. They are usually there, & usually nobody has written a page for them.
3. Code-switching within a single query
English nouns, Nepali connectors, English brand names. "SEO ko price in Nepal" is a real pattern. You cannot brainstorm these from an office; you find them in your own query data, in autocomplete, & by listening to how customers phrase things on the phone.
4. Discovery starts on Facebook & TikTok
A meaningful share of Nepali product discovery does not begin on Google at all. It begins in a Facebook group or a TikTok video, & Google gets used later to verify the business exists. That changes what you research: brand-plus-modifier searches ("[brand] review", "[brand] price", "[brand] contact") matter more here than in markets where Google owns the whole journey. It is also why TikTok search deserves its own keyword work.
5. "Near me" is doing heavy lifting
Nepali address conventions are loose, so people search by landmark & proximity rather than street. "near Baneshwor", "opposite Bhatbhateni", "near me". This is local SEO territory rather than content territory — the fix is a properly maintained profile, covered in Google Business Profile for Kathmandu businesses.
Building the seed list
Before any tool, spend an hour collecting raw language from places where real demand is recorded.
- Your sales conversations. The exact phrasing customers use on calls & on WhatsApp. This is the highest-quality source you own & almost nobody mines it.
- Search Console. Performance report, Queries tab, last 12 months, no filters. Everything here is demand you already receive. Sort by impressions with low clicks to find pages that are close but not good enough.
- Google Business Profile. The searches report shows the literal local phrases people used before finding you.
- Competitor pages. Not their meta keywords — their page titles & H1s. Those reveal which intents they decided were worth a page.
- Your own support inbox. Every repeated question is a page waiting to be written.
You should finish this step with thirty to eighty phrases in real human language. That is the seed.
The tools worth using, & what each is good for
You do not need an expensive stack to do this well in Nepal.
- Google Search Console. Free, & the only source that reports your actual queries rather than an estimate. Start here every time.
- Google Keyword Planner. Free with an active Ads account. Ranges rather than exact numbers unless you are spending, but the relative ordering is reliable & you can filter by Nepal specifically.
- Autocomplete & People Also Ask. The cheapest source of real phrasing that exists. Type your seed term, note every suggestion, then repeat with "how", "why", "best", "price" appended. Set the location to Nepal or the results will mislead you.
- Bing Webmaster Tools. Includes a free keyword research tool with data that sometimes fills gaps Google leaves.
- Paid tools — Ahrefs, Semrush. Genuinely useful for competitor gap analysis & link data. Treat their Nepali volume figures with the same suspicion as everything else.
One warning: no tool understands Nepali market context. Every list needs a human pass to remove terms that are technically relevant & commercially pointless.
Sorting by intent, which matters more than volume
Every query carries an intent. Get this wrong & you will write a beautiful guide for a term where Google only ranks product pages, & you will never rank at all.
- Informational. "what is off page seo". They want to understand. Serve a guide or an article.
- Commercial investigation. "best seo agency in kathmandu", "seo price in nepal". They are comparing before buying. Serve a comparison, a pricing page, or a genuinely useful evaluation guide. This is the most under-served intent in Nepal & the most profitable.
- Transactional. "seo services kathmandu". They are ready. Serve a service page, not a blog post.
- Navigational. "[brand] contact". They want a specific place. Make sure it is easy to find.
The reliable way to classify intent is to search the term yourself with the location set to Nepal & look at what Google is already rewarding. If page one is nine product pages, that is the format required. Google has already told you the answer; the only mistake is arguing with it.
Clustering: from phrases to pages
Group your terms by the answer they need, not by the words they share. "seo price nepal", "how much does seo cost", "seo package rate kathmandu" & "is seo expensive" are four phrasings of one question & belong on one strong page.
A rough method that works without any software:
- Put every term in a spreadsheet with a column for intent.
- For the top twenty terms, note the top three ranking URLs.
- Where the same URLs rank for several of your terms, those terms belong together — Google is telling you it considers them one topic.
- Name each cluster after the page you will build, in plain English.
- Decide which cluster is the pillar & which are supporting pages, then link them deliberately. The mechanics are in topical authority & content clusters.
One page per intent. Two pages targeting the same intent compete with each other, split their signals & both rank worse — the problem covered in content pruning.
Prioritising when everything looks equally attractive
Score each cluster out of five on four things: commercial value, realistic difficulty, how much of it you can already answer credibly, & how quickly you could ship it.
The pattern that emerges in Nepal is consistent. The head terms — short, broad, high volume — are usually held by established sites & take a year or more of authority building. The specific commercial-investigation terms are frequently unclaimed, because everyone chased volume instead.
So start where intent is highest & difficulty is lowest. Win those, build authority, then move up. Attacking the head term first is the single most common way a content budget gets spent with nothing to show for it.
Turning research into a brief
A keyword without a brief becomes a vague article. Each page in your plan should carry five lines before anyone writes a word: the primary intent in a sentence, the specific question the page must resolve in its first paragraph, the subtopics competitors cover, the internal links in & out, & the single action you want the reader to take.
That last one is where most content quietly fails. A page ranking for a commercial term with no clear next step is a page that generates traffic & no revenue.
Keyword research when the answer comes from an AI
Queries typed into ChatGPT, Perplexity & Google's AI Overviews are longer, more conversational & framed as complete questions. "which seo agency in kathmandu is good for a small hotel" rather than "seo kathmandu".
This does not invalidate the work, it extends it. Collect the question forms, not just the noun phrases, & make sure each one is answered directly in a self-contained passage that a machine can lift. The approach is set out in structuring content so machines can read it & the difference between the two disciplines in AEO & GEO explained.
The mistakes that cost the most
- Chasing volume over intent. Traffic that never buys is a cost, not an asset.
- Trusting Nepali volume figures literally. They are directional at best.
- Ignoring what Search Console already tells you. Free, specific, & about you rather than a market average.
- Writing an article for a transactional query. Check the results page first, every time.
- One page per keyword. Produces thin, cannibalising pages that all underperform.
- Doing it once. Language changes, competitors publish, & your own query data grows every month.
Frequently asked questions
What is keyword research, in plain terms?
Working out the words people actually type when they want what you sell, & how much demand sits behind each one. It is demand mapping, not list building. The output should be a set of pages you are going to build, not a spreadsheet of 900 phrases nobody will ever use.
Why do keyword tools show zero volume for Nepali searches?
Because Nepal is a small market & most tools round anything under roughly 50 monthly searches down to zero or "10". The searches are happening; the tool simply cannot see them reliably. Treat Nepali volume data as a rough ranking of relative popularity, never as a literal count.
Should I target Nepali or English keywords?
For most commercial searches in Kathmandu, English & romanised Nepali dominate, because people type on English keyboards. Devanagari matters more for news, government, health & religious topics, & for older or rural audiences. Check Search Console for what you already receive before assuming.
Which free keyword research tools work in Nepal?
Google Search Console for what you already rank for, Google Keyword Planner for relative demand, Google autocomplete & People Also Ask for phrasing, Bing Webmaster Tools for a second data source, & your own Google Business Profile search queries report for local intent. Together they cost nothing & beat any single paid tool for a Nepali market.
How many keywords should one page target?
One intent, not one keyword. A single well-written page will naturally rank for dozens of variations of the same question. Build pages around what the searcher wants, & split them only when the underlying intent genuinely differs.
Are low-volume keywords worth targeting?
In Nepal, often they are the whole game. A phrase with 20 searches a month & clear buying intent is worth more than one with 2,000 searches & none. Ranking for ten of those beats losing to a national brand on one broad head term.
How often should I redo keyword research?
A full review once or twice a year, plus a monthly fifteen-minute look at Search Console queries. The monthly check is where the real finds are, because it shows you the questions people are already asking you that you never thought to answer.
Does keyword research still matter with AI search?
Yes, but the unit changed. Queries into ChatGPT, Perplexity & AI Overviews are longer & phrased as full questions. So research the questions, not the two-word phrases, & answer each one cleanly enough to be quoted. The demand research is the same work; the output format is different.
Where to start
Open Search Console, set the range to the last twelve months, sort your queries by impressions, & look for the terms where you appear often but are rarely clicked. Those are searches Google already associates with you & where you are not yet good enough. That list is the fastest keyword research you will ever do, & it is free.
Once you have the clusters, the next question is what to build first. Our content marketing work starts exactly here, & bottom of funnel content covers which pages to write when you need enquiries rather than readers. If you would like us to run the research against your site & send back the page list, get in touch.