A reader emailed me last spring about a site he had spent seven months building. He picked his topic the way almost every guide told him to. Low keyword difficulty, respectable search volume, no big brands sitting on page one. By month seven he had 61 published articles and roughly 340 visitors a month. Every term he ranked for was a question, and almost every one of those questions had an AI Overview sitting above his result.
He did his niche research correctly by 2019 standards. That was the whole problem. This guide covers how to do niche research in a search environment where the old test no longer holds. That test rewarded topics stuffed with easy informational keywords, and those are exactly the searches Google now answers on its own page. What follows uses free methods where free methods work, names the points where they stop working, and tells you when to walk away from an idea instead of spending seven months finding out the hard way.
Niche research is the work of testing a market segment before you commit to it. You check whether people search for it, whether they spend money in it, and whether you can realistically outrank whoever is already there. In 2026 one more test matters: whether the niche contains buying queries, not just questions.
What Niche Research Actually Means
Niche research is the process of narrowing a broad market into a specific audience, then proving that audience is worth serving. You are answering four questions. Do enough people want this? Do they spend money? Can you compete? Do you know enough to say something useful? A niche that fails any one of those is not a niche. It is a hobby.
Most people skip straight to the third question. They open a keyword tool, sort by difficulty, and pick whatever looks easiest. That produces a list of topics you can rank for and no evidence that ranking will be worth anything.
Market, niche, and micro niche
A market is the whole category. Coffee. A niche is a defined slice of it with its own needs and vocabulary. Home espresso for small kitchens. A micro niche is a slice of the slice. Espresso setups for apartments without a dedicated water line.
The distinction matters because it decides how much content you will ever have. Coffee gives you thousands of possible articles and thousands of competitors. The apartment espresso angle might give you 40 articles and almost nobody to fight. Going too narrow is a real failure mode, and it usually shows up around month nine when you run out of things to write.
Why the phrase itself confuses search results
Search for niche research and you get three unrelated things mixed together: business niche selection, academic research specialization, and ecology. One page currently ranking is aimed at researchers choosing a PhD focus area. That ambiguity is worth knowing about, because it means the search results for the bare phrase are unusually soft and the results for the how-to phrasing are a different set of pages entirely.
Why Niche Research Changed in 2026
The classic method told you to find a topic full of low-competition informational keywords. That advice inverted itself. Independent measurement now shows AI Overviews absorbing a large share of clicks on informational queries, which are precisely the queries the old method told you to chase. A niche made entirely of questions is a niche where ranking well may produce very little traffic.
The single most useful number here is not about clicks at all. Ahrefs reported that 99.2% of keywords triggering an AI Overview have informational intent. Read that alongside their click data and the implication is direct. The queries most exposed to AI Overviews are the same queries beginners are told to build a site around.
99.2% of keywords that trigger an AI Overview are informational in intent.
Ahrefs, April 2025
What Google says versus what studies measured
I checked Google’s own announcement rather than relying on how others summarize it. In the May 2024 post introducing AI Overviews in the US, VP of Search Elizabeth Reid writes that links included in AI Overviews get more clicks than if the page had appeared as a traditional web listing. That is the company’s position, stated plainly. Independent researchers have not reproduced it. Ahrefs analyzed 300,000 keywords and found the presence of an AI Overview correlated with a 34.5% lower click-through rate for the top-ranking page, published April 2025. Ahrefs later announced a re-run using December 2023 and December 2025 data reporting a 58% drop. I have read the original study in full but not the update page, so treat the 58% figure as the company’s own reported result rather than something I verified line by line.
Other measurements land elsewhere. Across the published studies the estimated impact ranges roughly from the mid teens to the high eighties in percentage terms, depending on query type and method. Pew Research Center tracked real browsing rather than modeled curves and reported a lower click rate when an AI Overview appeared, though sources describe its sample differently and I have not opened the Pew report directly. The honest summary is that every independent study found a decline and none of them agree on the size.
| Claim | Support level | What that means for you |
|---|---|---|
| AI Overviews trigger overwhelmingly on informational queries | Measured by Ahrefs across 300,000 keywords | Screen niches for commercial queries, not just easy ones |
| AI Overviews reduce clicks to the top result | Measured repeatedly, size disputed | Assume a real cut, do not plan around a precise number |
| Links in AI Overviews get more clicks than normal listings | Stated by Google, not independently reproduced | Do not build a traffic forecast on it |
| Passion niches outperform money niches | Practitioner opinion, no public measurement | Useful heuristic, not evidence |
Step 1: Build a Wide List Before You Judge Anything
Start by generating far more ideas than you need, and resist scoring them. Aim for 30 to 50 candidate topics in a single sitting. Judgment at this stage kills good ideas early, because the strongest niches usually sound unglamorous. Nobody gets excited about drainage for sloped yards until they see what the customers spend.
Three sources that cost nothing
- Your own last two years. What have you researched obsessively, bought repeatedly, or been asked about by friends? This is the only source that produces first-hand knowledge, which matters more than it used to.
- Complaint threads. Reddit, Facebook groups, and product review sections. You are hunting for the same frustration stated three different ways by three different people.
- Category trees on marketplaces. Click down four levels into Amazon or eBay subcategories. Level four is where niches live. Level one is where corporations live.
Google’s own guidance on creating helpful, people-first content asks whether your content demonstrates first-hand expertise, giving the example of expertise that comes from having actually used a product or visited a place. That is a documented ranking consideration, not a vibe. It is also the strongest argument for weighting your own experience heavily in step one.
Step 2: Check Demand Without Paying for a Tool
Demand checking answers one question: are enough people looking for this, and is that number growing or shrinking? You can answer it for free. Google Trends gives you direction over time. Google Keyword Planner gives you rough volume. Autocomplete and the related searches at the bottom of a results page give you the actual language people use.
The five-year trend rule
Set Google Trends to the past five years, not the past twelve months. A twelve-month window makes seasonal topics look like growth and makes dying topics look stable. Five years shows you the shape. Flat is fine. Rising is better. A clean downward slope is a reason to cross the idea off, no matter how easy the keywords look.
Where the free tools mislead you
Keyword Planner rounds volumes into wide buckets unless you are running ads, so a term showing 10 to 100 searches could be either. More importantly, keyword tools underreport specialist topics badly. A phrase can show zero volume while People Also Ask boxes surface it constantly, which tells you real humans are asking it. Treat a zero from a keyword tool as missing data, not as proof of absence.
Step 3: Screen for Money Queries, Not Just Easy Ones
This is the step that separates 2026 niche research from the older version. Sort your candidate niche’s keywords by intent and count them. You want a meaningful share of commercial and transactional terms, the ones with words like best, review, cost, price, near me, alternative, or versus. A niche that is 95% questions is heavily exposed to AI Overviews.
There is a simple free proxy for this. Search a handful of your niche’s main terms and note two things: whether an AI Overview appears, and whether there are shopping results or ads. Ads mean somebody is paying for that click, which means the click is worth something. No ads anywhere in a niche is a quiet warning.
A rough intent mix to aim for
| Intent type | Example phrasing | Healthy share |
|---|---|---|
| Informational | how to, what is, why does | 50% to 70% |
| Commercial | best, review, comparison, versus | 20% to 35% |
| Transactional | buy, price, cost, discount, near me | 10% or more |
These proportions are my own working rule from screening niches, not an industry benchmark anyone has measured. Use them as a sanity check. If your niche comes out at 98% informational, you are not disqualified, but you need a plan that does not depend on search clicks.
Step 4: Test Whether You Can Actually Compete
Competition analysis means reading the first page of results and identifying who owns it. You are not looking at difficulty scores. You are looking at page types. Ten established brands with deep coverage is a closed door. A page one full of forums, marketplaces, and general-interest sites is an open one, because it means no dedicated site has claimed the topic.
What a winnable results page looks like
- Reddit threads or forum posts ranking in the top five
- Wikipedia, Pinterest, or Amazon category pages holding positions they have no business holding
- Results that are technically relevant but clearly answer a different question
- Pages last updated years ago with no sign of maintenance
Google’s Search Essentials documentation is worth reading once before you commit, because it sets out the baseline every competing page is measured against. Nothing in it is exotic. It is a useful reality check on niches where the incumbents are already doing the basics well.
One caution on the forum signal. Reddit ranking well used to mean weak competition. It now also reflects Google actively favoring discussion content, so a page one of Reddit threads is more ambiguous than it was three years ago. Read it as an opening, not a guarantee.
Speeding this up with search commands
Checking a competitor’s true size takes about a minute if you know the right queries. You can estimate how many pages a site has indexed, find every article they have written on a subtopic, and spot thin sections they never filled in. Our guide to Google search operators covers the specific commands and how to combine them for competitor audits.
Step 5: Confirm the Niche Has Keyword Depth
Depth means the niche can support enough articles to build authority. A useful floor is 30 to 50 genuine article topics that a new site could plausibly rank for. Fewer than that and you will exhaust the niche before Google trusts you. This is where niche keyword research and niche research overlap, though they are not the same job.
Niche research asks whether the market is worth entering. Niche keyword research asks which specific pages to build once you have decided. Doing the second before the first is the most common sequencing mistake, and it is why so many people end up with a technically rankable site nobody wants to buy anything from.
Validating individual keywords
For the keyword-by-keyword work, the Keyword Golden Ratio is the most useful free method I know, because it uses the number of competing pages that actually target a phrase rather than a proprietary difficulty score. It has real limits, and it works better on some niches than others. Our full breakdown of the Keyword Golden Ratio method covers the formula, the thresholds, and where it fails.
Step 6: Disqualify Fast
The most valuable skill in niche research is rejection speed. Every guide teaches you how to find niches. Almost none teach you how to kill one quickly, which is where the actual time savings live. Give each candidate a fifteen minute review against a fixed list, and cut anything that fails a hard criterion instead of talking yourself into it.
Five reasons to walk away
- Five-year trend clearly declining. You would be renting space in a shrinking room.
- Health, money, or safety topics with no credentials. These face the strictest quality assessment, and personal opinion carries little weight.
- No commercial queries anywhere. Interest without spending is an audience, not a business.
- Page one is ten strong brands with recent updates. Save that fight for when you have authority.
- Fewer than about 30 real article topics. You will run dry before you rank.
Once a niche survives all six steps, the next decisions are practical ones. What you call the site affects whether people click your result at all, which is covered in our look at how domain names influence click-through rates.
Niche Research for Local and Service Businesses
Local niches follow different rules and are frequently the better option in 2026. Search volume is far lower, but competition is thinner, the queries carry buying intent, and AI Overviews interfere less with a search for a nearby provider than with a general how-to question. A niche of two hundred monthly searches can outperform one with twenty thousand if every one of those searches is somebody ready to hire.
What to check that is specific to local
Look at the map pack rather than the standard results. Count how many businesses appear, how complete their profiles are, and how many reviews they have. Three competitors with eleven reviews each is a very different situation from three with four hundred. Our guides to Google Maps ranking factors and building local citations cover what actually moves the needle once you have chosen a local niche.
How Long Niche Research Should Take
Full niche research on a shortlist takes most people somewhere between six and twelve hours spread over a week. Not six weeks. The people who spend months on this are usually avoiding the harder work of publishing. At some point the research stops producing new information and you are just looking for permission.
A realistic timeline after you choose
Expect little from a new site for six to twelve months. That is not pessimism, it is how domain trust accumulates. Plan for meaningful traffic somewhere in months nine to eighteen, and treat anything faster as luck. Because search clicks are less reliable than they were, build an email list from day one so a single algorithm change cannot erase your audience. If you go that route, our guide to email deliverability covers the setup that keeps those messages out of spam.
Frequently Asked Questions
What is niche research?
Niche research is the process of testing a narrow market segment before committing time or money to it. You examine search demand, buying intent, competition strength, and your own knowledge of the subject. The output is a decision: build here, or reject and move on. It is market validation applied to a specific audience rather than a general industry.
How do I know if my niche is too small?
Count how many distinct article topics you could write that people actually search for. Under roughly 30 and the niche is likely too narrow to sustain a site. Also check whether products or services exist that your audience already buys. A niche with plenty of interest but nothing to sell is too small in the way that matters most, regardless of its search volume.
Is niche research the same as keyword research?
No. Niche research decides whether to enter a market at all, weighing demand, money, competition, and your own expertise. Keyword research decides which individual pages to build after that choice is made. Keyword research is one input into niche research, not a replacement for it. Running keyword research first is the most common sequencing error beginners make.
Can I do niche research for free?
Yes, for most of it. Google Trends, Keyword Planner, autocomplete, People Also Ask, related searches, Reddit, and manual results-page reading cover the majority of the work. Paid tools mainly save time and give you competitor traffic estimates. They are a convenience at the research stage, not a requirement. Budget for them later if the niche proves out.
Should I pick a niche I am passionate about or a profitable one?
Pick something profitable that you can tolerate for three years. Pure passion niches often lack buyers. Pure money niches usually get abandoned before they mature. Familiarity matters more than passion, because first-hand knowledge produces content that thin competitors cannot copy. Note that this is practitioner consensus rather than anything measured, so weigh it against your own situation.
How do I know if a niche is too competitive?
Read page one for your main terms. If every result is an established brand with content updated in the last few months, treat it as closed for now. If you see forums, marketplaces, general-interest sites, or stale pages, there is room. Difficulty scores from tools are useful for sorting but should never override what you can see on the actual results page.
What is a micro niche?
A micro niche is a subdivision of a niche, narrowed to a highly specific audience or use case. Coffee is a market, home espresso is a niche, and espresso for apartments without plumbing access is a micro niche. Micro niches face very little competition but risk running out of content. They work best as an entry point you widen later.
How many keywords does a niche need?
There is no official threshold, but 30 to 50 genuinely rankable article topics is a reasonable working floor for a content site. What matters more than the count is the mix. Thirty keywords including ten with commercial intent beats two hundred that are all questions, because the commercial terms are the ones least affected by AI Overviews.
How do I check if a niche is dying?
Set Google Trends to a five-year window and look at the slope. Then check whether new products still launch in the space and whether community discussion is current. A topic can hold steady search volume while its commercial ecosystem quietly collapses, so look at both signals. Recent forum activity is often the earliest warning sign available.
Does niche research work for dropshipping and affiliate sites?
The six steps apply, but the screening criteria differ. Dropshipping adds supplier availability, shipping weight, and return rates. Affiliate sites add commission rates and cookie duration. Both weight commercial intent far more heavily than a general content site would. Run the same process, then layer your model’s specific requirements on top before making a final call.
Where This Leaves You
The reader with 61 articles and 340 visitors did not fail at execution. He wrote consistently, published on schedule, and ranked for what he targeted. He failed at the screening step, because the test he used was built for a version of search that no longer exists. His niche had no buyers in it, only askers, and the askers now get their answers before they reach anybody’s website.
The fix is not more work. Knowing how to do niche research now means applying a different test, earlier, with far more willingness to reject. Check demand, then check whether anyone in that demand spends money. Read the actual results page instead of a difficulty score. Give yourself permission to kill an idea in fifteen minutes. Then commit properly, because a mediocre niche you work for two years will beat a perfect one you abandon in three months.
He is now rebuilding around the small commercial corner of that same topic, which was there the whole time. So here is the question worth sitting with before you start: if search sent you no traffic at all for the first year, would your niche still be worth building?
