I once did a pretty embarrassing review. Users were repeatedly searching “how to buy backlinks” on our site, and we’d never written about that topic at all — they searched, got no results, and left. This demand was completely invisible in Search Console — users only revealed it after entering the site, and GSC only records the search engine side of things.
On-site search is demand that users actively speak out — more precise than any keyword tool. External tools give you “people might search for this,” while on-site search is “someone actually searched for it, right here on your site.” The credibility difference between these two is an order of magnitude. This article thoroughly explains how to use on-site search keywords to feed back into topic selection — you can go pull a keyword list from your backend today.
The core is four sentences: on-site search keywords are “real demand from users who already entered your site,” an order of magnitude more credible than external tool estimates; three must-see dimensions are no-result keywords (most urgent), high-frequency keyword coverage gaps, and transactional intent in long-tail keywords; after getting the keyword list, first match against existing content — for keywords that are covered but buried deep, optimize internal links, only write new articles for keywords with zero coverage; start recording today, two weeks of data is enough to see direction, no need to wait for complete history.
Figure: Processing flow from on-site search keywords to topic selection (compiled by 运营GO)
Where the data comes from
The easiest way is the site search report in GA4. The prerequisite is that your on-site search box carries a parameter (WordPress default is ?s=, other systems might use a /search/ path), and you can see the search_term parameter in GA4 events. If it’s not set up, you can start with search records in the WordPress backend or stats from plugins like SearchWP — or even simpler, install a lightweight plugin that records search terms.
How to enable it specifically in GA4: go to “Configure → Data streams → Web data stream,” find the “Site search” setting, enter your search query parameter (usually s or q), save, and GA4 will show what users searched, how many times, and whether there were results in “Behavior → Site search.” This step takes five minutes, but it lets you “see” users’ real demand for the first time — more direct than any third-party tool.
Not having complete history doesn’t matter — start recording today, two weeks of data is enough to see direction. I’m used to exporting a search term CSV at the end of each month, comparing with last month, and newly emerging keywords are often that month’s hot demand. Keyword research can’t only look at competitors and tools — signals from your own site are the cheapest, most accurate first-hand material (the competitor keyword filtering funnel article covers the external perspective, this one covers the internal perspective — combining both gives you the complete picture).
There’s a counterexample worth sharing: a SaaS site had “how to choose tools” ranked #1 in their competitor keyword list, the team wrote a long article, and traffic was mediocre. Later they pulled on-site search terms and found users were actually searching combination questions like “choosing tools vs XX tool which is easier” and “is it enough for small teams” — they didn’t care about generic selection at all, only their own scenario. After changing that long article to “small team tool selection pitfalls,” conversion rate doubled. External keywords give direction, internal keywords give scenario — missing either means you can’t write accurately.
Three must-see dimensions
Prioritize no-result keywords. These are the most urgent: users searched, the site has no content — it’s like pushing away demand that walked right to your door. Organize these keywords, and any that fit your business direction go directly into the topic schedule. The judgment standard is simple — is the person searching this keyword your potential customer? If yes, write it; if not, ignore it.
Compare high-frequency keywords against content coverage. For keywords users search every day, does the site have a corresponding article? If yes, what page does it rank on? Many on-site high-frequency keywords actually have content, but it’s buried deep and the search box can’t match it — in these cases, adding an internal link or adjusting the excerpt can solve it, much faster than writing a new article. A practical action: search each high-frequency keyword on the site one by one, record “has content but ranks on which page,” and prioritize internal link authority-building for those ranking beyond page three.
Find intent clues in long-tail keywords. On-site search terms are usually more colloquial and specific than external ones, like “how much does SEO outsourcing cost per month” — this carries clear transactional intent. Putting these keywords into content gives higher conversion value than generic terms. Don’t dismiss the small search volume — people searching these terms are one clear-pricing article away from placing an order.
How to align with existing content
After getting the keyword list, don’t rush to write new articles — first do a matching pass: search each keyword on the site and see if existing content can cover it. For keywords that can be covered but rank deep, it means “you have the goods but didn’t display them” — optimizing internal links and excerpts can catch the traffic; only keywords with zero coverage enter the topic pool. I have a habit of marking the keyword list in three columns “covered/partially covered/not covered,” spending half an hour each week going through it, and the red-marked not-covered keywords are next week’s topic source. If the same keyword has two articles on your site competing for it, first fix keyword cannibalization before talking about coverage — otherwise the new one you write will get stuck in it too.
Also watch for one pitfall: on-site search terms often expose “combination demand” content gaps. For example, users search “seo outsourcing how much small company” — each individual term has an article, but the combination doesn’t — these long-tail combinations are precisely the lowest-competition blue ocean, worth a dedicated article answering “how to budget for SEO outsourcing for small companies.”
Here’s a specific matching process: suppose your site already has “SEO outsourcing” and “SEO pricing” articles, and an on-site search term “SEO outsourcing monthly payment” was searched 40 times but got zero results. You don’t need to write something new — first add a section “monthly vs annual payment” to the “SEO outsourcing” article, then merge the payment-cycle content from “SEO pricing” into it, then add internal links between the two articles. One small change catches 40 searches/month of real demand, much faster than opening a new article. I use this method every month to catch hundreds of searches, at almost zero cost.
How to interpret search term signals
| Search term signal | Underlying intent | Content action |
|---|---|---|
| “How to + X” | Learning / operation | Write tutorials, step-by-step articles |
| Brand + product name | Comparison / pre-purchase | Write reviews, comparison pieces |
| Industry broad term, no verb | General understanding | Write explainers, definition articles |
| Long-tail question words | Specific sticking point | Write Q&A, FAQ |
Implementation is six steps: today, go to GA4 or your search plugin backend, export the last 30 days of search terms; mark all “no-result keywords,” and schedule the business-relevant ones into topics this week; pick 10 high-frequency keywords, search them on the site, record coverage status and ranking depth; build a monthly topic pool for not-covered keywords, merge and deduplicate with the competitor keyword list; for keywords where “you have the goods but didn’t display them,” add internal links and excerpts within this week to catch traffic; set a monthly end-of-month review of search terms, track newly emerging demand changes.
