Keyword Research for New Websites: Expert Tips & Strategies

My process for a new niche is a topical map first, a keyword list second. I start by mapping the entire content space into 5 to 7 high-level pillar topics (the core categories the niche cares about), then 3 to 5 sub-clusters under each pillar with one cornerstone post target and 3 to 8 supporting variations per sub-cluster, before pulling a single keyword.

Only then do I pull keyword volumes (free via Google Autocomplete and Brave Search, or GSC if the site has history) to validate which clusters have real demand and which are aspirational. The crucial aspect: keyword volume is a CONSEQUENCE of getting the topical map right, not the input that drives it. Sites built keyword-list-first end up with hundreds of thin posts that never compound; sites built topical-map-first end up with 50 to 80 posts in tight clusters that rank as a structure rather than as individuals.

Starting keyword research for a new website usually feels like guessing, but we’ve learned it works better when you start with real customer language, not keyword tools.

When we worked on a home improvement site, we didn’t begin with search volume. We read customer emails, support questions, and competitor reviews to see how people actually described their problems. That gave us phrases that didn’t look impressive in tools but clearly showed what people needed help with.

From there, we used keyword tools to expand those phrases and find close variations, but we stayed grounded in the original wording. One phrase looked small on paper, but it matched exactly how customers spoke, and it ended up bringing in more qualified leads than broader keywords.

The most important thing to focus on is how closely your keyword matches real-world language. If it sounds like how customers actually describe their problem, it will perform better than a higher-volume term that sounds generic.

I don’t really start keyword research with a process. That’s the honest answer. Most of the “process” talk makes it heavier than it needs to be.

I usually just open Google and start typing whatever the niche is. Half the time I don’t even finish the thought, just see what autocomplete throws at me. That alone already shows how people are actually searching, not how tools think they search. Then I scroll through “people also ask” and related searches at the bottom. That part is messy but it’s real. You can see how one idea splits into ten different directions.

I also look at competitor pages, not to copy anything, just to notice what keeps repeating. If every page in a niche keeps circling the same topics, that’s usually where attention is sitting. Sometimes I’ll even jump to YouTube search or Reddit just to see how people phrase things when they are not trying to sound “SEO correct.” It’s often more honest there.

I don’t spend much time organizing anything at this stage. I just collect phrases and move on. If you try to structure too early, you end up filtering out useful stuff without even realizing it.

The one thing I always come back to is intent. That’s the only part that actually matters.

Same keyword can mean completely different things. Someone searching “SEO tools” could be learning, comparing options, or ready to buy. If you miss that, you can rank and still get useless traffic. I’ve seen that happen a lot. Pages look fine on paper but bring nothing that matters.

So I’d rather target something smaller with clear intent than chase big volume keywords that are vague. Volume looks nice in reports, but intent is what decides if traffic is worth anything.

After that, I do a quick check using SeoSets. That’s my own platform, around $10 a month, built it because most tools felt too heavy for no reason. I just use it to scan what’s actually ranking and if there’s real movement or just noise. I don’t sit there comparing difficulty scores for hours. If something looks forced or empty, I drop it.

Then I group things together instead of treating every keyword as separate. One page usually ends up covering multiple related searches anyway. People don’t search in clean isolated terms, they search in patterns. I learned this after building a few pages that were too narrow and basically went nowhere.

So yeah, nothing fancy. Start messy, follow real search behavior, focus hard on intent, and don’t get too attached to keyword lists.

Keyword research for a new niche works best when treated like audience decoding. First, identify the questions a brand must answer to be seen as credible, then separate those questions into discovery, evaluation, and confidence-building themes. Search tools confirm demand, but the richer insight often comes from studying weak competitor pages, comment sections, and repetitive customer language that signals unmet expectations.

I pay special attention to what is missing from the current search results. If every ranking page repeats the same angle, that usually signals content fatigue rather than complete coverage. The crucial aspect is differentiation, because search visibility grows faster when a page fills a gap instead of echoing the market.

I start with the client’s goals and my own niche knowledge to identify the top commercial and transactional queries first, since these are the clearest route to revenue. Informational queries matter too, but they’re not where I focus my energy early on.

From there, I validate volumes in Ahrefs and Google Ads Keyword Planner, but I treat that data with caution. Volume figures are inconsistent, location-based queries often return nothing, and AI prompt volumes are basically untrackable. This is why I always cross-reference with Reddit and forum threads. This is the step most people skip, but it consistently surfaces real demand and natural phrasing that keyword tools lag behind on. If people are asking something repeatedly in forums, it’s a signal worth acting on regardless of what Ahrefs says.

Then I look at competitors to see what they’re targeting and how much space they occupy. The goal is to find gaps my client can fill quickly and build a counter-strategy from there.

Finally, I check SERP features for priority queries to assess whether a keyword is actually worth chasing. If a query is dominated by AI Overviews, local packs, or ads, the organic opportunity might not justify the effort, and for newer sites, I also need to be realistic about whether the site has the authority to compete at all.

The post-launch keyword research strategy

Launching a new niche leads you to not use third-party tools for keyword research. The tools just show heavily searched terms that you’ll compete against, and there’s no chance of discovering the early language in the niche. The critical part of our keyword strategy comes ~30-60 days after launch, when you mine Google Search Console (GSC) with Regex filters for long-tail variations.

The trick involves opening GSC, going to Performance > Search Results, then clicking “New” under Queries, then “Custom (regex)”. Then type in the regex `.{25,}` which tells GSC “show me all queries with 25 or more characters”.

Third-party tools often say phrases this long have zero search volume, but GSC gives you real queries that customers are typing that lead to your site. For instance, a recent new-niche launch we did for an accounting SaaS product showed us the phrase, “how to automate recurring invoices in accounting software for small businesses,” once we put the above filter in. We didn’t know that phrase existed until then, and it was sitting at position 13, driving about 4 clicks per month. From there, we created a dedicated page that answered that super-long keyword question in the 25+ character query. Over the next few months, the organic clicks from Google to this accounting niche topic cluster grew from around 22 per month to 480 per month, and the conversion rate from visitor to lead on these pages went from 0% to a very profitable 3.2%.

In other words, if you’re a new-niche entrepreneur launching in a new topic, hoovering up these long queries from Google via regex is the fastest way to get revenue-generating organic traffic. The best launch tool is to launch a new site, then do keyword research, and then go.

The process I use is called “search intent mapping.” I don’t start with volume or competition. I start with the question: what does a user actually want when they type this?

Rough sequence:

1. List every permutation of the core topic in a spreadsheet — not with a tool, literally typing what a customer would type. 30 minutes of human brain-dumping. It always beats the tool-first approach because you pick up phrasing and framing a keyword tool misses.

2. Classify each query by intent: informational, commercial, transactional, or navigational. Commercial and transactional go to service or product pages. Informational goes to pillar or blog content. Navigational is handled by your own brand footprint. Most people skip this step and end up with a blog full of informational content that never converts.

3. Only now run the list through Ahrefs Keywords Explorer and Google’s “People Also Ask.” Filter by KD realistic for the site’s DR. If you’re DR 30, don’t chase KD 60. You’re wasting months.

4. Cluster related queries into “page units.” A page unit typically owns 15-40 related queries, not one. That matches how Google and AI-assisted search actually read content in 2026.

5. Sanity check each top candidate by searching it in Google. Ask: “would my page sit naturally alongside these results, or would it look like the odd one out?” If the SERP is all product roundups and you’ve planned a 2,000-word guide, you’ll lose.

The one crucial aspect most people get wrong: intent changes over time. A query that was informational two years ago may now be transactional — AI Overviews and personalisation have moved intent. I re-check intent on our top 50 target queries every quarter. Last autumn we caught a shift on a client’s flagship page — query had moved from “learn about X” to “compare X vs Y” — and a focused rewrite pushed us from position 6 to position 2 in three weeks. 94 extra qualified leads a month from one refresh.

Keyword research isn’t a one-and-done spreadsheet. It’s a living map. Treat it like one.

I’m Kim McNeil, founder of Hatchify. I help companies build in-house marketing engines that last.

For brand-new sites in unfamiliar niches, I treat specificity as a strategic advantage. Instead of chasing high-volume head terms, I focus on long-tail queries that reflect a clear problem, audience, or use case, and, importantly, have weaker SERPs (low domain authority sites, forum results, or thin content).

For example, rather than targeting “project management software,” I’d go after “project management for remote design teams.” The volume is lower, but the competition is lighter, the intent is clearer, and it’s far more realistic to rank in the short term.

From an SEO standpoint, this isn’t just about quick wins. It’s about building topical authority – grouping related long-tail keywords into clusters, interlinking them, and expanding coverage over time. Individually, these terms are small, but together they compound into meaningful, high-converting traffic because they closely match how people actually search.

Most keyword research advice gets the priority order backwards. Tools like SEMrush and Ahrefs train people to optimize for search volume and keyword difficulty — but I learned a different framework working inside Google for 10+ years with the country’s top advertisers.

My process is three stages, and the order matters:

Stage 1 — Map the buyer journey, not the keyword universe. Before opening any tool, I list out the 5-7 specific questions a buyer asks at each stage: problem-aware (“why are my Google Ads not converting”), solution-aware (“Google Ads agency vs in-house”), and decision-stage (“Google Ads agency Sydney pricing”). The decision-stage queries have lower volume but 10-30x higher conversion rates. Most agencies skip this and chase top-funnel volume that never converts.

Stage 2 — Score each keyword by intent, not just difficulty. I use a simple 1-5 buyer intent score: 5 = “ready to buy now” (e.g., “[product] free trial”, “[service] near me”), 1 = “casually researching” (e.g., “what is [topic]”). Then I cross-reference with KD (keyword difficulty). The sweet spot for new sites is intent 4-5 with KD under 20 — these are commercial queries competitors are ignoring because the volume looks small.

Stage 3 — Validate with paid search before investing in SEO. This is the crucial aspect most miss: run a $200 Google Ads test on your top 10 candidate keywords for 2 weeks before writing a single SEO article. Real conversion data beats every keyword tool’s projection. If a keyword doesn’t convert in paid search, it won’t convert from organic either, no matter what SEMrush says about its volume.

The mistake I see constantly: founders spending 6 months ranking for a keyword that has volume but no buyer intent — then wondering why traffic isn’t producing leads. Volume without intent is vanity. Intent without volume compounds.

For a new website or niche, I start by validating which topics have consistent demand and which ones are seasonal, and Google Trends is one of my go to tools for that. It gives a fast read on whether interest is rising, flat, or dropping over time, so I can prioritize keywords that match real search behavior. In our OSHA training space, we saw terms tied to jobsite safety climb in early spring as construction activity picked up, so we planned and refreshed content months before the spike. That lead time matters because it gives pages time to be indexed and gain traction before demand peaks. The crucial aspect is timing your content to the seasonality of the keywords, not just picking keywords based on a snapshot of search volume.

We still start with keyword data. Volume and difficulty from Ahrefs and Semrush are directional and can point you to where demand lives. That part has not changed. But we do not stop there anymore because search is in flux, not only from the surfaces where users are asking their questions, but the funnel and mechanism of the ask, too.

A keyword like “running shoes” has volume. But the user sitting in front of ChatGPT is not simply typing “running shoes.” And even if they were, they likely shared other context in the same session. For instance, they might have shared the desire to lose weight, improve their health, or their bad knees. So their search may be, “What are the best running shoes for someone with bad knees?” But the model is going to search “best running shoes for weight loss,” “best running shoes for bad knees,” “best running shoes for beginners,” because the model is using that session data and history from the user to provide the best answer based on intent. The model breaks down the keywords into sub-queries based on that data to provide the best answer.

So after I pull the keyword data, I map intent at the sub-query level. Not just informational versus commercial versus transactional. What are the actual component questions a model will generate when someone asks the parent query in natural language? Those components become the content architecture. Each section of a page answers one of them, completely enough to stand alone if a retrieval system pulls it out of context.

Looking at keyword research through the same traditional search lens does not account for AI-mediated search and query decomposition. Targeting only high-volume keywords without having intent and query breakdown in mind, or the longer conversational and detail-driven searches humans actually type, puts brands at a disadvantage even in the most specific niches.

For a new website, we begin with search economics. We focus on queries that can move the business within twelve months, not ones that only look good in reports. Then we group the niche into four lanes which are discovery, evaluation, comparison, and validation. Next, we review the search results to understand what Google and AI systems prefer for each lane.

This helps us see if the space rewards education, proof, expert input, or brand trust. Keyword research should match the structure of the results page, not just the term itself. Two keywords may seem similar but need different signals to rank well. If we ignore this, we may target terms that look easy but do not fit a new site.

Hyperlocal research starts with how people in one suburb, postcode, or service area would search when they are ready to act, not with broad national keywords that look good in a spreadsheet. My process is simple: map the core service, map the location modifiers, check what Google is already rewarding in local results, then use Keyword Planner to expand real variations and volumes around those terms. The crucial aspect is intent. A lower-volume phrase with clear local buying intent will beat a bigger vanity keyword almost every time because it is closer to the customer and easier to support with the right proof, page structure, and local signals.

Before using any tool, I map user intent. Some examples of seed terms are forums, reddit, sales notes, and site search, from where I not only expand and cluster them with SERP overlap but also using the volume. Per another powerful tool in your toolbox, you put keywords into “decision paths” (problem – comparison – purchase), structure your content accordingly, and avoid cannibalizing the early stage of their journey.

My main concern is how useful a keyword is beyond search. A product that feeds email content, ads, or social posts is more valuable to me. I choose topics that can be reused across channels. With this approach, traffic isn’t just driven by Google, and each piece of content can serve multiple purposes.

One of the most underrated ways to conduct correct keyword research is to analyze your competitors first. Type in the search term or keyword that you are trying to rank for, see which are the top five organically ranking websites, and then identify the keywords they are ranking for. You can use many tools for this purpose.

What I recommend is downloading all the keywords that the website is ranking for and determining which ones you want to start using. In my opinion, this is the fastest way to perform correct keyword research because you’re starting from a foundation that is already working. You’re not reinventing the wheel; you’re just making it spin faster.

Honestly the biggest thing I wish somebody had told me before I started doing keyword research for a new domain: check the SERP for AI Overviews before you commit to anything. Since January, Google started putting them on roughly half of queries and it changed the math completely. You can rank number one on an informational keyword and lose more than half your traffic to the summary box because the user already got their answer and didn’t scroll down. Not every query though. Transactional stuff mostly doesn’t trigger them, local search basically never does. But on the informational side, it’s rough out there.

And you can’t trust KD tools to flag AIOs, half of them say there’s no AIO and then I load the SERP and there’s one sitting at the top. Just look manually. Open the page, scroll, see if there’s a summary. If there is, ask if the query is still worth chasing for you. Sometimes yes if the intent’s really commercial. Usually no.

Backing up. The actual process I ran for our new domain a month or so ago. Pulled four competitors. Not the top four Google search results, the companies your sales team actually names on the losing-to-these-people slide. Their ranked keyword lists are on Ahrefs if you pay for it, or DataForSEO for like five cents per query if you don’t. Actually that’s ranked_keywords specifically, their bulk endpoints price differently, not important. Stack the four lists, find where they all overlap. That overlap is your real niche. Everything else is noise or a two-year bet, and most of the two-year bets don’t pay off anyway, we dropped almost all of them.

Then filter. And I mean actually filter, not sort-by-volume-and-take-the-top-20. You can’t rank for KD 40 with a brand-new domain and no backlinks, I don’t care what the blog posts claim. So anything with a KD over like half your realistic DR, skip it. Then look at buyer intent. Is the searcher trying to buy or trying to learn? If they’re learning, skip it unless you have a take nobody else is offering. If they’re buying, keep it. This is the rule I should’ve been stricter about.

SERP volatility too. This sounds dumb but just load the page and look. Three DR 80 sites at positions 1-2-3 for three years means they are not moving. Your article won’t move them no matter how good it is. Go find a weaker SERP.