Three recent polls on loneliness, dating apps and AI romance show what to check before you believe a number about love
Every week, a new statistic about love arrives: how many people are lonely, how many use dating apps, how many consider an AI romance to be cheating. Some come from careful research, and some come from companies with products to sell. Telling them apart is a skill worth having. This guide uses three recent examples to show what to check. The first is the Human Connection Survey for Match Group by Ipsos, reported by Medianews4U. The second is a poll on AI romance reported by ProPakistani. The third is a 2024 report on lonely young adults, whose headline and body did not match.
Question One: Who Paid for It?
The Human Connection Survey was sponsored by Match Group, which operates dating apps. The AI romance poll was sponsored by AI Girlfriend Coach, a review site that tests AI companion apps. The 2024 loneliness report was by Social Discovery Group, which operates over 70 online dating and social discovery sites and apps. In each case, the sponsor has a commercial interest in the subject. That does not mean the findings are false. It means they should be read with the question in mind: would this finding be convenient for the sponsor, and were the questions framed in a way that favours it?
Independent academic studies have their own pressures, such as the need to publish, but they typically undergo peer review and disclose methods. A sponsored poll that is also run by a reputable polling firm, like Ipsos in the Match Group survey, is better than one with no named fieldwork partner. Still, the firm carries out the questions the client asked for.
Question Two: Who Was Asked?
The Match Group survey asked 1,000 single adults aged 18 to 45 in major metropolitan and Tier 1 cities in India. That is a reasonable sample size, but it describes urban singles, not all Indians, and not people who are partnered, older or rural. The AI romance poll began with 2,150 US adults, excluded 441 suspicious responses and analysed 1,709. Excluding bad responses is good practice, but it raises the question of how the respondents were recruited. The 2024 report did not state its sample size at all, which is a red flag.
Question Three: What Exactly Was Measured?
Look for the wording. A statement that 63 percent of Indian singles say dating apps help them feel more connected and less alone is a statement about opinion, not about outcomes. It tells you what people believe, not whether the apps reduce loneliness. Similarly, saying that 50.5 percent of Americans consider a partner’s AI romance to be cheating describes an attitude to a imagined situation, not behaviour. Surveys are good at measuring attitudes and poor at measuring cause.
Also check whether the numbers in the headline match the numbers in the article. In the 2024 report, the headline cited 61 percent of young adults feeling isolated, but the body never mentioned that figure and gave 67 percent as a different statistic about turning to dating platforms and AI to reduce loneliness. When a headline and a body disagree, the safest course is to find the original report.
Question Four: Is It a Snapshot or a Trend?
A single survey cannot show change over time. The Human Connection Survey is a one-time snapshot, as is the AI romance poll, which its own report describes as measuring attitudes at a single point in time in one country. A claim that something is rising or falling needs data from more than one period collected in the same way. Where a company says that time spent on its platforms rose 35 percent since 2022, that may be true of its own users while telling you little about society.
Question Five: Does the Interpretation Go Too Far?
Surveys that find an association are often described as if they show a cause. A finding that people who use apps report more connection could mean that apps help, or that more socially active people use apps, or that both reflect something else. Longitudinal studies that follow the same people over time, like the 2,149-adult study of chatbots and loneliness covered elsewhere in this series, can say more about direction, though still not about proof.
A Quick Checklist
Before sharing a statistic, check the sponsor, the sample size and how people were recruited, the exact wording of the question, the date, whether the numbers in the headline match the body and whether the claim goes beyond what the data can show. If you cannot find the original report, treat the number as provisional. If the finding is surprising, look for a second source.
What Good Evidence Looks Like
Better evidence tends to come from studies that state their hypotheses in advance, share their data or methods, use large and varied samples, measure outcomes over time and report limitations honestly. The authors of the chatbot study, for instance, described their own analyses as exploratory and urged caution. That kind of candour is a sign of good faith.
None of this means commercial surveys are useless. They can show what people say they want, flag emerging concerns and prompt better research. The aim is to use them for what they are. Our London pages have long enjoyed poking at statistics about love, including coverage of ten leading AI models answering a question, couples who met online and a dating expert’s advice on loneliness. See also our guide to UK political comedy shows, and for American coverage Bohiney Magazine.
For some lighter reading, Newsthump has the news of the day.
SOURCE: bohiney.com