The data used for this research are located on public platforms (ie, consumer review sites). Therefore, informed consent for this study was not obtained. However, as the use of direct quotes from consumer reviews could potentially make them identifiable, the quotes that were illustrated in this study were minimally amended to ensure users’ anonymity. This study was approved by the National University of Ireland Galway Research Ethics Committee (reference number R20.JUN.12).

To date, most research that aims to understand experiences using social robots has traditionally been researcher driven [10]. By contrast, this study utilized the large volume of information uploaded by users of the JfA cat onto publicly accessible online consumer review sites. These sites contain a sizeable body of anecdotal evidence from users who have purchased and used lower costed pet robots. These individuals shared detailed accounts of their experiences, for the primary benefit of other potential users who might be seeking to gather information about the product. Examining this valuable source of information during the study was an opportunity to develop knowledge shifting away from regarding researchers and health care professionals as the sole producers of information toward eliciting the voice and empowerment of nonprofessionals [21]. This approach has been used in other research fields, such as business or consumer research, however it is a novel methodology in the field of health and social sciences which allowed for an examination of user-led content.

Data collection involved 3 key steps. First, online consumer review sites were identified through a Google search, using the search terms “Joy for All cat” and “user review”. The researcher’s (WK) internet browsing history and cookies were cleared, and the search was conducted in the incognito mode. Next, the first 100 consumer sites identified from the Google search that contained consumer reviews of the robotic cat were selected as data collection sites. All reviews were manually extracted into Microsoft Excel. This step was essential to ensure a clear audit trail, as the content of a webpage may change depending on what the researcher searches for and researcher’s location [21]. Consumer reviews of all languages that were submitted up to July 24, 2020, were extracted using a standardized data extraction form (Multimedia Appendix 1) containing the following data fields: (1) review title, (2) review text, (3) star rating given, and (4) review date. Demographic information about users of the technology, such as their age group, diagnoses, and setting, was also collected if these data were available. If these were not available, the data field was left empty. To ensure anonymity, no potentially identifying information, such as the reviewing authors’ name and photo attachments, was collected. Finally, all reviews were screened to identify the sampling frame for data analysis.

Reviews were included if they contained information about the use of the robotic cat for older adults or people with dementia in any settings and were published in the English language.

As not all reviews contained information regarding users’ age and diagnoses, innovative approaches had to be undertaken to ensure that all relevant reviews were adequately considered for inclusion. First, as the average age of becoming a grandparent is between 50 and 69 years in several countries [22-24], it seemed reasonable for the researcher to include reviews that mentioned about the use of the robotic cat for this group (ie, grandparents) as older adults. Next, reviews that contained information about the use of JfA cat in institutional care were also included, as the large majority of people living in assisted living facilities or care homes are of an older age group [25-29]. Hence, reviews that met any of the following inclusion criteria were included in the sampling frame:

All reviews that did not meet these inclusion criteria were excluded. Reviews that were included were cleaned and formatted on Microsoft Excel before being exported into NVivo 12 (QSR International) for data analysis.

Descriptive statistics was applied to characterize the number of reviews, available demographic information about users of the JfA cat, and the average star ratings given by users. Textual data were qualitatively analyzed using inductive content analysis, as described by Hsieh and Shannon [30], on the NVivo12 software. This method of data analysis was chosen as it guides systematic categorization of large volumes of text-based data and facilitates the identification of patterns of occurrences [31].

The data analysis proceeded as follows: First, 3 coders (WK, SW, and PH) immersed themselves in the data by reading all data repeatedly to obtain a sense of the whole and to allow new insights to emerge [31,32]. The first 5% of reviews were read word by word by each coder, who independently generated key thoughts or concepts for each phrase, and labeled them using descriptive and low-inference codes [33,34]. After that, all coders met to discuss similarities and differences, and agreed on codes that formed the initial coding scheme [30]. Next, this coding scheme was tested by WK, SW, and PH, who independently coded another 10% (n=137) of all data using the coding scheme. Data that did not fit into an existing code were assigned a new code. After this, intercoder reliability test (ICR), using the kappa coefficient (κ), was conducted to assess the similarity between the coding produced by the authors. Although there is no set consensus on what proportion of data should be analyzed to yield a reliable estimate of ICR [35], an analysis of 10%-25% of the data set is typical [36]. Conducting this test allowed the rigor and transparency of the coding framework to be ascertained [36-38]. The kappa coefficient of 0.60 was obtained, which demonstrated substantial agreement between coders [39]. Following this, all coders met to discuss and agree upon the final coding framework. In particular, they ensured that all data within the codes and categories were distinctive and that they had good coherence [40,41]. The final coding scheme (Multimedia Appendix 2) was tested by WK and SW, who independently coded another 5% (n=66) of the data set. Strong intercoder reliability was established (κ=0.7). Thereafter, the coding framework was applied to the remaining reviews by WK. Research rigor was ensured through prolonged engagement with the data [42], and frequent meetings with all coders throughout the creation of the coding framework, and to develop and refine the codes and categories.

Figure 3 shows the flowchart that reports the data identification and collection. A total of 100 websites were identified, of which 15 were consumer review sites for the JfA robotic cat (Table 1).

A total of 2445 consumer reviews were submitted over a 5.5-year period from December 4, 2015, to July 24, 2020. Of these, 1327 reviews met the inclusion criteria and were included for data analysis. Most reviews were derived from consumer sites from the United States (n=948), Canada (n=132), the United Kingdom (n=80), and Australia (n=13). Most reviews contained information about review date and star rating (n=1309). Overall, the number of reviews increased steadily from 2015 to 2020, and its average star rating was 4.75 (Table 2).

Information about the review authors and users is presented in Table 3. Most review authors were family members of the primary users of the JfA cat. The majority were children (n=770), grandchildren (n=120), and partners (n=52) of older adults or people with dementia. Only 2% (n=22) of all reviewers identified themselves as users of the robotic cat. Information about the relation of other review authors with the older person or person with dementia was not available in 247 (18.61%) cases.

The JfA cat was described as being for the use for older adults in 44.16% (586/1327) of reviews, while 51.77% (687/1327) described their use for people with dementia, cognitive impairment, or memory issues. The majority (n=1109) contained information about users’ gender, of which 89.09% (n=988) were females. Less than half (n=539) provided explicit information about the setting in which the device was used (Multimedia Appendix 3). Most were used in care settings, including long-term care facilities (n=399), specialized memory care facilities (n=56), retirement homes (n=16), or other care facilities (n=49).

This is the first study to use a novel web-based approach to explore the usability and impact of a low-cost robotic pet for older adults and people with dementia, based on perceptions and experiences of its use. Most of the review content was derived from consumer sites that were based in the United States, and most reviewers were family members of older adults and people with dementia. Overall, most reviewers had positive perceptions and experiences of using the JfA cat and found it to be beneficial and practical for older adults and people with dementia. Nevertheless, not all were satisfied with this technology.

Users’ previous experiences of pet ownership were frequently reported as a circumstantial reason for purchasing the JfA cat for the intended user. This finding aligns with previous findings that users’ like of animals influenced their acceptance of a robotic pet [43]. Therefore, it may be worth screening users’ likes and dislikes of animals as a predictor to gauge their acceptance of the robotic pets [44]. Reviewers also acknowledged that pragmatic deterrents, such as institutional regulations and a lack of capacity to care for a live animal, propelled them to seek robotic alternatives. This echoes the proposition that a recognition of the relative advantage of an innovation can facilitate its adoption [45].

Most perceptions about the JfA cat were positive, which suggests its design as a familiar animal was acceptable. In previous studies, familiarly designed robotic animals, such as the JustoCat and the NeCoRo cat, were also well received by older adults and people with dementia [46,47]. Likewise, other studies have highlighted preferences for familiarly designed pet robots [20,48,49]. These findings contrast with the notion that people are more likely to accept less familiarly designed robots because they would have fewer prior conceptions or expectations [50]. However, this hypothesis has not been widely evaluated, as few studies have investigated design preferences of older adults or people with dementia. Indeed, in most research studies, participants were typically given a single pet robot to engage with, which was selected based on the needs of the research rather than the preference of the participants. In line with a person-centered approach to care [51], older adults and people with dementia should be given the autonomy to choose their preferred robotic pet design. People with dementia, especially in the advanced disease stages, may not be able to articulate their preferences for pet design. However, they should still be given opportunities to participate in decisions relating to their care [52], to allow for the maintenance of self-identity, dignity [53], and personhood [54]. Moving forward, more considerations should be made to identify pet robot design preferences of individuals.

Use of the robotic cat offered older adults and people with dementia opportunities to participate in meaningful activities. Older adults and people with dementia participated in an array of activities with the JfA cat, such as talking to it and about it, cuddling, and stroking it. These findings resonate with results from studies which used other robotic pets [46,48,55,56], suggesting the potential of the JfA cat to elicit similar activities. Other activities identified included brushing the cat, sleeping with it, and taking it to places. Some reviewers supported these meaningful activities by getting a brush for users to brush the cat, and getting a cat bed and a personalized collar to allow for easier identification in care facilities. Such activities were not reported in previous studies and appeared to be unique to this study. This might be attributed to more opportunities for interaction with the cat over an extended period, made possible due to individuals owning their own robotic cat and not sharing it with others. Individual ownership may have provided users with the opportunity to take ownership of the robotic pet and be actively involved as care providers, in contrast to their traditional role as passive recipients of care [57]. Furthermore, the consistent and proximate presence of the JfA cat might have enabled such additional activities involving its use to be scaffolded naturally.

The relationship between engagement in meaningful activities and health outcomes has been established [58-62]. Similar to findings from previous studies [10-12,18], participating in activities with the JfA cat elicited positive emotions among users, and also provided comforting and calming effects. This is an important finding, because it highlights the potential of the JfA cat to elicit therapeutic benefits that are similar to costlier and more technologically sophisticated robotic pets. This raises an important question—In consideration of potential cost benefits, what degree of technological sophistication is required for a robotic pet to be therapeutic? Further research and randomized controlled trials should be conducted to evaluate and compare the effectiveness of low-cost robotic pets on the mental and social health of older adults and people with dementia, with other more technologically advanced robots.

The movements and vocalizations of the JfA cat appeared to be perceived positively by users as behaviors of reciprocity. Reciprocity, or the give and take that occurs between individuals, can influence the maintenance of social relationships [63,64]. This may explain why interactive robotic pets have been able to elicit more user engagement as compared with noninteractive or plush alternatives [65,66]. Interestingly, the lack of predictable responses to touch and movement was interpreted by some users as resemblant of a live cat’s behavior, and was well received. Nevertheless, the JfA cat’s interactive features also resulted in some negative impacts, particularly among those who perceived it as a live animal. When the robotic cat ran out of batteries, some users exhibited emotional distress as they perceived it to be dead. The meowing sounds worried or caused annoyance to some users, who sometimes perceived the robotic cat to have unmet needs. Similar issues have also been raised previously in relation to other robotic pets [13,48,67,68]. Furthermore, some users became concerned that the cat was not eating and attempted to feed it. These issues may be due to individual ownership of the robotic cat, where perceived responsibility for pet care may place a burden on people with cognitive impairment [69]. In such instances, reviewers provided mediation and support. This suggests that unattended, prolonged interactions with the robotic pet may have the potential to cause negative impacts. In turn, this raises the question as to what amount of robot–human interactions, especially for people with cognitive impairments, should be conducted completely without the support of caregivers. Findings from this study suggest some degree of facilitation and mediation by caregivers may still be necessary.

The JfA cat also positively impacted caregivers, providing them with a sense of relief and positive emotions, which included feelings of happiness and contentment. There is currently a lack of research that has focused on how robotic pets impact caregivers. More research is needed to increase understanding, especially since one of the key premises for developing social robots is to supplement and support the care of older people with dementia [66].

Finally, despite the overall positive perceptions and experiences, some reviewers reported negative opinions about the cat’s design. This included comments about its “hardness” and lack of sophistication, such as audible mechanics during movements and unrealistic “meowing” sounds. These issues did not appear to influence most users’ interaction with the robotic cat, suggesting that reviewers may have a higher expectation than the end users in wanting the robotic cat to behave more realistically and autonomously. Nevertheless, these issues resulted in the rejection of this technology by a minority of users. Comments about the robustness of the technology were mixed, with some reviewers being dissatisfied with its durability. Some elaborated that users’ handling of the JfA cat, such as holding it too tightly or dropping it, affected its functioning. The relatively short longevity of the device has potential to cause negative impacts such as emotional distress, especially among users who have developed an attachment toward it [70]. The understanding of such issues are useful to inform future robot development to ensure technological robustness [18].

Despite the valuable new knowledge that was generated through this study, there are limitations that should be acknowledged. Data that were used for this study were self-reported information that was gathered through publicly available sources. The anonymity of users makes it difficult to verify the authenticity of the content, and to verify the ages and diagnoses of the users of the robotic cat. Most reviewers were family members, and as such, their perceptions and experiences might differ from actual opinions of the primary end user (ie, older adults or people with dementia). Although most included reviews were shown as verified purchases, it is not possible to confirm the authenticity of review or distinguish potentially deceptive reviews. There could also be a bias in terms of the representation of data, as not all consumers will upload their reviews on consumer websites. Nevertheless, given the analysis of the large number of reviews from multiple websites across a 5-year period, as well as the richness of the data contained in these reviews, it may be reasonable to infer that the findings from this study represent real-world perceptions and experiences of using the JfA cat for older adults and people with dementia.

This study provides important knowledge about the usability and impact of a low-cost robotic pet for older adults and people with dementia based on perceptions and experiences of its use. It analyzed user-driven content to access a unique perspective toward an understanding of this phenomenon. We found that circumstantial reasons, such as inability to care for a pet, have prompted the use of the robotic cat, and that familiarly designed robotic pets can be accepted by older adults and people with dementia. Although the JfA cat is less technologically advanced than other robotic pets, its interactive features were generally well received. Use of the JfA cat facilitated participation in meaningful occupations, as it provided older adults and people with dementia opportunities to participate in various activities. These activities elicited positive psychosocial impacts on both users and caregivers. Nevertheless, facilitation by caregivers may be necessary to monitor for and mitigate potential negative impacts. Although perceptions and experiences were mainly positive, negative aspects of the JfA cat’s design and interactivity were raised. Experiences of its durability were also mixed, which highlights the need to improve the technical robustness of this device.

These insights are vital in helping researchers, robot developers, and clinicians to understand the viability of using low-cost robotic pets to benefit older adults and people with dementia. Future research should consider evaluating design preferences for nonfamiliarly versus familiarly designed robotic pets. It will also be valuable to conduct a randomized controlled trial to compare the impacts of low-cost robotic pets with other more technologically advanced robotic pets, to understand any similarities or differences of their impacts on the mental and social health of older adults and people with dementia. A process evaluation may also be conducted to identify factors that may explain any outcome variations. This has the potential to influence equal access to technology if their impacts on the psychosocial health of users are comparable.