One customer question, a dozen different searches

Search Console shows you every search your site appeared in, one row per wording, and almost none of those rows mean anything on their own. People ask the same question a dozen different ways, each wording gets its own row, and each row is too small to act on. What’s worth acting on is the question underneath, and it only shows up once the rows are grouped together.

Take an accounting firm in Oakville. Over a month, its Search Console report might list searches like these.

Most people would sort these into three or four questions without thinking about it. Someone is looking for an accountant in town, a business owner wants help with the books, and someone else wants their personal return done. Search Console sorts them into twelve rows. Each row gets its own count of how often the firm appeared, its own average position and its own clicks. On its own, a row with six appearances and no clicks tells you nothing. Put it back with the rest of its question and those same six appearances might belong to something customers ask every week that the firm almost never wins.

Most searches are too small to act on, and they still add up

The same pattern shows up at scale. In August 2026, one business we monitor appeared in Google for 24,301 different searches. Of those, 20,134 (83%) were seen fewer than 10 times in the whole month. None of them is worth an hour of anyone’s attention on its own. Together, they accounted for just over a fifth of every time the site appeared in search results that month.

Search Console’s own table stops at 1,000 rows, sorted by clicks. For this business, that’s about 4% of the list. The list is also incomplete, because Google withholds searches too rare to show without risking identifying someone. Across nearly 900,000 Search Console properties, Ahrefs found that those hidden searches accounted for close to half of all clicks.

Grouped by the question underneath, most of this business’s searches fall into about 2,500 topics, roughly eight wordings to a topic. That’s still a long list, and the weight on it is very uneven. Just 85 topics account for half of every time the business appeared in search. About 450 account for four-fifths. A list of 24,000 rows hides that shape completely, and the topics bring it to the surface.

Words that look alike can be different questions

Grouping sounds like a sorting job, until you notice that wording is a poor guide in both directions. “cpa oakville” and “accounting firm oakville” share one word and ask the same thing. “tax accountant oakville” and “accountant oakville” share almost every word, and one may well be a person with a return due in April while the other is a company looking for someone to handle its year-end.

Google provides one useful test. If two searches bring up largely the same pages, Google is treating them as the same question, whatever the words say. That’s one of the signals DepthFinder weighs. The hard part is everything around it. Some searches have too little data to test, some wordings sit between two questions, and a long tail of one-off searches belongs to no group at all. Google has said that about 15% of the searches it sees each day are brand new, so the list keeps growing new rows.

A grouping is only useful if it holds still. When this month’s topics are drawn differently from last month’s, there’s nothing to compare against. DepthFinder keeps topics stable from one month to the next, so when a topic’s numbers move, the movement comes from your results.

Google groups searches too, but only if your site is big enough

Since late 2025, Search Console has grouped similar searches for some sites, in a card in its Insights report. Google offers it only to sites with a large volume of searches and hasn’t said where the line sits. The business above, at more than 24,000 searches a month, doesn’t qualify. Where the card does appear, groups are ranked by clicks, so they’re built from searches you already win traffic on, and Google notes they may change over time.

Google’s reasoning is that grouping adds less for smaller sites. For a small business it runs the other way. The smaller the site, the thinner each row, and the more it depends on adding rows together to see anything at all. For most small and medium-sized businesses, Google leaves that work to you.

What a topic can tell you that a row can’t

Grouped searches have a stage, and most topics hold a mix of both. Some searches come from people still working out what their problem is, which DepthFinder calls Understanding. Others come from people looking for someone to buy from, which it calls Buying. For the business we monitor, four in every five times it appeared in search were Buying searches, and Buying is the larger share in about two thirds of its topics (1,647 of 2,548). A business can show up well for one stage and badly for the other for years without knowing which. That distinction, covered in what stage your customers are at when they find you, only means something across a group of searches, because a single wording rarely carries enough to tell.

Grouping also puts the gaps in view. Search Console only reports searches where your site already appeared. Set a topic against what the wider market searches for, and who turns up on page one when you don’t, and the missing ground shows up as a handful of topics instead of a scatter of rows you were never shown.

The topic is also the unit AI answers work in. When someone asks an assistant a long question, it turns that question into a few shorter searches before choosing its sources (how AI search works). Nobody asks ChatGPT “cpa oakville,” but the searches it runs on their behalf land among the ones people type. DepthFinder tracks AI answers in Google AI Overviews, ChatGPT, Claude, Gemini and Perplexity on the same topics as your search results, so you can compare the two directly.

You can see the shape of this in your own data in an afternoon. Export the top rows from Search Console’s performance report and sort them by hand into the questions behind them. Most people end up with a short list of questions from a long list of rows, and a few of those questions have no page on the site that answers them.

The plan lives at the level of the topic

A list of searches tells you which exact words brought someone to your site. Very few decisions about a business depend on that. The questions that decide what to work on are what customers ask about and whether they’re ready to hire someone when they ask, and both have answers only once the rows are grouped. For the business we monitor, 76 of its 2,548 topics are both among the most searched and mostly searches from people looking to buy, and those 76 carry more than half of all its Buying demand, which is a list short enough to work through.