I call bullshit on the idea that AI search has changed how people use your shop's search box. But you'll be getting a bill for that idea soon nevertheless.
We track every angle of over 25 billion searches a year. What we see in that data is plain: customers still treat the shop search differently from their AI agents.
What Google's numbers are about
In Alphabet's Q3 2026 earnings remarks, its CEO said: "Queries in AI Mode are on average three times longer than traditional searches, and one in five AI Mode searches globally use voice, images, or documents."
But be careful: Google AI-Mode is not your ecommerce shop's search box. It would be presumptious to make such a correlation.
You're forgiven if you made such a leap. AI Mode queries are longer, so the queries in your search box must be getting longer too.
So let's have a look at what people really type when the box belongs to a shop.
Still short
In our data, ecommerce search still sees customers typing 1 to 3 word queries the most. A 4 word query is already an outlier, not to mention a complete sentence.
Andreas Wagner put it this way on this blog in September 2024!: "While many search vendors assert a significant increase in natural language understanding query volume, our production data does not support this claim. We observed a slight upward shift, but not substantially or statistically significant." No significant change since 2024.
A fix for a problem that doesn't exist
Our conclusion: the market is calling for solutions and spend for a problem that doesn't exist (yet). Shoppers typing whole sentences into your search box is a story we don't see in the data (yet).
Don't get me wrong: the day may come when shopper behavior changes. I can't tell you when, and neither can anyone selling you the fix today.
Until then, the job in front of you is an older one, and it's harder than the story makes it sound.
The name of the game is still intent
You have a limited window to work out what a shopper wants, and truncated context to work with. You get a fragment, rarely a complete sentence.
Take "trousers", one of the most common queries across our customer base. For whom, and for what? The query doesn't say, and your engine still has to answer.
Then comes the long tail: requests that may only ever appear once or a handful of times a season. Our Dead ends page puts it plainly: "most queries appear only a handful of times, so no single interaction carries enough signal to call intent reliably."
The same KPIs that give us the 3 word query length, also account for 750M unique phrases. Most of them will never reach the top of the report anyone reads on a Monday.
The monsters are already there
In our view, the more profitable and economical move is still to improve the search engine you already have. You don't need to replatform, and you don't need to pay a fortune for analysts to go looking for monsters under the bed.
The monsters are already there. We know them because we've found them, and none of them is a shopper typing a paragraph:
- One purchase intent split across thousands of variants, each treated as its own search.
- Searches that return something, just not the right thing, so they never show up in a zero-result report.
- Rare queries, too thin to judge one at a time.
- Synonym lists kept by hand, where a fix in one place quietly breaks something in another.
searchHub optimizes search and product discovery. It groups every spelling and variant of a search into one intent. Then it rewrites each variant to the query that has proven it sells, before your engine looks anything up, so the engine receives a query it already handles well.
Because it works on the query itself, it can run a test that ranking alone can't: same shopper intent, two different queries. You keep the one that sells, and a change that doesn't measurably improve results is rolled back.
You've gotta get yourself a partner who can do this kind of optimization without pulling money out of your pocket unnecessarily.
Keep the engine. Fix the input.
The short version
Queries in Google's AI Mode are longer. In our data, the shop search box still sees 1 to 3 word queries the most.
The day that changes, we'll say so. Until then, spend the money on the queries you have.
If you want to see what your own queries hold, ask for the business case for your shop. Any online shop can ask, on whatever engine it runs, and the offer is open, with no end date. We read your real query data, not a sample, and no sales call is needed to receive it.
