There is a small irony in our industry that I find quietly amusing. We spend hundreds of pounds a month on keyword tools that estimate, with great solemnity, what people might be searching for. We pore over volume figures and difficulty scores and intent classifications produced by algorithms that have never met our customers.
And meanwhile, sitting on most websites, completely ignored, is a free and infinitely more honest source of the same information. The internal site search box. The little input field at the top of the page that the customer types into when they cannot find what they are looking for.
The queries that go into that box are, in a way that no external tool can match, the real questions your real customers are asking, in their real words, at the real moment they expected to find something on your site and did not. There is no estimation, no extrapolation, no machine-learned guesswork. It is direct testimony from the people who matter.
And almost no small business pays attention to it.
I want to talk about why this is, what changes when you start paying attention, and how to do so in a way that does not require any expensive software or any special skills.
The reason it is ignored
The first thing to understand is why internal search data has been so neglected, because the reasons explain how to overcome the neglect.
The first reason is that the data is, by default, invisible. Most content management systems have a built-in search function, but the queries that flow through it are not, in most cases, logged anywhere a non-technical person would think to look. The data exists, but it requires either a small technical setup or a configuration step in Google Analytics to surface it.
The second reason is that the queries are scattered and look, at first glance, like noise. A given small-business site might receive between fifty and five hundred internal searches a month, with most queries appearing only once. There is no obvious aha moment when you first open the data. It reveals itself slowly, through patient reading, in a way that does not produce a satisfying dashboard.
The third reason is that the queries reveal uncomfortable things. Why is your shop closed on Sundays. How much does this actually cost. Do you have a return policy. The internal search data is a record of every question your site failed to answer, and looking at it requires accepting that those failures are real. Many businesses, understandably, prefer not to.
These reasons explain the neglect. They do not justify it. The same qualities that make the data hard to look at also make it the single most useful intent signal you can collect.
What you actually learn
Let me describe what the internal search data of a real business — a small bookstore in Bologna we work with — revealed in its first month of being properly logged.
The five most common queries were, in order: gift voucher, English books, audiobooks, opening hours Sunday, and kids' section. These were searched dozens of times each.
The most striking thing about this list, to me, is that none of the five queries were prominently answered on the site at the time the data was first captured. The gift-voucher page existed but was buried three clicks deep under a services dropdown. The English-language section, despite being a third of the inventory, did not have a dedicated category page. The audiobook offering was new and not yet linked from the main navigation. The Sunday hours were correct on the contact page but were not shown on the homepage, despite obviously being a recurring question. The children's section had a page but it was tagged for SEO purposes as literatura per ragazzi — the correct Italian term but not the term that customers, including many tourists and expatriates, were typing.
In other words, the internal search data identified, in a single month, five high-priority content and navigation fixes. Each of them addressed a question the bookstore was demonstrably failing to answer for a measurable number of real customers. None of them would have surfaced from a keyword tool, because keyword tools do not know what is and is not visible on a particular site.
Within six weeks of fixing these five things, the bounce rate on the homepage had fallen by eleven percent and the conversion rate on the gift-voucher product specifically had more than tripled. No new marketing spend. No new traffic. Just the elementary act of giving the customers what they had been asking for.
This is the pattern, more or less, with every small business that begins to take internal search data seriously. The fixes are obvious, the data is honest, and the results are disproportionate.
How to capture the data
The mechanics are, mercifully, simple. There are two basic methods, depending on how your site is built.
If your site uses Google Analytics 4 — and almost all sites do — internal site search tracking can be enabled in the GA4 admin under enhanced measurement. Once enabled, GA4 will automatically capture queries from any standard search box, provided the query is passed as a URL parameter (commonly q or s or search). You can then view the captured queries under the engagement section, filtered to the view_search_results event.
If your search box does not use a URL parameter — some sites submit searches via JavaScript without changing the URL — a small bit of additional setup is required to fire a GA4 event when the search is performed. Any web developer can do this in under an hour. If you do not have a developer, ask your CMS provider; most modern platforms have a documented way to enable search-event tracking.
The data, once it starts flowing, accumulates quickly. After a single week you will have signal on the most common queries. After a month you will have a usable dataset. After three months you will have the rhythm — the queries that recur seasonally, the ones that spike around particular events, the ones that emerge as new customer concerns.
How to read the data
Reading internal search data well requires resisting two temptations.
The first temptation is to focus only on the top ten queries. The head of the distribution will tell you the most urgent fixes — and those should, of course, be addressed first — but it will not tell you the most interesting things. The interesting things live in the long tail: the dozens of low-volume queries that, looked at together, reveal a pattern of confusion you would never have noticed from the head alone.
The bookstore data, for example, had a long tail of queries about parking, bicycle parking, bus stop, easiest way to reach you, and similar variations. None of these were individually frequent. Together they revealed that customers — particularly the older customers who were the bookstore's most loyal demographic — were uncertain about how to physically get there. A small how to find us page, written with care, addressed the whole cluster. The keyword tools, naturally, would never have surfaced this; the queries were too small and too varied to register.
The second temptation is to assume that zero results queries are failures of the search engine. They might be — your internal search is, in most CMSs, not very good — but more often they are failures of your content. If a customer searches for vegan options on your restaurant site and gets zero results, the problem is rarely that the search engine could not find the vegan options. It is usually that the vegan options are not described on any page in a way that the search could match. The customer's failed search is a request for content you have not yet written.
I keep a separate column in the spreadsheet for zero-result queries that should have results. This column, in my experience, is where the most actionable content opportunities live.
What to do with the findings
The temptation, having captured and read the data, is to immediately produce a long list of new content to write. I would urge you to resist this temptation for at least the first month.
The most productive first step is almost always navigation and surfacing, not new content. Many of the queries in the data are for things you already have on the site but have not made visible. The gift-voucher page in the bookstore example existed; it just was not in the navigation. The first wave of fixes should be a careful audit of what is already there and how to expose it. This produces results faster, costs less, and is, in my experience, a better introduction to the discipline of acting on internal search data than committing to a content programme.
Only after the surfacing work is done does the new-content programme begin to make sense. And when it does, the queries themselves — the actual words the customer typed — are the best brief you will ever have for the new pages. Use the query as the page title where natural. Use the question as the H2. Answer the question in the first paragraph. The pages written this way perform substantially better than pages written from keyword-tool briefs, because they are built around what the customer actually asked, not what an algorithm guessed they might ask.
A note on the limits
I want to be honest about one limit of internal search data. It tells you what people who already came to your site were looking for. It does not, on its own, tell you anything about the much larger universe of people who never made it to your site at all.
For that universe, you do still need the external keyword tools — or, in 2026 more usefully, you need to be tracking your AI-citation share and your impressions data in Search Console, both of which give you a window into the queries you are not yet visible for. The internal search data and the external visibility data are complements, not substitutes.
But of the two, the internal data is the one you almost certainly have access to right now, the one you almost certainly are not using, and the one that will produce the most disproportionate return for the next few months of work. The external data is the harder game. The internal data is the low-hanging fruit nobody is picking.
The closing thought
The keyword tools have their place. I use them. I will keep using them. But the longer I do this work, the more I have come to believe that the best keyword research a small business can do is to read, with patient attention, the questions its own customers are typing into its own search box.
It costs nothing. It requires no special expertise. The data is more honest than any external estimate. And the actions it suggests are, almost without exception, the kind of small, useful, low-risk improvements that compound quietly and reliably over time.
If you do nothing else this month, enable the internal-search tracking on your site. In thirty days, come back and read what your customers have been asking. I would be surprised if it does not change at least three things about how you think of your website.
The little box at the top of the page is, in the end, a much better teacher than any tool you can buy.