We constructed a term list consisting of a combination of “Coronavirus,” “COVID-19,” “COVID19,” and “COVID” with misinformation terms obtained from the WHO Mythbusters website [8]: wine, hot weather, antibiotics, chlorine, garlic, ginger, sun, 5G, hydroxychloroquine, pepper, houseflies, mosquito, hand dryer, supplement, and saline. We selected topics that were clearly defined and for which there was available data. For example, someone searching for “COVID AND Alcohol” might be interested in the amount of alcohol needed to make hand sanitizers at home and not necessarily trying to verify the claim that drinking alcohol might cure or prevent COVID-19. Topics such as 5G are reported to have spread quickly through social networks on Twitter [16]. Additionally, searches for terms such as mosquito and hand dryers were prevalent in some countries and completely absent in others.

After assessing the quality of search data across the countries, we focused on four misinformation topics: claims that (1) drinking alcohol (specifically, wine) increases immunity to COVID-19; (2) sun exposure prevents spread or that COVID-19 is less likely to spread in hot, sunny areas; (3) home remedies may prevent or cure COVID-19; and (4) COVID-19 is spread via 5G cellular networks. We also discussed searches for hydroxychloroquine separately because, unlike the other misinformation terms, it did not appear on the WHO website until July 31, 2020. There was also much confusion about its potential benefit while it was being evaluated by clinicians, unlike the other misinformation topics.

We focused on 8 English-speaking countries from five continents: Nigeria, Kenya, South Africa, the United States, the United Kingdom, India, Australia, and Canada. Weekly search data was obtained from the Google Trends application programming interface from December 2019 to October 2020.

We assumed the search data represented trends in exposure to misinformation. This implies that if someone is seeking information on a particular misinformation topic, they have been exposed to it. However, we cannot deduce a person’s intent or whether or not they believe the misinformation. Additional data is needed to deduce the personal motives of individuals engaging in searches (see Discussion section). We inferred the week of the first peak and then fitted an exponential growth model to both sides of the time series curve: before and after the peak. The exponential regression model is defined as follows: log(y) = r × t × b, where y represents searches (or postings) for the misinformed phrase, and r, t, and b represent the growth rate, number of days since Google reported a search volume greater than one, and the intercept, respectively. This approach was implemented in the Incidence package in the R software (R Foundation for Statistical Computing) and used for analyzing incidence data for epidemics [17,18]. We compared the start week, doubling time, and first peak across the 8 countries and four topics. We referred to the peak of the search data as the initial peak, since similar to an epidemic, there can be multiple peaks. New searches might be initiated at a later time during the pandemic.

Searches for “coronavirus AND 5G” started at different times but peaked in the same week for 6 of the countries (Figure 1). For example, searches for 5G in Australia, the United Kingdom, and Canada were first reported during the week of January 19, 2020, while searches in South Africa, India, and the United States started the following week. In contrast, searches were recorded from Kenya and Nigeria a month later, during the weeks of February 16 and 23, respectively. Despite the different start dates, India, Australia, Canada, Kenya, Nigeria, and the United States observed the first search peak during the week of April 5, 2020. The United Kingdom and South Africa observed a peak during the same week: March 29, 2020.

Similarly, searches for “coronavirus AND ginger” started in the same week in several countries. The United States, the United Kingdom, Canada, Australia, and India noted initial searches during the week of January 19, 2020. However, initial searches in South Africa, Nigeria, and Kenya occurred several weeks later, during the weeks of February 9, February 23, and March 8, respectively. The peak week was earliest for the United Kingdom (March 22), followed by Canada (March 29) and the United States and Australia (April 5). South Africa, Nigeria, India, and Kenya peaked during the weeks of April 12, April 12, April 19, and April 26, respectively. Searches did not always grow exponentially after the initial week. For some countries such as Nigeria, zeros were noted during 1 or 2 consecutive weeks after the initial search. This might be due to Google’s scaling algorithm and might not represent no searches during those weeks.

Furthermore, searches for “coronavirus AND sun” started in the United States, Canada, Australia, India, and the United Kingdom during the week of January 19, 2020, in Nigeria and South Africa during the week of January 26, and a month later in Kenya (February 23). Multiple countries noted a peak during the same week: the week of March 15 for the United States, South Africa, and Canada; March 22 for Australia, the United Kingdom, and Nigeria; and April 12 for India and Kenya.

Lastly, search trends for “coronavirus AND wine” were inconsistent across the 8 countries. Nigeria and Kenya had low search volume and were therefore excluded. The United States noted the earliest searches during the week of January 12, 2020. Other countries noted initial search during the weeks of January 19 (Canada, India, the United Kingdom), January 26 (Australia), and February 9 (South Africa). The peak weeks also differed across the countries with no obvious groupings across regions. The peak weeks were March 15 (Canada and South Africa), March 22 (the United Kingdom), April 5 (Australia and the United States), and April 12 (India).

Searches for 5G had the shortest doubling time across all misinformation topics (Figure 2). Nigeria and South Africa noted the shortest doubling time: between 4 and 5 days. Searches for ginger doubled at approximately the same rate for the United Kingdom and the United States. Searches for sun doubled much more slowly in Canada compared to the other countries. The confidence interval was also wider, suggesting sparse searches in these contexts and less confidence in the estimates. Similar observations were noted for searches of ginger in Australia. Searches for wine were more prevalent in the United Kingdom, the United States, India, and Australia. The data for Kenya, Nigeria, Canada, and South Africa were sparse.