“Do You Have Any Numbers For Japan?”
I pitched AI search as a priority in a planning meeting once, and every chart on the screen was from outside Japan.
That was the first question back.
If you’d pulled the deck right there, honestly, that would have been a sound call. Japanese-language measurement of this is still thin — there still aren’t many studies to point to.
But one survey did ask 2,204 working adults in Japan directly. Every one of them had already used generative AI to look something up.
Among People Who Already Ask AI, Is It The First Stop?
It is, inside that group. Among the 2,204 working adults who had used generative AI to research something, 58.1% said they use it “often” or “almost every time” they look something up (Money Forward’s survey of generative-AI and search-engine use, fielded as of January 2026, R).
77.3% of the same group also rated what the AI tells them as helpful.
The same survey asked why people reach for it, and what worries them about doing so.
- Top reason: “I want the gist fast” — 51.5%
- Top worry: “whether the information is accurate” — 45.6%
So the pattern is: grab the summary first, then wonder if it’s right.
Every one of those numbers, though, is a share of the same 2,204 people.
Which means “58.1% of Japan’s working adults use AI to research” isn’t a sentence this survey supports. The denominator is 2,204 working adults who already research with generative AI, not the country.
What happened to plain search, inside that same group?
The Search Drop Splits By Age, Not By The Average
Among the 2,204, 44.8% said they now use a search engine less often than before. Broken out by age, only two brackets were published: 57% for people in their twenties, 30% for people in their fifties.
“Less often” is as far as the question goes — it doesn’t say where that search-time went instead.
44.8% is the figure for the full 2,204 across every age group, and the twenties and fifties sit on opposite sides of it.
Reading that average onto your own customer base is a bit like checking a company’s average commute time and assuming everyone lives the same distance from the office.
I filed 44.8% into a slide once and left the age breakdown out of it.
A company whose customers skew twenties, and one whose customers skew fifties, are reading two different numbers out of the same survey.
So which one are you? Pull anything with an age field — a membership list, a purchase record — and read off the split.
Re-checking The Answer Also Splits By Age — And So Does The Reason
More than 80% of the 2,204 said they re-search “depending on the situation” after getting an AI answer.
The people who said they re-check often: 56.9% of twenty-somethings, against 26.9% of fifty-somethings — again, more than double.
Checking more could just mean being more careful. Except the reason people gave for checking flips the story.
The single most common reason overall was “I want to confirm the information is reliable,” at 28.4%. Split by age, that reason was picked by 22.6% of the twenties group and 38.9% of the fifties group.
Those two sets of numbers aren’t measuring the same thing. 56.9% and 26.9% are shares of people who say they re-check often; 22.6% and 38.9% are shares of people who picked reliability as one reason among several for re-checking.
One read: younger people check by volume, older people check to settle a doubt. The survey itself doesn’t draw that line — it’s a read, not a finding.
Why the reason flips by age isn’t something this survey measured at all.
Either way, the place people go to check is a third-party article or your own site. Is your own site one of the pages they land on?
How Far Can This Number Travel To Your Own Case
What was asked is “does it feel like it dropped” — not a count of actual searches. It’s a self-report, not a log of behavior.
And the 2,204 people answering already research with generative AI — they’re not a cross-section of Japan’s internet users, let alone Japan’s population.
If you’re going to cite 44.8%, the line right next to it should say “among 2,204 working adults who already use generative AI to research things.”
According to the publisher’s own page, three more conditions come attached.
- The fielding period starts 2026-01-16; no end date is given
- How the 2,204 people were recruited isn’t stated
- The publisher sells accounting and HR-cloud software, which makes this a vendor’s own research
Without a stated sampling method, there’s no way to rule out the respondents skewing toward this vendor’s existing customer base — and no way to rule it in either.
What this survey doesn’t touch at all is the AI’s actual output: what a model says about any specific brand isn’t in here.
Even with those caveats, this fills a real gap. Our own past pieces on English-language measurement have had to say “nobody’s measured this in Japanese yet.”
This puts one Japanese number on the behavior-and-attitude side of that gap — I could be wrong about how far it stretches, but it’s a start where there was nothing before.
Pull Your Customer Age Mix Before You Pick An Order
Grab a membership list or a sales record — anything with an age field — and check where the split falls between twenty-somethings and fifty-somethings.
I didn’t know my own split off the top of my head either, until I looked.
For a customer base that skews young, I’d measure AI visibility first — the share of AI answers where the brand actually gets mentioned. That’s the group moving away from search fastest, so whether you show up matters before anything else.
For one that skews older, I’d flip the order: fix what a reader finds after the name comes up — third-party articles, your own pages — first. That’s the group most likely to go check.
Sitting in the middle doesn’t point either way — small moves on both sides make more sense than betting on one.
The drop in search itself already showed up in a separate study that tracked 2,014 households’ browser activity (SEO to GEO: ChatGPT Took A Fifth Of Search, And The Fix Isn’t Abandoning SEO). What this survey adds is the age split behind that drop.
This is one domestic data point sitting next to the international ones — not a replacement for them, just one more thing on the shelf.
Sources
- Money Forward, “Survey on Generative AI and Search Engine Usage (as of January 2026),” https://biz.moneyforward.com/research/2026-generative-ai-usage-survey/
