Science1 publisher2 min readPublished
Facebook health ads borrowing fringe-group wording drew more clicks in 11 of 12 tests
Researchers studying Australian Facebook groups found a vaccine-information ad drew about 52% more clicks when worded as a doubt-raising question. The paired design isolates wording well, though its outcome is the click, so any effect on trust or vaccination remains unmeasured.
The Scientist · Science desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- The team built three personas, a young mother, a retired teacher and a fitness-minded young man, after months watching Australian anti-vaccine and far-right Facebook groups.
- Each persona-styled ad was paired with a neutral version sharing the same image, government resource link, sponsor and ad settings, so only the wording differed.
- Wording that echoed the communities' own language drew more clicks in 11 of the 12 paired comparisons.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- decision A health agency already paying for Facebook ads can now treat wording as a variable to test, running neutral and community-voiced versions against the same page before scaling either.
- constraint With vaccination-ad click rates under 1%, a large relative lift still moves few people, and the evidence stops short of showing changed trust or vaccination decisions.
- exposure Borrowing doubt-raising questions puts that question in front of everyone who scrolls past, and those readers never see the accurate page the ad was paid to deliver.
The design rules out the usual confounds in an ad test. A better picture, a more appealing destination or different ad settings cannot explain the gap, because all of those were held fixed between each pair [1]. "Our findings suggest the wording matters, even when the information itself stays the same," the researchers wrote [2].
According to the researchers' own account, republished by phys.org, the 52% figure comes from one pair of messages [4]. It set an instruction, "Protect the community and your kids by booking their vaccinations today," against a question, "Do you know all the effects of childhood vaccinations on your babies?" [4]. Questions were one device among several. Ads written for the young-mother persona used personal, caring language. The retired teacher's took an authoritative, explanatory tone, and the fitness enthusiast's were motivational, urging readers to check the facts for themselves [7]. Matching a community's register is a broader claim than copying one misinformation trick. It fits what the team saw in the groups, where messages could sound like a friend sharing worries about their child [15].
The denominators are small. Familiar-language vaccination ads were clicked 0.98% of the time [10], under one click per hundred showings [18]. At rates like that, a large percentage lift adds only a few clicks per thousand impressions. The school-education ads moved more, gaining 4.9 percentage points over their neutral twins [16]. One of the 12 comparisons did not favour the familiar wording [17].
The thing this doesn't tell you is what happened after the click. A click is reach. The authors name the next barrier themselves: fact-checking can help, they wrote, but "people need to trust the source and understand the explanation." [11] I'd also want to see what the question version does to people who scroll past without clicking. For them, the doubt-raising question is the whole message.
Reach is the problem the study was built to test [3]. Meta stopped using independent fact-checkers in the United States in April 2025 and began introducing Community Notes [12]. Users got around earlier platform restrictions with coded language and links to less moderated platforms [13]. The researchers put the open problem plainly: "How do you get reliable information in front of someone who isn't looking for it?" [14]
In my view the evidence supports one narrow change for anyone buying health ads on Facebook. Run a neutral version and one written in the audience's own register against the same page, and compare. The conditions are specific: one platform, groups observed in one country [5], and an outcome that stops at the click.
What to watch
- Whether the researchers report post-click outcomes, such as time on the government page or stated vaccination intent, for the familiar-language ads.
- Replication of the paired-wording test outside Australian Facebook groups, on other platforms or in other countries.
- Whether health agencies publish their own A/B results after adopting persona-styled wording in campaigns.