Who Gets News From AI Chatbots?

Q: Who Gets News From AI Chatbots?

9% of US adults. Highest: Asian at 19%. Lowest: 65 and over at 4%. Source: News Platform Fact Sheet. Fieldwork to 2025-09-25.

CurrentPublished 2025-09-25 · 12 months oldVerified against the source 2026-09-09

9% of US adults. It runs from 4% among those aged 65 and over to 19% among Asian adults, a spread of 15 points. AI chatbots entered this survey in 2025 and already show the widest racial spread of any news channel.

By age

18 to 2913%+4n=480 ±5.6

30 to 4911%n=1,399 ±3.4

50 to 647%n=1,274 ±3.6

65 and over4%-5n=1,813 ±3

By race and ethnicity

White6%-3n=3,304 ±2.3

Black13%+4n=512 ±6

Hispanic14%+5n=757 ±5

Asian19%+10n=211 ±8.9

By education

High school or less10%n=1,175 ±3.8

Some college8%n=1,587 ±3.4

College graduate9%n=2,215 ±2.7

By gender

Men8%n=2,194 ±3

Women9%n=2,758 ±2.5

By party

Rep / lean Rep9%n=2,234 ±2.8

Dem / lean Dem9%n=2,446 ±2.8

This is one figure cut one way. The studio combines traits, shows the arithmetic behind any combination it estimates, and says plainly when nothing measures what you asked.

Build an audience in the studio

Source: News Platform Fact Sheet, published 2025-09-25. Subgroup margins are shown on each row. % who say they get news at least sometimes from the named platform.

What this is, and what it is not

Measured means a named study published that figure for that population, and the study is cited with its sample size, field dates and margin of error. Those numbers are theirs, not this site's, and are reproduced under the terms each publisher sets.

Estimated means this site combined published figures to answer a question nobody measured directly. The arithmetic is printed on every such row so it can be checked or rejected. It assumes traits shift the odds independently of one another, which is rarely exactly true, so an estimate is a direction rather than a measurement and its error is wider than the sampling error shown.

Gaps are real. Where no source measures something, this says so instead of filling it in. A trait marked inferred has no evidence behind it at all and exists to be tested, not believed.

Everything here describes US adults in aggregate at the time of the fieldwork shown. It says nothing about any individual, it is not a basis for decisions about any individual, and it must not be used for eligibility, credit, employment, housing, insurance or any other decision about a person. Surveys age; the date on each figure is the date it was collected, not the date you are reading it.

Provided as is, for research and planning, with no warranty of accuracy, completeness or fitness for any purpose, and no liability accepted for any decision, spend or outcome based on it. Verify anything load-bearing against the primary source, which is linked from every figure. Naming a source is not a claim that they endorse this analysis; they do not.

Independent research, not affiliated with or endorsed by any platform or publisher named. Trademarks belong to their owners. This is a plain-language statement, not legal advice. Full method, sources and changelog.

Every other question, answered

Same corpus, same discipline, one question each. The studio answers any combination of them, and the data page prints every cell.

Which platforms they use2025 data

How many people use them2025 data

Where they get news2025 data

How they get online2025 data