The group changes between the percentages
In the KB survey, 55.5 percent of Generation Z respondents say they have used AI for health or medical information. Among that generation’s health-related AI users, 37.1 percent report emotional or psychological consultation. The denominator narrows between those figures: respondents are first selected for having used AI for health, then asked how they used it. Reading 37.1 percent as the consultation behavior of all young people changes the population being measured. What interests me is the question this forces about representation. The preferences of people who reach a health-related tool do not automatically describe the needs of those who never reach it.[1]
The generational comparison needs the same care. Emotional or psychological consultation is reported by 12.6 percent of health-related AI users born in 1955–1964. That group’s overall experience of health-related AI use is 41.9 percent, compared with 55.5 percent for Generation Z. The consultation rates therefore come from user groups of different sizes selected through different patterns of adoption. Their difference is a useful signal about the questions younger users bring to a conversational tool. It cannot establish the ratio between the generations’ levels of need for mental-health services. Any bridge from reported preference to underlying need remains a hypothesis requiring its own test.[1]
Counting people who never reach the tool
The sample narrows that inference further. KB studied 2,000 people aged 25–69 with independent economic activity, including retirees. Online and mobile surveys and focus-group interviews were conducted in July and August. Generation Z is defined as people born in 1995–2004, yet the sample’s lower age boundary prevents assuming that the entire birth range is represented. Extending these results to all young people in the country, students or people without economic independence adds another representation assumption. Detailed focus-group accounts can illuminate reasons for use, but cannot by themselves determine the population weight of people excluded from the sample.[1]
One possibility is that younger users turn to AI to overcome difficulty accessing consultation. Another explanation fits the same observation: people familiar with digital tools may bring everyday concerns to a chat window because of convenience or expected privacy. Those explanations imply different service decisions. The first centres an access barrier; the second centres a preferred form of communication. Current usage rates cannot choose which dominates. Asking only users can also leave people with substantial needs but little willingness to use the tool invisible. I value the survey as an initial map; filling its blank spaces requires asking about the experiences of nonusers too.[1]
The question for the next measurement
A follow-up study could start with a sample assembled independently of AI-use status. Within comparable age and economic groups, users and nonusers could both be asked about care needs, barriers to services and preferences for digital tools. Responses from people reporting consultation use could then be compared on that common basis. This would begin to distinguish associations with need, access and habit. That is my proposed research design, rather than a measurement KB has already performed. For sensitive health information, the requested detail should serve the research question while taking account of the privacy burden imposed on participants.[1]
My decision threshold therefore does not rest on a rising usage rate alone. I would change my assessment if comparable questions across different access conditions clarified which unmet needs accompany consultation use. That would provide a firmer basis for designing health services. KB’s findings currently establish that health questions are reaching conversational tools and that user groups bring different purposes to them. The next step is to count the people behind those preferences more accurately. Preserving the denominator strengthens the usefulness of the research by specifying which decisions it can support.[1]