Keyword clustering
Keyword clustering is grouping searches that mean the same thing, so one page can target the whole group instead of one phrase at a time.
Keyword clustering is sorting a long list of searches into groups that share one meaning, so each group can be served by one page. "Best coffee grinder for beginners," "easy coffee grinder" and "first coffee grinder to buy" are three searches with one answer, so they belong in one cluster.
For a marketer, it turns a spreadsheet of thousands of keywords into a plan of a few dozen pages, and it stops you writing three pages for the same question.
How it's done
There are two common ways. One groups keywords by meaning, either by hand or with a tool that compares the words. The other groups them by results: if two searches bring up mostly the same pages on Google, Google treats them as the same question, so they go in the same cluster.
Results-based clustering is closer to how search engines think, because it uses what they actually show. For example, if "coffee grinder for french press" and "grinder for cafetière" share eight of their top ten results, one page can serve both.
From keyword clusters to prompt topics
The idea carries straight over to AI search. People ask AI platforms the same thing in many different words, and the answers tend to be similar for questions that mean the same. So tracking one question in ten wordings tells you little more than tracking one well. It's better to group prompts by the need behind them and track a handful of different questions in each group.
AI questions are also longer than keywords, so a cluster tends to be a need, like "choosing a first grinder," more than a phrase.
In AnswerOrbit
Topics are clusters for prompts. Group your prompts by a need your customers ask about, such as coffee subscriptions or office coffee, and you'll see your visibility for each topic as well as for each prompt. The Coverage tab points out duplicate prompts. You can also connect Google Ads to see the keywords you bid on next to your visibility on each topic. See Topics and prompts.