Household Pulse Survey, September 2026
Trust in health information from AI tools by age
In the 25 to 34 group, 47% answered “Very little”; in the 45 to 54 group, 27% did. Across the table the answers differ by age, a small association: χ²(12, N = 12381) = 149.35, p < .001, V = .06. These are unweighted respondents, not population estimates.
How results are checked
Named statistical review pending. Selected automated reference cases are published on the validation page.The table
The table
Question asked: In general, how much would you trust information about health or medical topics from AI tools such as ChatGPT, Google Gemini, or Claude?
Shown by age, as column percentages with the number of respondents beneath.
| Trust in health information from AI tools | 25 to 34 (A) | 35 to 44 (B) | 45 to 54 (C) | 55 to 64 (D) | 65 and over (E) | All |
|---|---|---|---|---|---|---|
| A great deal | 2.2%25 | 3.0%48 | 3.1%56 | 2.7%68 | 2.3%122 | 2.6%319 |
| Quite a lot | 11.1%128 | 14.7%a238 | 16.9%A311 | 17.9%AEb446 | 15.1%A801 | 15.5%1,924 |
| Some | 40.1%461 | 50.3%A815 | 52.5%Ae964 | 50.0%A1,243 | 48.9%A2,589 | 49.0%6,072 |
| Very little | 46.6%BCDE535 | 32.0%C519 | 27.5%504 | 29.3%728 | 33.6%CD1,780 | 32.8%4,066 |
| Respondents | 1,149 | 1,620 | 1,835 | 2,485 | 5,292 | 12,381 |
A letter beside a percentage names the columns it is significantly higher than (uppercase 95%, lowercase 90%; pairwise column tests, Bonferroni-corrected). 1,663 of 14,044 respondents did not answer one of the two questions and are left out.
Analysis
Results
Chi-square test
χ² = 149.35
df = 12, p < .001
Effect size
Cramér's V = 0.063
a small association
APA-style report: χ²(12, N = 12381) = 149.35, p < .001, V = .06. N = 12,381. Smallest expected count 29.6.
How to read it
The chi-square test rejects independence at the 0.05 level (p < .001): differences this large between the groups are unlikely if the answer did not depend on age. Cramér's V of 0.063 is in the small range, which is real but modest: knowing someone's group tells you only a little about their answer.
With 12,381 respondents, even small differences reach statistical significance. Read the percentages and the effect size before the p-value.
Limits
- The file carries a person weight (PWEIGHT) and replicate weights. These tables are unweighted respondent counts, so they describe the people who answered, not the U.S. adult population.
- The test treats respondents as a simple random sample. The survey's design (weights and replicate weights) would widen the uncertainty, so treat borderline results with care.
- A difference between groups is not evidence that group membership causes the answer.
Next steps
Work with this table
Open this table in the calculator
The link fills in every count and label. Merge answers, drop a group, and the statistics update.
Open this question in the full file, weighted
The workspace opens all 14,044 respondents in your browser with this question selected and the person weight (PWEIGHT) applied, so the percentages estimate U.S. adults. Change the group, add a filter or export from there.
Open in the workspace →Run this on your own data — free, no signup
Upload a CSV, Excel, SPSS, Stata, SAS, R, Parquet or JSON file. Everything runs in your browser; your file never leaves your device.
Open the workspace →The same question by other groups
Other questions that differ by age
Source
U.S. Census Bureau. Household Trends and Outlook Pulse Survey, September 2026 public use file (HTOPS_HPS_2609). Tabulated by crosstabs.com. Variables HLTHTRUST_AI (Trust information about health or medical topics from Super Intelligence or SI (formerly collected as Artificial Intelligence or AI) tools) and TAGE. The Census Bureau asks people who make their own estimates from its data to cite it as the source of the original data only; conclusions are the responsibility of whoever does the analysis.