Aiso research commentary · Israel · November 2026

78% of Israeli AI users ask health questions. Only 25% use it to code.

Benjamin Tannenbaum, Founder and CEO, Aiso
By Benjamin Tannenbaum · Founder and CEO, Aiso · LinkedIn
7 min read

First published .

We talk endlessly about AI writing code. In a new Israeli survey, health questions were far more common. The same research also found that fewer than four in ten AI users regularly check the answers they receive.

78%

of surveyed AI users asked health or medical questions

38%

frequently or always verified AI answers with another source

25%

used AI to write code

A study presented on November 19 by Shenkar, the Samuel Neaman Institute for National Policy Research, Wikimedia Israel and the Israel Internet Association looks at how people in Israel consume information, use AI and decide what to trust.

Its headline finding is about ordinary questions. Among surveyed people who had used AI in the previous month, 85% used it to satisfy their curiosity and 78% asked about health. Only 25% used it for coding. If you work in AI, the last number may sound surprisingly low. If you think about the people around you, it makes more sense.

What people use AI for

Share of surveyed recent AI users reporting each activity. Multiple responses were possible.

Satisfying curiosity85%
Health and medical questions78%
Understanding unfamiliar concepts71%
Tips and advice70%
News and current affairs55%
Writing code25%

Source: 2026 information-consumption survey by Shenkar, Samuel Neaman Institute, Wikimedia Israel and ISOC-IL, as reported by Ynet and Mako.

AI has become a place to ask everyday questions

People use AI when they don't understand a medical term, want a straightforward explanation or have a question they would once have typed into a search engine. It is quicker than looking through ten different tabs, and people can ask a follow-up immediately.

In the survey, 71% used AI to understand unfamiliar concepts and 70% to get tips or advice. Another 55% asked about news and current events. These activities are closer to ordinary search and learning than to a specialized productivity tool.

Some users now start with AI instead of a search engine

When respondents described how AI had changed their search and learning habits, 73% described a positive change and just 2% a negative one. Speed, convenience and easier learning came up often. In those open-ended responses, 14% explicitly described AI as replacing search engines for them.

That is a self-reported change in habit, not a measured loss of Google queries. It still tells us something useful: for some people, an AI answer has become the first stop.

The distinction matters for websites. A person can learn about a brand, provider or product directly inside an AI response without clicking through to the source. Traffic alone no longer captures every encounter with a business.

The trust gap is hard to ignore

Only about 38% of AI users said they frequently or always verify answers against another information source. Just 53% recognized that AI can produce incorrect information.

Those numbers are especially important alongside the health finding. AI can be useful for explaining medical terminology and helping people prepare questions for a professional. But a convincing answer is not necessarily an accurate one, and health decisions deserve reliable evidence.

The researchers also found a difference in verification habits among Wikipedia readers: 55% of daily readers cross-check AI answers, compared with 33% of people who never read Wikipedia. That is an association, not evidence that Wikipedia use itself causes better verification.

What this means for AI search visibility

At Aiso, we study how brands appear in AI answers and what people ask along the way. The survey is a useful reminder that commercial discovery often starts with an informational question. Someone might begin by asking about a medical symptom, a tax rule or an unfamiliar destination, then ask for providers and products later.

That creates a practical content problem for brands: can an assistant find an accurate, specific explanation when the user first encounters the topic? And when a user eventually asks for a recommendation, is the brand supported by evidence beyond its own website?

It's also a measurement problem. A brand mention in an educational response is different from a recommendation. A recommendation is different from a visit or a sale. These should not be merged into a single visibility score. For a closer look, see our guide to how AI questions affect citation sources and our analysis of how to compare AI search with Google.

What the study can and cannot tell us

The survey included 1,226 participants in total. The adult sample contained 1,071 respondents surveyed in Hebrew and Arabic using online quota sampling, with 155 teenagers aged 15 to 17 in a separate comparison group. The teenage group was not part of the adult statistical tests.

These are reported behaviors and attitudes, not logs of actual AI conversations. They describe people surveyed in Israel and should not be treated as a global estimate. Use-case percentages refer to surveyed AI users, rather than everyone in the country.

Even with that scope, the pattern is striking. Curiosity and health outrank coding by a wide margin. AI is already part of how people find explanations. Our ability to check those answers has not caught up with the convenience of getting them.

Sources