We searched “cancer cure” on Amazon.com (10,000+ search results) and retrieved the top 1000 English-language book search results, including sponsored books, on August 3, 2023. Similar to other studies, we performed the search without logging into the platform to avoid the search biases of an existing profile [34]. The Amazon.com search function matches customers to products that fit their typed description. Our search mimics the searches of patients with cancer who may be seeking cancer cure information on Amazon.com. To collect the data, MZ created 2 Data Miner scrapers [35] that retrieved the product information of each result, including its title, author, description, rating score, number of ratings, price, year published, number of pages, publisher-supplied “about the author” information, and the page number of the Amazon.com search result that the book appeared in. Data Miner is a program that allows its user to create simple scrapers that automatically collect certain fields from a web page. The first Data Miner scraper collected the URLs of each result listed from the “cancer cure” Amazon.com search. The second Data Miner scraper was created to collect the aforementioned web page fields (eg, book name) from each of the URLs collected from the first scrape.

MZ reviewed the books’ descriptions and titles to determine whether the books provided misleading cancer cure or treatment information. As the books are for sale and their titles and descriptions represent their value to a potential customer, we considered this information appropriate to determine whether the books for sale are misleading or not. We considered a book to be misleading if it suggested scientifically unsupported cancer treatment approaches that cure or meaningfully treat cancer. To determine which books are misleading, we consulted a list of alternative and natural treatments named in a study examining which alternative cancer treatments patients use [36], along with the National Cancer Institute’s list of alternative and complementary cancer treatments [37]. While it is not possible to identify every unproven cancer cure, our sources contain the standard and most well-known examples. We then applied the lists against the results. We additionally reviewed books for misleading cues. This included overt statements that are contrary to scientific or clinical consensus and present conspiratorial narratives. To ensure consistent inclusion decisions, TC blindly audited 10% of inclusion decisions. TC found no disagreements, demonstrating high agreement and consistent coding. TC additionally reviewed the code applications MZ marked as requiring a second opinion.

We conducted a qualitative, inductive latent thematic analysis [38] to determine the informational value the books sought to offer to potential customers. We followed the thematic analysis steps outlined by Clarke and Braun [39], consisting of data familiarization, generation of initial codes, searching for themes, reviewing themes, defining and naming themes, and reporting. To increase the trustworthiness of the analysis, the authors adopted the recommendations of Nowell et al [40]. MZ, GH, TC, and NM reviewed the data, undertook data review, and iteratively drafted and redrafted a coding frame until a final frame captured themes. MZ coded the data. GH audited nearly half (236/494, 47.8%) of thematic book coding, finding agreement with 95.8% (1,808/1,888) of code decisions, signifying consistent and accurate coding. All disagreements were resolved through discussion. GH reviewed the thematic code applications MZ marked as requiring a second opinion. Our study followed the Standards for Qualitative Research Reporting guidelines (Multimedia Appendix 1) [41].

This study did not require ethical approval because all data collected and analyzed are publicly available and posted without the expectation of privacy.

Our analysis found that 49.4% (494/1000) of the sampled “cancer cure” books for sale on Amazon.com contained misleading cancer treatment or cure information. Misleading books were consistent throughout the search result pages (Figure 1), with 17 (51.5%) out of 33 Amazon.com results pages having 50% or more of the books coded as misleading. Results in the first 3 pages contained the highest percentage of misleading books (first page: 23/33, 69.7%; second page: 23/33 69.7%; third page: 22/33, 66.7%). The lowest percentage of misleading books across any page found was 35.5% (11/31).

These results demonstrate that misleading cancer cure and treatment books are for sale, visible, and seemingly prevalent on Amazon.com. Nearly half of the “cancer cure” books for sale offered misleading and potentially harmful cancer treatment misinformation. Misleading books directly claimed to have efficacious cures for cancer, undermined scientifically supported treatments, misapplied scientific reasoning, oversimplified cancer and cancer treatments, and promoted conspiracy theories. Books offered information that may delay or encourage patients to opt out of best-standard treatments and create false hope.

Our study contributes to the increasing research documenting the presence, spread, and mechanisms of medical misinformation on Amazon. Notably, Amazon’s algorithm has been reported to amplify books with vaccine misinformation [23,42]. In other cases, Amazon has been found to host products selling fake autism cures [43] and promote COVID-19 misinformation [25]. Amazon’s search results have been found to rank products with misinformation higher than sources debunking misinformation [34]. Concerningly, misinformation in Amazon-hosted products may be amplified by fake product reviews, which are associated with a causal increase in product sales [44,45].

Amazon effectively gives legitimacy affordances [46] to authors without medical credentials to freely write medical information and advice books. Previous studies have documented how other technology platforms, like Meta and Google, may inadvertently enable persons selling disproven or unproven cancer therapies to appear to be legitimate medical providers [10]. Many books in our study were self-published, thus undergoing limited or no external or publisher review [22]. Amazon may enable such books to be put up for sale, advertised, matched to user search queries depending on keywords, and displayed alongside books written by qualified authors. The blurred differentiation between books offering misleading or credible information creates unfair difficulties for patients with cancer evaluating sources of web-based information during a highly stressful time in their lives.

The presence of misleading cancer cure books for sale on Amazon occurs despite Amazon policies that prohibit non–FDA-approved medical claims [21]. Presumably, this policy would include books whose market value is derived from the cancer treatment information they offer, although this is not immediately clear. The findings of the study may support Amazon to expand or revise its existing policies to protect its users. The presence of misleading cancer cure books also occurs after previous criticism of Amazon for allowing the sale of books with misleading information during the COVID-19 pandemic [47,48]. This context suggests that Amazon’s disinformation and misinformation policies related to the sale of health-related books are at this time limited, inadequately protecting its users.

Amazon benefits from a distinct form of “platform power” [49] that complicates regulators’ abilities and their political support to act upon health misinformation. Platform power refers to the “tacit allegiance” of a platform’s consumers. Confronting the mechanisms that make Amazon successful, such as its algorithms, affordances to sellers, advertisements, and convenience, is likely not in the interest—and would feel beyond the power—of many Amazon users. This also limits the public support of political entities to act upon health misinformation on Amazon. Therefore, while receiving this distinct platform power, Amazon, like other big tech platforms, may not be incentivized [50] to act upon health misinformation until its users demand such actions or until its business interests are threatened. It has been argued that digital platforms, which are reliant on automated, algorithmically driven, paid advertising for income, are themselves important commercial determinants of health [50,51]. Furthermore, they may have an inherent conflict of interest, where meaningfully curbing exploitation might require greater resources committed to oversight and regulatory compliance or reductions in the profitability of certain platform features. In light of their scale, power, and contributions to the wider digital ecosystem, such platforms may have cumulative harmful effects on the spread and consistent appearance of such misinformation. This may also occur in ways that make the independent investigation of effective cancer treatment, by those who may have just received a diagnosis, highly challenging in inequitable ways. Inherent to these issues is the tension between the informational needs of the public and their right to be protected from the harms associated with the spread of misinformation on the one hand, and the interests of corporations like Amazon to maximize profits on the other hand.

Our results and those of other studies documenting harmful web-based ecosystems for patients with cancer suggest that cancer-focused organizations should engage in preemptive treatment misinformation debunking [52]. Dispelling cancer misinformation before patients are exposed to it can result in a form of psychological inoculation against ineffective treatments [53]. Patients with cancer, especially those with terminal prognoses, are inevitably susceptible to exploitive actions from bad actors selling false hope [54]. Web-based technology platforms, such as specific social media platforms, Amazon, or Google, inadvertently assist such bad actors in selling their products [10]. Health care providers and cancer support organizations are well placed to preemptively debunk such content, ensuring a first-line defense against cancer misinformation. Proactive surveillance of cancer treatment misinformation and health literacy initiatives can help curb misinformation effects.

Our study has several important limitations. We collected and analyzed the publicly available information from a book’s Amazon.com listing page. It was not feasible nor ethical to purchase and read the entire book’s content. However, the product information collected, including the book’s title and description, is suitable for analysis due to its conveyance of purpose and content. In deciding whether the book was misleading or not, we relied upon external lists and signifiers of alternative cancer treatments. We sought to mitigate this limitation by auditing misleading coding application decisions. Finally, as Amazon.com search results are personalized based on prior behavior, it is possible that recommendations or highlighted products may appear even more frequently than reported here for users who previously engaged with such products or those closely related to them.

Future research into cancer misinformation on Amazon could include a focus on 3 areas in seeking to assess the potential dangers of such books: best sellers, advertising, and impact. Amazon has several lists for general and specific types of cancer, each of which has its own “best sellers” list that evolves over time [55]. It is prudent to track the best-selling, cancer-related “information” books for sale on Amazon for misinformation. Second, this study did not examine paid advertising of books containing cancer misinformation. Future research should determine the extent of paid advertising involving misleading cancer cure books on Amazon. Finally, the impact of misleading cancer books on their readers could be evaluated using survey or focus group methods. Other data sources, such as the reviews of misleading cancer cure books available on Amazon, could also be worthy of study for accuracy, provenance, and potential manipulation, to determine whether and how book reviews disseminate cancer misinformation.