Keyword Research Masterclass for 2026
Keyword research has not stopped working. What has changed is that search volume is a less reliable guide than it used to be, because a growing share of queries are answered without a click and AI-generated answers sit above the traditional results. A page ranking second may now receive far less traffic than the same position did three years ago.
The response is not to abandon keyword research. It is to weight it toward queries where a click is still likely and where you can demonstrate something a summary cannot.
What changed in 2026
Answer engines absorb informational queries. A question with a short factual answer is increasingly answered in place. Volume estimates for those terms still look attractive and convert poorly.
Branded and navigational queries still click. Someone searching for your product by name wants your site, not a summary.
Comparison and evaluation queries still click. When someone is choosing between two tools, they read the detail. This is why comparison pages remain among the highest-value content types.
Volume estimates are rounder than they look. Many tools show the same banded figure for genuinely different terms, and low-volume terms are frequently extrapolated from a tiny sample. Treat volume as a rough rank order, not a number.
Start from intent, not from volume
Classify before you rank. Four buckets, and each needs different content.
- Informational. "how does DMARC work". High volume, heavily absorbed by answer engines, weak commercial value.
- Commercial investigation. "mailchimp alternatives for saas". Click-heavy, high value, where comparison content competes.
- Transactional. "email marketing pricing". Lower volume, highest intent.
- Navigational. Your brand plus a feature name. Always worth owning.
A term's value is intent multiplied by how likely a click still is. A 200-volume commercial term is often worth more than a 20,000-volume informational one.
Finding terms your competitors have not covered
The highest-return technique is not finding what everyone targets, but finding the gaps inside topics they cover shallowly.
- Take a topic your competitors rank for. Read their page fully.
- List the questions it raises but does not answer. Every "you should also consider configuration" without detail is a gap.
- Check whether those questions appear in search suggestions. If people search them, they are viable.
- Write the page that answers them properly. Depth on a sub-topic beats another overview of the parent topic.
This works because broad terms are saturated and sub-topics usually are not. It also produces content an answer engine cannot summarise away, because the value is in specifics.
Evaluating a term before committing
Four questions, and a term needs to pass all four.
- Can I demonstrate something a summary cannot? Original data, a worked example, specific configuration. If not, an answer engine will absorb the query.
- Is the intent commercial? Otherwise high rankings produce traffic that does not convert.
- Do I have any realistic path to ranking? On a young domain, a competitive head term is not a plan.
- Does it connect to a page I already have? A term with no natural home means creating a page for a keyword, which is how thin content gets made.
Mapping terms to pages
One page should target one primary intent, supported by related terms in the subheadings. Two pages competing for the same intent split their own authority, which is a common self-inflicted problem on sites that have grown by accretion.
Before writing, search the term and see what the results are. If the top results are listicles, a long-form guide will struggle regardless of quality, because the intent is to compare options quickly rather than to read.
What to do with low-volume terms
Do not discard them. A cluster of twenty related low-volume terms often generates more qualified traffic than one head term, and it is far easier to rank for. Publish them as genuine pages rather than as near-duplicate variations of each other, since a set of pages differing only in a location or a year is exactly the pattern search engines treat as low quality.
A repeatable process
- List your topics, not your keywords.
- For each topic, find the top three ranking pages and list the questions they answer shallowly.
- Check those questions for search demand.
- Classify by intent and drop anything purely informational with no commercial connection.
- Map each surviving term to one page, merging duplicates.
- Write the deeper version, with specifics and worked examples.
- Review in three months against actual conversions, not impressions.
Why your own data beats every keyword tool
Third-party volume estimates are extrapolated from a sample that may not represent your market at all. Your own search console data is measured.
Start from what you already rank for, including queries where you sit outside the top ten. Terms where you appear on page two or three are the most actionable in any account, because you have demonstrated some relevance and need improvement rather than a new page from nothing.
Also read the queries your pages appear for but are not targeting. That list frequently contains the exact phrasing your audience uses, which is more useful than a tool's suggestion because it is real.
Internal search is the most underused source
If your site or product has a search box, its query log is a list of terms people used your own product to find. Support tickets are the same signal in a different form. Both describe what your audience wants in their own words, with no estimation involved.
The practical move is to look for terms appearing repeatedly that you have not written about. A question asked in support ten times is a page, not a documentation gap.
Intent shifting over time
Searcher intent for a term is not fixed. A phrase that meant one thing five years ago may now be dominated by a different need, and the results page reflects that shift before any tool's volume estimate does.
Before writing, look at what currently ranks. If the top results are all a format you cannot easily produce, the intent has moved and your plan needs to change or the term should be dropped. Checking this takes two minutes and prevents writing a page that cannot win regardless of quality.
The short version
Keyword research still works, but volume is a weaker guide because answer engines absorb informational queries. Weight toward commercial and comparison intent where a click is still likely. Find gaps in how competitors cover a topic rather than competing on the parent term.
Before committing to a term, check that you can demonstrate something a summary cannot. Use your own search console data and internal search logs ahead of third-party estimates. Map one intent per page, and judge the result on conversions rather than traffic.
Continue learning
Related Cresca resources
External references