How Many Lonely Americans Depends on Who Answered the Phone

Ask what share of American adults are lonely and the answer ranges from about one in three to well over half, depending on which survey supplied it. The 2023 Surgeon General’s advisory put the figure near half. The Cigna Group’s 2025 report, describing fieldwork from mid-2024, found 57%. A 2025 analysis of the federal Health Information National Trends Survey found roughly 37% reporting moderate-to-severe loneliness. None of these figures is wrong. They are answers to different questions, asked of different people, by different means.

The gap matters because national loneliness figures increasingly function as inputs to policy. The World Health Organization’s Commission on Social Connection cited a global loneliness burden in its 2025 report; the CDC’s Morbidity and Mortality Weekly Report treats its own 2022-derived estimate as a surveillance baseline. When a number is used to justify a national strategy, the sampling frame behind it deserves the same scrutiny as the effect size.

Who gets asked, and how

Start with mode. The Cigna Group’s 2025 report drew on fieldwork conducted between 29 May and 14 June 2024, surveying more than 7,500 U.S. adults, and reported 57% experiencing loneliness. The CDC’s MMWR estimate, published in June 2024 but describing 2022 data, comes from the Behavioral Risk Factor Surveillance System and a related instrument, a different data collection architecture entirely, built for public health surveillance rather than a proprietary corporate wellness index. The HINTS-6 study, published in the American Journal of Preventive Medicine in November 2025 but drawing on 2022 data, used the National Cancer Institute’s Health Information National Trends Survey, a nationally representative mail and web panel designed around health communication rather than loneliness specifically.

Three different vehicles, three different years of underlying data, three different published dates for that data. A reader treating all three as current, comparable snapshots of “how lonely Americans are right now” would be wrong on both counts: the data is not current, and the instruments are not the same.

The instrument question compounds the problem. The AARP Foundation’s 2018 survey of adults 45 and older used the 20-item UCLA Loneliness Scale, the same validated instrument used across the academic literature going back decades, which is precisely why the Center’s earlier coverage has treated AARP’s figures as unusually comparable to peer-reviewed research. Its 2025 follow-up, Disconnected, used the same instrument again, which is what makes a longitudinal claim about rising loneliness among this age group credible rather than an artifact of a new question wording. Cigna’s index, by contrast, is a proprietary composite the company has run since 2018, not the UCLA scale, and its year-over-year comparisons are internally consistent but not directly transferable to studies using a different tool. The BMC Public Health review of the field’s state of research in 2023 named inconsistent measurement as one of the central barriers to comparing findings across studies. That is not a minor caveat. It means the entire debate about whether loneliness is rising, falling, or flat nationally is partly a debate about which yardstick was used, dressed up as a debate about the underlying phenomenon.

The frame itself

Sampling frame is a separate issue from instrument, and it gets less attention. A phone or web panel recruited through address-based sampling reaches a different slice of the population than an opt-in online panel incentivized to complete surveys quickly. Adults without stable housing, without reliable internet access, or without English fluency are underrepresented in nearly every general-population survey on this list, and they are plausibly among the most isolated people in the country — which would suggest most national estimates undercount rather than overcount loneliness among the most affected groups. None of the surveys cited here claim to have solved this. The Harvard Making Caring Common report from 2021, which found 36% of Americans reporting serious loneliness and 61% of young adults aged 18-25 reporting the same, was fielded during the acute phase of the pandemic; its sample and its historical moment are both unrepeatable, which limits its use as a baseline for anything other than that specific period.

Age structure of the sample frame also drives headline numbers in ways that are easy to miss. Cigna’s 2025 report found Gen Z and Millennial respondents reporting more loneliness than Gen X and Baby Boomer respondents, despite greater technological connectedness among the younger cohorts. A survey oversampling working-age adults recruited online, where younger respondents are easier to reach and cheaper to convert, will produce a higher topline loneliness figure than a survey built around a landline-plus-cell RDD frame skewing older, purely as a function of who picked up. This is not an accusation against any specific survey on this list. It is a structural feature of how national loneliness numbers get built, and it is rarely disclosed prominently enough for a reader to adjust for it.

What the range actually tells us

The honest reading is not that one of these figures is correct and the others are noise. It is that “loneliness prevalence” is not a single measurable quantity in the way that, say, unemployment is defined by a fixed and widely agreed operational rule. The Surgeon General’s 2023 advisory settled on “approximately half” as a summary figure precisely because it was synthesizing across sources with this range, not because a single authoritative survey produced that number.

What can be said with more confidence, because it holds across instruments and frames, is the direction of every gradient. Younger adults report more loneliness than older adults in nearly every survey design examined here, from Harvard’s 2021 pandemic-era figure to Cigna’s 2025 report to Gallup’s global work. That consistency across differently constructed samples is more informative than any single topline percentage, because a pattern that survives changes in mode, instrument, and sampling frame is much less likely to be an artifact of any one of them.

The same logic applies to the AARP comparison over time. Because the 2018 and 2025 AARP surveys of adults 45 and older used the same UCLA scale on a comparably constructed sample, a reported rise in loneliness in that population between the two waves is a claim the Center is willing to treat as a real finding rather than a measurement artifact. That is a rarer situation in this literature than it should be.

What would settle it

A single, harmonized national loneliness module fielded consistently across a general-population sample, using a validated instrument such as the UCLA scale, repeated on a fixed schedule with the same recruitment method each wave, would let the field distinguish real trend from instrument churn. Something close to this already exists for isolated years within HINTS and BRFSS, but neither uses the same instrument as the academic gold standard, and neither has run long enough on a fixed design to serve as a trend line on its own.

Until then, the responsible way to cite a national loneliness figure is to name the instrument, the sampling frame, and the year the data was actually collected — not just the year the report was published. A reader who sees “57% of Americans are lonely” and “37% of Americans are moderately-to-severely lonely” cited in the same paragraph without that context will reasonably conclude the research is incoherent. It is not incoherent. It is measuring adjacent but distinct things, through different windows, and reporting them on different clocks.