The shift to intent-based search

For about twenty years, we taught people to search like machines. You'd shrink a question down to a few keywords, guess at the words a page might use, and pick through the links for something close. We all got good at speaking machine.

Now the machine speaks back. Ask Google a real question and the AI Overview answers it on the spot, with no list to wade through. ChatGPT does the same. Quietly, asking and getting an answer has become what people mean by search.

When someone asks your own search a question, they hand you their intent in their own words. What people ask, and what they ask for and can't find, is the clearest read on what they actually want. Search engines used to pass those words to site owners, and then they stopped. On your own site, you still get every one. Reading them is where we want to go next.

People bring that habit with them. They land on your website, look for the search, and expect to ask a question and get an answer. The moment your own search can do that, a few old assumptions start to wobble.

Take your navigation. We've spent years building careful menus and category trees so people can find their way around. If someone can just ask and get the answer, how much of that do they still need? Maybe less than we assume, which is a strange thing to realise after all those menus.

Integrating smart search solution into your website might sound complex, but it is actually the easy part. The hard part is what you now have to decide.

What should it answer, and how? An answer is an opinion in a way a list of links never was. If someone asks your site which product suits them, you're giving advice with your name on it. So the limits are yours to set. It shouldn't quietly send someone to a competitor, and it shouldn't invent something to look helpful, because a confident wrong answer does more damage than a page of honest links ever did. Sometimes the right reply is "we don't have that", and someone has to decide it's allowed to say so.

Then there's what comes back. If the result is still a list of search results, nothing has really changed. People wanted an answer, so you give them one: a recommendation, a short comparison, the next step. That is a different thing to design than a results page.

Under all of it sits one question: do you actually know what people will ask? Not the keywords you wish they'd type, the real questions, in their own words. And is your content ready to answer them? A smart search is only as good as what sits behind it. If the answer isn't somewhere in your own data, the search can't conjure a good one, and you really don't want it to try.

This is also where a blank "ask us anything" box tends to fall over. Faced with all that freedom, plenty of people freeze and don't know what to type. The kinder version helps them get to the question, the way a good shop assistant narrows things down with you.