Abstract
Background
Crisis Pregnancy Centers (CPCs) have expanded their digital presence using social media to promote services, advance ideological messages, and disseminate potentially medically inaccurate claims. Despite extensive use of platforms such as Instagram for outreach, little empirical work has examined the content and engagement patterns of CPC‐related posts. We systematically analyzed Instagram posts to identify key themes, assess engagement patterns, and determine factors influencing the reach and impact of CPC‐related content.
Methods
We conducted a mixed‐methods content analysis of publicly available Instagram posts published between June 24, 2022 and October 15, 2025. A total of 716 posts met inclusion criteria. Posts were coded using an iteratively developed codebook that yielded four overarching themes and 13 subthemes. Hashtags were classified into service, wellness, advocacy/identity, and location categories. Engagement outcomes (likes and comments) were log‐transformed to address skew. Multivariable ordinary least squares (OLS) regression models estimated associations between content characteristics and engagement, controlling for account features, post timing, and hashtag use.
Results
Service Promotion & Outreach was the most common theme (49%), followed by Emotional Appeals (23%), Ideological Messaging (18%), and Misinformation & Misrepresentation (10%). Posts containing misinformation or misrepresentation received significantly higher engagement, with a 46% increase in predicted likes compared with Service Promotion & Outreach posts (β = 0.38, p < .05). Log followers was strongly associated with engagement (β = 0.68, p < .001). Reels received substantially more likes than still‐image grid posts. Certain posting months and wellness‐ or location‐based hashtags were associated with lower engagement. Subtheme analyses revealed that content portraying abortion as immoral or harmful and posts exposing or correcting CPC misinformation generated the highest predicted engagement. Results for comments followed similar directions but were generally smaller in magnitude and not statistically significant.
Conclusions
On Instagram, CPC posts containing misinformation and misrepresentation received disproportionately higher engagement compared to service‐oriented content. These findings highlight the visibility advantage of misleading content in reproductive health contexts. They also underscore the need for proactive public health communication strategies that counter misinformation in the reproductive health ecosystem.
Keywords: crisis pregnancy center, misinformation, pregnancy, qualitative research, reproductive health services, social media
1. INTRODUCTION
Crisis Pregnancy Centers (CPCs) present themselves as providers of pregnancy support services, but are widely understood to operate with the primary purpose of discouraging abortion. Frequently referred to as “pregnancy resource centers,” CPCs are non‐healthcare organizations that receive funding from religiously affiliated groups and federal and state budgets [1, 2]. In 2024, 2633 CPCs operated in the United States, an increase of 3.3% from 2021 [3]. Although CPCs advertise free services, centers frequently provide misleading, incomplete, or medically inaccurate information about reproductive health [4, 5]. Despite this concern, one survey study estimated that up to 20% of ever‐pregnant individuals have attended a CPC [6]. In some areas of the United States, CPCs are commonly found in maternity care deserts and may be positioned as the sole source of reproductive health information in the area [7, 8, 9, 10]. Finally, CPCs have been shown to delay access to medically legitimate prenatal and abortion care [11].
Following the Dobbs decision in June 2022, CPCs have become more visible actors in the reproductive health ecosystem [4, 12, 13]. As digital communication has become central to how individuals seek health information [14, 15], CPCs have increasingly expanded their activities onto social media platforms, where strategic messaging can rapidly reach individuals searching for pregnancy‐related support [16, 17]. Marketing firms specializing in CPC promotion openly target individuals searching for abortion‐related information on social media platforms, emphasizing tactics that optimize discoverability and shape user perceptions [18]. Social media platforms also host a parallel and equally important stream of user‐generated content from individuals who received CPC care [14]. Collectively, CPC‐generated content and user‐generated experiences create a dynamic informational environment in which persuasive messaging, testimonial narratives, and misinformation coexist and potentially influence reproductive decision‐making.
We conducted a systematic content analysis of post‐Dobbs CPC‐related Instagram posts, examining (1) CPC‐created posts that promote services, convey ideological messages, or present medically inaccurate information, and (2) user‐generated posts describing personal encounters with CPCs. By combining qualitative content analysis of themes, emotional appeals, ideological framing, and accuracy of CPC‐related messaging with quantitative modeling of engagement metrics, we provide a comprehensive characterization of CPC communication strategies on social media and offer insights into how such messaging may shape public understanding of reproductive health options.
2. METHODS
We used a mixed‐methods design that integrated qualitative content analysis with quantitative regression modeling to systematically assess CPC‐related social media content. Our methodological approach is consistent with other recent social media analyses [19, 20, 21]. We collected publicly available Instagram posts because this platform is a primary target for CPC outreach and engagement. To minimize personalization from prior user activity, we created new Instagram accounts solely for data collection [22]. Because these accounts had no prior search history, engagement history, followers, or followed accounts, they reduced the likelihood that retrieved content reflected user‐specific algorithmic personalization rather than the study search terms. We collected posts from June 24, 2022 (the day the Dobbs decision was rendered) to October 15, 2025 to ensure that the data reflect CPC messaging and user experiences in the post‐Dobbs landscape. We retrieved posts containing relevant keywords (e.g., pregnancy resource center, free pregnancy services, free pregnancy ultrasound, free pregnancy clinic) and hashtags (e.g., #CrisisPregnancyCenter, #PregnancyHelp, #ProLife, and #ReproductiveRights).
For the purposes of this study, we operationally defined a CPC as an organization or account that advertised pregnancy‐related support services while not providing referral for abortion services and that appeared to promote continuation of pregnancy. CPC‐affiliated accounts were identified using one or more standardized indicators: use of common CPC self‐descriptors such as “pregnancy resource center,” “pregnancy help center,” “pregnancy care center,” or “women's center”; promotion of free pregnancy tests, ultrasounds, STI testing, counseling, parenting classes, material assistance, or post‐abortion support; absence of abortion provision or referral information; and messaging that framed abortion negatively or emphasized alternatives to abortion. When account affiliation was ambiguous, coders reviewed the account biography, linked website when available, captions, hashtags, and service descriptions to determine whether the account met the operating definition. Posts from individuals were included only when the caption or visual content explicitly referred to an encounter with a CPC, pregnancy resource center, or comparable organization meeting these criteria.
We included English‐language posts with content that originated from CPC‐affiliated accounts or individuals explicitly discussing their experiences with CPCs. Exclusion criteria include advertisements without user‐generated content and posts from private or restricted‐access accounts, such as closed groups. For each post, we extracted the caption, post type (grid post vs. reel), date published, number of likes and comments, user profile characteristics, and all hashtags appearing in the caption. In total, 716 posts met the inclusion criteria and were included in the analytic dataset. Duplicate posts appearing under multiple keywords were removed.
To ensure systematic identification, we conducted structured keyword and hashtag‐based searches at regular intervals throughout the study period [23]. Searches were performed manually using Instagram's native search functionality because Instagram does not provide a public Application Programming Interface for full‐text retrieval. When a search returned more than 100 posts for a given term, we deferred to a saturation approach and reviewed the most recent 100 posts, as Instagram sorts results primarily by recency and engagement rather than at random. This process resulted in a comprehensive sample of posts that reflected the range of content disseminated by CPC‐affiliated accounts and individuals reporting CPC experiences.
2.1. Coding process
We developed the initial set of codes deductively based on the published literature about CPC communication strategies, misinformation, and reproductive health communication [3, 4, 6, 11, 13]. The initial codes included nine themes and 20 subthemes, grouped into categories such as service promotion, emotional appeals, moral or religious messaging, decision‐framing, and medical misinformation. Before formal coding began, both coders completed a training phase using 40 posts not included in the analytic dataset. This phase allowed coders to pilot the initial codebook, discuss ambiguities, and refine operational definitions for each theme and subtheme. Consensus meetings were held iteratively during training and again after double‐coded posts were reviewed. During these sessions, the team clarified decision rules for distinguishing between overlapping constructs. In particular, we clarified definitions of emotional appeals, ideological content, and decision‐framing, which ensured stable code application across the full sample.
Two coders independently reviewed a subset of posts to refine code definitions and example indicators. To ensure consistency, we double‐coded 20% of posts (n = 144). Coders independently assigned a theme and subtheme from the initial 20‐category subtheme structure. The coders agreed on labels for 133 out of 144 posts, yielding a 92% agreement rate, which exceeds conventional thresholds for qualitative reliability in applied social research [24]. The team discussed and resolved codes for the remaining eleven posts and made refinements to both code definitions and the decision rules governing classification.
2.2. Themes and subthemes
To characterize the communicative strategies used by CPCs on Instagram, we applied a structured content analysis. Through iterative review of the initial coding system, coder discussions, and consolidation of conceptually overlapping categories, all posts were ultimately classified into four overarching themes comprising 13 subthemes. Three potential subthemes from the original scheme were excluded because no posts were coded into them.
The consolidation process followed an inductive–deductive procedure. First, the team reviewed coder memos and discrepancies from double‐coding to identify conceptually redundant or rarely used subthemes. Subthemes that reflected similar communicative functions were merged. Three subthemes were eliminated because no posts were coded into them during primary coding. The final coding structure therefore retained only those subthemes that were both conceptually meaningful and empirically represented in the data. Each overarching theme was operationalized through specific subthemes, enabling finer‐grained analysis of how CPCs deployed varied communication strategies on social media.
2.3. Hashtag classification
Hashtags were extracted from every post caption and compiled into a master list of unique tags. We developed a structured, rule‐based classification system to categorize hashtags into conceptually meaningful groups. The groups formed an iteratively constructed keyword dictionary that reflected common themes in how CPCs use hashtags to position themselves, target specific audiences, and shape their platform visibility.
We matched hashtags to each category using deterministic rules based on the keyword dictionary and followed with a manual review of ambiguous cases. Across all posts, we classified a total of 2709 hashtags into the four categories, and each post received counts representing the number of hashtags belonging to each category. Although classification relied primarily on deterministic keyword matching, we conducted a manual audit of all hashtags assigned to each category to confirm accuracy. Ambiguous hashtags (e.g., pregnancyjourney, choosejoy) were reviewed in context and assigned a category based on the predominant semantic function observed across posts. Because hashtags apply platform‐level indexing rather than sentence‐level meaning, we prioritized consistent functional classification over literal interpretation of individual words.
2.4. Data analysis
We conducted descriptive analyses to summarize account characteristics (e.g., follower count, user profile type), post attributes (e.g., post type, posting date), thematic codes, subthemes, and hashtag classifications. We included publication year, month, and day of week as temporal control variables, which allowed us to account for variation in engagement that may reflect platform use patterns, current events, holidays, or changes in public attention to reproductive health issues during the post‐Dobbs period. Continuous variables (likes, comments, follower counts, and hashtag counts) were evaluated for normality and log‐transformation (log(x + 1)) was applied to normalize the distribution of likes and comments prior to regression analysis.
Bivariate analyses were used to examine differences in engagement across post characteristics and thematic categories. The primary multivariable approach used ordinary least squares (OLS) regression with log‐transformed likes as the dependent variable. Independent variables included the four major themes, log‐transformed number of followers, post type (grid vs. reel), publication year, month, and day of the week, and counts of hashtags in each of the four hashtag categories. To enhance interpretability of the log‐transformed coefficients, we calculated percent change values for all predictors in the primary OLS model. Percent change was computed by exponentiating each coefficient and subtracting one (eβ − 1) × 100, which reflects the percentage difference in expected likes associated with a one‐unit change in each predictor. Percent change values are presented alongside raw coefficients and standard errors. Subthemes were entered as mutually exclusive categorical indicators in a fully adjusted OLS regression model with log‐transformed likes and comments as the outcomes. We calculated adjusted predicted values of log‐transformed likes and comments for each subtheme.
Prior to model interpretation, we confirmed that key OLS assumptions were met. Due to heteroskedasticity, all regression models were estimated with robust standard errors. Model coefficients were substantively unchanged with robust estimation, confirming that inference was not sensitive to heteroskedasticity. As a sensitivity analysis, we estimated negative binomial regression models using raw engagement counts (likes and comments) as outcomes. Results were substantively consistent with the primary OLS models. Because the OLS models with log‐transformed engagement produced more stable confidence intervals and easier‐to‐interpret predictive margins, those results are presented as the primary analysis. All analyses were conducted using Stata SE version 19 (StataCorp LLC).
2.5. Ethical considerations
Only publicly accessible posts were analyzed. Usernames and other identifying information were not reported. The study protocol was approved as non‐regulated research by the Institutional Review Board.
3. RESULTS
The final four themes represented distinct communication domains observed in CPC messaging: Service Promotion & Outreach, Emotional Appeals, Ideological Messaging, and Misinformation & Misrepresentation (Table 1). Examples of service promotion and outreach were testimonials that CPCs provided “all the proper information so you know all your options,” with targeted outreach toward families in need. Appeals to emotion were posts urging the audience to recognize inspiration in the miracle of life. Ideological messaging was conveyed in posts claiming the Bible states that abortion is wrong. Examples of misinformation were claiming childbirth may provide long‐term relief from severe menstrual cramps or promotion of “abortion pill reversal” as possible within a 24‐h window. An example of misrepresentation was listing ultrasound, STI testing, pregnancy testing, education, resources, and support without clarifying that abortion care and referral for comprehensive reproductive care are not provided.
TABLE 1.
Crisis pregnancy center themes and subthemes in Instagram posts collected from June 24, 2022 to October 15, 2025.
| Theme | Subtheme | Description |
|---|---|---|
| 1. Service Promotion & Outreach | 1A. Service Listing | Promotion of free services (e.g., pregnancy tests, ultrasounds, STI testing), resources, or appointment scheduling. |
| 1B. Reassurance of Care | Messages emphasizing compassion, nonjudgment, “we're here to help,” or emotional support from staff. | |
| 1C. Appeals to Urgency / Uncertainty | Posts framed around crisis pregnancy, immediacy, fear of the unknown, or the need for rapid decision‐making. | |
| 1D. Decision Framing / Counseling Tone | Messaging that presents CPC counseling as neutral while subtly framing abortion as less desirable. | |
| 1E. Digital Engagement / Optimization | Platform‐oriented strategies (hashtags, algorithm‐optimized captions, SEO‐style language) to increase visibility. | |
| 2. Emotional Appeals | 2A. Positive Emotional Appeals | Encouragement, empowerment, hope, strength, community, or celebration of pregnancy. |
| 2B. Negative Emotional Appeals | Fear, grief, shame, uncertainty, or posts leveraging distress‐based emotional influence. | |
| 3. Ideological Messaging | 3A. Anti‐Abortion / Religious Messaging | Religious framing, scripture references, or explicit opposition to abortion. |
| 3B. Abortion as Immoral or Harmful | Claims that abortion causes physical, emotional, or spiritual harm; moral framing. | |
| 3C. Pro‐Life Identity Promotion | Statements reinforcing pro‐life identity, branding, or movement‐building. | |
| 4. Misinformation & Misrepresentation | 4A. Medical Misinformation | False or misleading medical claims (e.g., abortion risks, fetal development misinformation, “abortion pill reversal”). |
| 4B. Misleading Appearance / Advertising | Posts portraying CPCs as medical clinics (white coats, clinical exam rooms) or creating confusion about their medical legitimacy. | |
| 4C. Correction / Exposure | Posts responding to or attempting to correct misinformation about CPCs (rare but present). |
Hashtags fell into one of four mutually exclusive categories: Service‐Related, Wellness‐Related, Advocacy/Identity, and Location‐Based. These categories represented the dominant communicative intent or semantic content of the hashtag (Table A1 in the Appendix).
3.1. Descriptive characteristics of posts
A total of 716 Instagram posts were included in the analysis (Table 2). Most posts were grid posts (92%) and 8% were reels. Posting activity was relatively evenly distributed across and within years, with each month representing between 6.8% and 10% of total posts. On average, posts had 4551 followers, received 84.9 likes (SD = 1159), and 5.0 comments (SD = 83). Advocacy/Identity related hashtags were the most frequently used (mean = 1.79), followed by Service‐Related (mean = 1.02) and Location‐Based tags (mean = 0.86), whereas Wellness‐Related hashtags were rare (mean = 0.11).
TABLE 2.
Descriptive characteristics of Instagram posts collected from June 24, 2022 to October 15, 2025, N = 716 Posts.
| Categorical variables | Category | N = 716 total | % |
|---|---|---|---|
| Post type | Grid | 660 | 92.2 |
| Reel | 56 | 7.8 | |
| Post year | 2022 | 199 | 27.8 |
| 2023 | 201 | 28.1 | |
| 2024 | 159 | 22.2 | |
| 2025 | 157 | 21.9 | |
| Post month | January | 59 | 8.2 |
| February | 72 | 10.1 | |
| March | 60 | 8.4 | |
| April | 63 | 8.8 | |
| May | 54 | 7.5 | |
| June | 50 | 7 | |
| July | 60 | 8.4 | |
| August | 69 | 9.6 | |
| September | 51 | 7.1 | |
| October | 66 | 9.2 | |
| November | 49 | 6.8 | |
| December | 63 | 8.8 | |
| Day of week posted | Sunday | 42 | 5.9 |
| Monday | 112 | 15.6 | |
| Tuesday | 167 | 23.3 | |
| Wednesday | 122 | 17 | |
| Thursday | 121 | 16.9 | |
| Friday | 123 | 17.2 | |
| Saturday | 29 | 4.1 |
| Continuous variables | Mean | SD | Range |
|---|---|---|---|
| Followers | 4,550.90 | 38,135.80 | 29–708,000 |
| Likes | 84.9 | 1,158.70 | 0–28,888 |
| Comments | 5 | 83 | 0–1988 |
| Hashtags: Service | 1.02 | 1.88 | 0–17 |
| Hashtags: Wellness | 0.11 | 0.48 | 0–6 |
| Hashtags: Advocacy | 1.79 | 3.28 | 0–27 |
| Hashtags: Location | 0.86 | 2.65 | 0–11 |
3.2. Distribution across themes and subthemes
Nearly half of posts (49.3%) aligned with Service Promotion & Outreach, followed by Emotional Appeals (22.8%), Ideological Messaging (17.9%), and Misinformation & Misrepresentation (10.1%) (Table 3). The most common subthemes were Digital Engagement/Optimization (21.5%), Positive Emotional Appeals (20.4%), and Anti‐Abortion/Religious Messaging (13.6%). Few posts contained Medical Misinformation (1.0%) or presented Abortion as Immoral/Harmful (1.7%).
TABLE 3.
Themes, subthemes, and engagement statistics of Instagram posts collected from June 24, 2022 to October 15, 2025, N = 716 Posts.
| n (%) | Total likes | Total comments | |
|---|---|---|---|
| Themes | |||
| 1. Service Promotion & Outreach | 353 (49.3%) | 5548 | 133 |
| 2. Emotional Appeals | 163 (22.8%) | 3233 | 104 |
| 3. Ideological Messaging | 128 (17.9%) | 33,583 | 1095 |
| 4. Misinformation & Misrepresentation | 72 (10.1%) | 18,408 | 2244 |
| Subthemes | |||
| 1A. Service Listing | 96 (13.4%) | 1455 | 33 |
| 1B. Reassurance of Care | 54 (7.5%) | 563 | 9 |
| 1C. Urgency / Uncertainty Appeals | 15 (2.1%) | 430 | 16 |
| 1D. Decision Framing / Counseling Tone | 34 (4.8%) | 394 | 6 |
| 1E. Digital Engagement / Optimization | 154 (21.5%) | 2706 | 69 |
| 2A. Positive Emotional Appeal | 146 (20.4%) | 1860 | 33 |
| 2B. Negative Emotional Appeal | 17 (2.4%) | 1373 | 71 |
| 3A. Anti‐Abortion / Religious Messaging | 97 (13.6%) | 31,086 | 1024 |
| 3B. Abortion as Immoral or Harmful | 12 (1.7%) | 684 | 12 |
| 3C. Pro‐Life Identity Promotion | 19 (2.7%) | 1813 | 59 |
| 4A. Medical Misinformation | 7 (1.0%) | 178 | 6 |
| 4B. Misleading Appearance / Advertising | 24 (3.4%) | 2397 | 111 |
| 4C. Correction / Exposure | 41 (5.7%) | 15,833 | 2127 |
3.3. Engagement by theme and subtheme
Engagement patterns differed sharply across themes (Table 3). Posts categorized as Ideological Messaging accounted for over half of all likes (33,583 likes, 55.3% of total) and a substantial portion of comment activity (1095 comments, 30.6%). Posts containing Misinformation & Misrepresentation generated the highest number of comments overall (2244 comments, 62.8%). At the subtheme level, Correction/Exposure content documenting misleading practices produced the highest volume of comments (2127) and was among the most liked type of post (15,833, 26.1%). Anti‐Abortion/Religious Messaging generated the highest number of likes (31,086, 51.1%) and the second highest number of comments (1024, 28.6%).
3.4. Multivariable regression results
In multivariable regression models, several content characteristics were significantly associated with higher engagement (Table 4). Compared with posts categorized as Service Promotion & Outreach, posts containing Misinformation & Misrepresentation generated substantially higher engagement, with an estimated 0.38‐unit increase in log‐likes (p < 0.05), corresponding to an estimated 46% increase in predicted likes after exponentiating the coefficient. Emotional Appeals and Ideological posts did not differ significantly from the reference group.
TABLE 4.
Multivariable regression models examining associations between primary themes, post characteristics, and engagement (likes and comments), N = 716 Instagram posts.
| Predictor | Log likes β (SE) | % Change a | Log comments β (SE) | % Change a |
|---|---|---|---|---|
| Themes (vs. service promotion as referent group) | ||||
| Emotion | −0.021 (0.066) | −2.05% | −0.016 (0.045) | −1.57% |
| Ideology | 0.136 (0.071) | 14.52% | 0.022 (0.053) | 2.23% |
| Misinformation | 0.376 (0.151) * | 45.63% | 0.234 (0.140) | 26.41% |
| Account & post characteristics | ||||
| Log followers | 0.675 (0.062) *** | 96.46% | 0.384 (0.068) *** | 46.80% |
| Grid (vs. reel) | −0.601 (0.147) *** | −45.15% | −0.286 (0.142) * | −24.84% |
| Posting year (vs. 2022) | ||||
| 2023 | −0.158 (0.090) | −14.63% | −0.074 (0.065) | −7.16% |
| 2024 | 0.253 (0.093) ** | 28.80% | −0.060 (0.068) | −5.80% |
| 2025 | −0.088 (0.127) | −8.41% | 0.030 (0.091) | 3.05% |
| Posting month (vs. January) | ||||
| February | −0.031 (0.136) | −3.04% | −0.052 (0.141) | −5.09% |
| March | −0.044 (0.145) | −4.30% | −0.151 (0.151) | −14.04% |
| April | 0.061 (0.137) | 6.31% | −0.139 (0.138) | −12.94% |
| May | −0.183 (0.142) | −16.68% | −0.224 (0.144) | −20.07% |
| June | −0.233 (0.154) | −20.76% | −0.140 (0.162) | −13.10% |
| July | −0.416 (0.166) * | −34.00% | −0.202 (0.150) | −18.28% |
| August | −0.202 (0.141) | −18.25% | −0.168 (0.136) | −15.49% |
| September | −0.286 (0.146) * | −24.87% | −0.163 (0.141) | −15.04% |
| October | −0.178 (0.128) | −16.34% | −0.240 (0.128) | −21.36% |
| November | −0.095 (0.133) | −9.05% | −0.186 (0.132) | −17.00% |
| December | −0.215 (0.141) | −19.38% | −0.143 (0.134) | −13.29% |
| Day of week (vs. Sunday) | ||||
| Monday | −0.263 (0.159) | −23.16% | −0.134 (0.122) | −12.52% |
| Tuesday | −0.177 (0.157) | −16.25% | −0.087 (0.124) | −8.32% |
| Wednesday | −0.318 (0.161) * | −27.24% | −0.150 (0.129) | −13.94% |
| Thursday | −0.172 (0.157) | −15.82% | −0.126 (0.124) | −11.85% |
| Friday | −0.205 (0.155) | −18.50% | −0.084 (0.124) | −8.07% |
| Saturday | 0.091 (0.238) | 9.49% | 0.019 (0.174) | 1.95% |
| Hashtag counts | ||||
| Service hashtags | −0.004 (0.019) | −0.40% | 0.000 (0.014) | 0.02% |
| Wellness hashtags | −0.121 (0.064) * | −11.41% | 0.003 (0.045) | 0.26% |
| Advocacy hashtags | 0.018 (0.016) | 1.85% | 0.024 (0.012) * | 2.46% |
| Location hashtags | −0.064 (0.014) *** | −6.18% | −0.043 (0.010) *** | −4.23% |
| Constant | −1.17 (0.473) * | — | −1.920 (0.409) *** | — |
| R 2 | 0.558 | 0.397 |
Note: OLS models with log‐transformed engagement outcomes. Standard errors in parentheses.
Percent change calculated as: (eβ − 1) × 100.
p < .05.
p < .01.
p < .001.
Platform characteristics demonstrated additional, meaningful associations with engagement. A one‐unit increase in log followers was linked to a 96% increase in expected likes (β = 0.68, p < 0.001). Post format also mattered: grid posts received substantially fewer likes than reels (β = −0.60, p < 0.001), representing 45% fewer expected likes. Post timing, by contrast, only showed modest associations with engagement. Posts published in 2024 had higher engagement than those in 2022 (β = 0.25, p < 0.01), while posts published in July (β = −0.42, p < 0.05) and September (β = −0.29, p < 0.05) of every year had significantly lower engagement than those published in January. Posts shared on Wednesdays of every year received fewer likes compared to those shared on Sundays (β = −0.32, p < 0.05). Among hashtag categories, Wellness‐Related hashtags (β = −0.12, p < 0.05) and Location‐Related hashtags (β = −0.06, p < 0.001) were associated with significantly lower log‐likes. Service‐ and Advocacy‐Related hashtags showed no independent association with engagement.
Results for log‐transformed number of comments (Table 4) showed a similar pattern of influence from account‐ and post‐level characteristics, although thematic content played a more limited role. None of the four major themes showed statistically significant differences in comment volume relative to Service Promotion & Outreach. Account‐ and platform characteristics, however, demonstrated variation in comment volume. A one‐unit increase in log followers was associated with a 38% increase in expected comments (β = 0.38, p < 0.001). Grid posts received fewer comments (β = −0.29, p < 0.05) than reels. Posting month, regardless of year, and day‐of‐week did not demonstrate significant changes in comment volume from the referent groups, respectively. Hashtag patterns showed two notable associations: Advocacy‐Related hashtags were positively associated with comment volume (β = 0.02, p < 0.05), while Location‐Related hashtags were negatively associated with comment volume (β = ‐0.043, p < 0.001). Wellness and Service hashtags were not significantly related to comments. Sensitivity analysis in which we estimated negative binomial regression models using raw engagement counts (likes and comments) demonstrated results that were substantively consistent with the primary OLS models (Table A2 in the Appendix).
In fully adjusted models using robust standard errors, two subthemes were significantly associated with higher predicted log likes: Abortion as Immoral or Harmful (predicted log likes = 3.11; 95% CI, 2.83–3.38; p < 0.01) and Correction/Exposure content (predicted log likes = 3.21; 95% CI, 2.82–3.60; p < 0.01) (Table A3 in the Appendix). No subtheme was significantly associated with predicted log comments.
4. DISCUSSION
In this systematic examination of CPC‐related content on Instagram in the post‐Dobbs era, we identified several key findings. First, we found that engagement was strongly shaped by account size, with each one‐unit increase in log followers on Instagram nearly doubling predicted likes. This is consistent with general patterns in digital marketing and health communication research [14, 15]. Second, Instagram posts generated by CPC‐affiliated accounts and user‐generated posts that were classified under Misinformation & Misrepresentation generated the highest engagement, with 46% more likes than Service‐Oriented posts, even after adjusting for account size, posting date, and hashtag use. Thus, irrespective of intent, CPCs’ most engaging content is not supportive or service‐based messaging. Rather, ideological and misleading posts provide better marketing results even while distorting public understanding of reproductive health.
This pattern parallels broader research on health misinformation, which consistently finds misleading health content more engaging on social media than factual information [25, 26]. In the context of CPCs and reproductive health, our findings suggest that posts mimicking medical authority, exaggerating risks of abortion, or presenting misleading claims about CPC services may systematically benefit from platform engagement algorithms. Whether intentional or not, a dynamic where CPCs reach more people with obfuscated information is highly concerning from a public health perspective [27, 28, 29]. High‐engagement misinformation can create the illusion of consensus or credibility, elevate misleading content in users’ feeds through social media company algorithms, and influence perceptions of reproductive health services [25, 30]. The fact that these posts substantially outperform Service‐Oriented or purely emotional content underscores the importance of monitoring CPC digital strategies, educating the public and policymakers on these strategies, and developing interventions to counteract misleading narratives elevated by social media platforms [28].
Posts using Ideological themes also received higher engagement than Service‐Oriented messages. Although the magnitude of engagement was smaller and did not reach statistical significance in the fully adjusted model, ideological posts constituted over half (55%) of all likes recorded in the study. This volume reflects the role of CPCs in broader anti‐abortion networks: they often function as movement‐aligned organizations that disseminate political identity cues, religious values, or moral arguments [31]. Consequently, pro‐life religious messaging may reinforce anti‐abortion identity networks and shape perceptions of abortion in ways that can affect information‐seeking and care navigation. Elevated user engagement with ideological messaging suggests that CPC social media activity contributes not only to service promotion but also to the reinforcement and circulation of a larger pro‐life identity online.
That two subthemes (Abortion as Immoral or Harmful and Correction/Exposure Content) were significantly associated with higher predicted log‐likes suggests that posts employing strong moral claims or that present themselves as part of a journalistic exposé attract more users. The effectiveness of such themes would make sense, since morality and exposure tactics are well‐documented as such in studies on misinformation and persuasive communication [30]. Conversely, Service‐Oriented posts and Reassurance Messaging showed consistently low engagement, suggesting that CPC services may not drive audience interaction.
When interpreted in the context of broader trends in access to reproductive healthcare and maternity care services, our findings paint a bleak picture. Public health agencies, reproductive health providers, and misinformation monitoring systems should note that the most visible CPC content misrepresents evidence. At best, these posts confuse public health discourse. At worst, they discourage evidence‐based care. There is an urgent need to identify and moderate medically inaccurate content, promote digital literacy initiatives for CPC messaging audiences, and improve transparency in CPC advertising.
Timing variables showed modest associations with engagement, suggesting that CPC‐related Instagram visibility may vary somewhat across the calendar and week. However, these findings should be interpreted cautiously because the study was not designed to identify causal seasonal effects. Rather, the inclusion of year, month, and day of week helped distinguish content‐related engagement patterns from temporal fluctuations in social media activity and public attention to reproductive health.
4.1. Strengths and limitations
A strength of this study is its mixed‐methods approach, which allows for a nuanced understanding of both the content and reach of CPC‐related social media posts. Limitations include the impermanent nature of social media content, although posts were saved for analysis when available. Additionally, privacy settings may restrict access to certain user‐generated content. For example, private discussions and groups are not accessible for analysis. We must rely on publicly available data, which may not fully represent the experiences of all individuals interacting with CPCs and may result in sampling bias. While we manually reviewed profiles to evaluate for potential bots and non‐user‐generated content, alternative methods have been described [32].
5. CONCLUSION
CPCs remain a significant presence in the digital reproductive health landscape. Misinformation posts generated the highest engagement, which reflects broader challenges in the online health information environment, specifically that sensationalism often outperforms accuracy. By identifying the themes that drive engagement and quantifying their impact, this study provides an evidence base for public health interventions aimed at countering deceptive narratives, protecting reproductive autonomy, and promoting access to accurate information in a post‐Dobbs United States.
AUTHOR CONTRIBUTIONS
Jim P. Stimpson: Conceptualization; formal analysis; methodology; writing—original draft; writing—review and editing. Aditi Srivastava: Data collection; writing—review and editing. Ketan Tamirisa: Data collection; writing—review and editing. Emily C. Baudot: Data collection; methodology; writing—review and editing. Kimberley Sladek: Data collection; methodology; writing—review and editing. Emily H. Adhikari: Conceptualization; writing—review and editing.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
FUNDING INFORMATION
The authors received no specific funding for this work.
1.
TABLE A1.
Hashtag categories among Instagram posts collected from June 24, 2022 to October 15, 2025.
| Service‐related | Included hashtags that highlighted CPC‐provided services (e.g., pregnancy tests, ultrasounds, counseling), service accessibility (e.g., free, confidential), or operational attributes of the center. These tags commonly reach individuals actively searching for pregnancy‐related support or resources. |
| Wellness‐related | Captured themes of emotional support, self‐care, healthy pregnancy, or general encouragement. Examples include terms related to one's pregnancy journey, empowerment, mindfulness, and maternal wellness. These hashtags function as broad engagement mechanisms, positioning CPCs as supportive environments rather than clinical or ideological actors. |
| Advocacy/Identity | Encompassed explicitly ideological, political, or identity‐signaling terms (e.g., prolife, chooselife, reproductivejustice, abortionfacts). This category also included expressions of institutional or community identity aligned with pro‐life or religious values, reflecting how CPCs use hashtags to embed themselves within broader activism or ideological networks. |
| Location‐based | Consisted of geographic identifiers used to target local audiences (e.g., city names, neighborhoods, state‐level tags). These hashtags are central to visibility algorithms and community‐specific outreach, helping CPCs appear in local searches and regionally relevant content feeds. |
Abbreviation: CPC, Crisis Pregnancy Center.
TABLE A2.
Negative binomial regression sensitivity analysis for likes and comments among Instagram posts collected from June 24, 2022 to October 15, 2025.
| Predictor | Likes IRR (SE) | p value | Comments IRR (SE) | p value |
|---|---|---|---|---|
| Themes | ||||
| Emotion | 0.85 (0.09) | 0.07 | 0.78 (0.18) | 0.168 |
| Ideology | 1.45 (0.16) | 0.001 | 1.12 (0.23) | 0.536 |
| Misinformation | 2.70 (0.29) | <0.001 | 2.47 (0.47) | <0.001 |
| Log of followers | 1.77 (0.03) | <0.001 | 1.54 (0.05) | <0.001 |
| Post type | ||||
| Grid (ref = Reel) | 0.53 (0.05) | <0.001 | 0.55 (0.10) | 0.004 |
| Post year | ||||
| 2023 | 0.88 (0.08) | 0.13 | 0.76 (0.14) | 0.134 |
| 2024 | 1.25 (0.11) | 0.014 | 1.28 (0.21) | 0.09 |
| 2025 | 0.94 (0.10) | 0.591 | 1.03 (0.18) | 0.854 |
| Post month (ref = January) | ||||
| February | 0.95 (0.12) | 0.707 | 1.03 (0.21) | 0.876 |
| March | 0.90 (0.12) | 0.452 | 1.10 (0.22) | 0.673 |
| April | 1.06 (0.14) | 0.649 | 0.93 (0.19) | 0.754 |
| May | 0.80 (0.11) | 0.086 | 0.67 (0.15) | 0.054 |
| June | 0.77 (0.11) | 0.043 | 0.64 (0.15) | 0.05 |
| July | 0.65 (0.09) | 0.002 | 0.59 (0.12) | 0.024 |
| August | 0.82 (0.11) | 0.135 | 0.73 (0.15) | 0.141 |
| September | 0.75 (0.11) | 0.049 | 0.91 (0.20) | 0.71 |
| October | 0.84 (0.11) | 0.243 | 0.88 (0.18) | 0.527 |
| November | 0.92 (0.13) | 0.591 | 0.67 (0.15) | 0.05 |
| December | 0.79 (0.11) | 0.084 | 0.61 (0.13) | 0.021 |
| Day of week (ref = Sunday) | ||||
| Monday | 0.76 (0.10) | 0.04 | 0.72 (0.15) | 0.137 |
| Tuesday | 0.84 (0.11) | 0.231 | 0.76 (0.15) | 0.199 |
| Wednesday | 0.77 (0.10) | 0.067 | 0.64 (0.13) | 0.036 |
| Thursday | 0.84 (0.11) | 0.238 | 0.68 (0.14) | 0.067 |
| Friday | 0.81 (0.10) | 0.126 | 0.66 (0.13) | 0.04 |
| Saturday | 1.11 (0.22) | 0.537 | 1.16 (0.35) | 0.628 |
| Hashtag categories | ||||
| Service | 0.99 (0.01) | 0.647 | 1.03 (0.03) | 0.426 |
| Wellness | 0.72 (0.11) | 0.04 | 0.69 (0.18) | 0.167 |
| Advocacy | 1.04 (0.01) | 0.058 | 1.04 (0.02) | 0.119 |
| Location | 0.94 (0.02) | 0.001 | 0.95 (0.03) | 0.151 |
| Constant | 0.17 (0.08) | <0.001 | 0.14 (0.10) | 0.09 |
| Dispersion (α) | 3.42 (0.36) | – | 5.71 (0.70) | – |
Note: Coefficients are incidence‐rate ratios (IRRs). p values shown for sensitivity‐analysis transparency. Reference categories: Themes = Service Promotion; post type = Reel; post year = 2022; month = January; day of week = Sunday.
TABLE A3.
Predicted log likes and log comments by subtheme (margins from multivariable linear regression models with robust standard errors) among Instagram posts collected from June 24, 2022 to October 15, 2025.
| Log likes | Log comments | |||||
|---|---|---|---|---|---|---|
| Subtheme | Margin | 95% lower CI | 95% upper CI | Margin | 95% lower CI | 95% upper CI |
| 1A. Service Listing | 2.60 | 2.47 | 2.73 | 0.23 | 0.13 | 0.33 |
| 1B. Reassurance of Care | 2.51 | 2.35 | 2.66 | 0.19 | 0.08 | 0.30 |
| 1C. Appeals to Urgency / Uncertainty | 2.80 | 2.34 | 3.26 | 0.37 | 0.00 | 0.74 |
| 1D. Decision Framing / Counseling Tone | 2.43 | 2.23 | 2.63 | 0.23 | 0.10 | 0.35 |
| 1E. Digital Engagement / Optimization | 2.66 | 2.55 | 2.78 | 0.24 | 0.15 | 0.33 |
| 2A. Positive Emotional Appeal | 2.53 | 2.45 | 2.62 | 0.17 | 0.11 | 0.22 |
| 2B. Negative Emotional Appeal | 3.05 | 2.49 | 3.62 | 0.68 | 0.25 | 1.11 |
| 3A. Anti‐Abortion / Religious Messaging | 2.66 | 2.53 | 2.79 | 0.23 | 0.12 | 0.34 |
| 3B. Abortion as Immoral or Harmful | 3.11 ** | 2.83 | 3.38 | 0.21 | −0.06 | 0.47 |
| 3C. Pro‐Life Identity Promotion | 2.99 | 2.63 | 3.34 | 0.41 | 0.09 | 0.73 |
| 4A. Medical Misinformation | 2.87 | 2.05 | 3.69 | 0.39 | −0.17 | 0.96 |
| 4B. Misleading Appearance / Advertising | 2.68 | 2.25 | 3.12 | 0.28 | −0.03 | 0.60 |
| 4C. Correction / Exposure | 3.21 ** | 2.82 | 3.60 | 0.62 | 0.20 | 1.04 |
Note: Margins were estimated from multivariable linear regression models with robust standard errors. Adjusted for log followers, post type, post year, post month, post day of week, and hashtag categories.
** p < 0.01.
DATA AVAILABILITY STATEMENT
The data sets generated or analyzed during this study are available from the corresponding author upon reasonable request.
REFERENCES
- 1. Rabinovitz, J. R. D. 2024. An Analysis of Federal Funding for Crisis Pregnancy Centers, 2017–2023. Accessed August 13, 2026. https://www.healthmanagement.com/wp‐content/uploads/CPC‐Federal‐Funding‐Report_20240614_hma.pdf. [Google Scholar]
- 2. Office USGA . 2018. Health Care Funding: Information on Crisis Pregnancy Centers, Fiscal Years Through 2024. U.S. GAO. Accessed August 13, 2026. https://www.gao.gov/products/gao‐26‐108137. [Google Scholar]
- 3. Lambert, D. N. , Luisi N., Johnson E. R., Baugh R., and Swartzendruber A.. 2025. “Crisis Pregnancy Centers in the United States: Post‐Dobbs Characteristics and Changes.” American Journal of Public Health 115(6): 924–35. 10.2105/ajph.2025.308055. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Bryant, A. G. , Narasimhan S., Bryant‐Comstock K., and Levi E. E.. 2014. “Crisis Pregnancy Center Websites: Information, Misinformation and Disinformation.” Contraception 90(6): 601–5. 10.1016/j.contraception.2014.07.003. [DOI] [PubMed] [Google Scholar]
- 5. Bryant, A. G. and Swartz J. J.. 2018. “Why Crisis Pregnancy Centers Are Legal but Unethical.” AMA Journal of Ethics 20(1): 269–77. 10.1001/journalofethics.2018.20.3.pfor1-1803. [DOI] [PubMed] [Google Scholar]
- 6. Yang, T. J. K. , Smith M. H., Kavanaugh M. L., Ricks J. M. , and Gallo M. F.. 2025. “Prevalence of Crisis Pregnancy Center Attendance Among Women in Four U.S. States.” PLoS ONE 20(6): e0324228. 10.1371/journal.pone.0324228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Stoneburner, A. L. R. 2024. Nowhere to Go: Maternity Care Deserts Across the US. Accessed August 13, 2026 https://www.marchofdimes.org/peristats/assets/s3/reports/mcd/Maternity‐Care‐Report‐UnitedStates.pdf. [Google Scholar]
- 8. Swartzendruber, A. , Luisi N., Johnson E. R., and Lambert D. N.. 2024. “Spatial Analyses of Crisis Pregnancy Centers and Abortion Facilities in the United States, 2021 (Pre‐Dobbs): Cross‐Sectional Study.” JMIR Public Health Surveillance 10: e60001. 10.2196/60001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Swartzendruber, A. and Lambert D.. 2021. “Crisis Pregnancy Center Map.” Accessed August 13, 2026 http://www.crisispregnancycentermap.com. [DOI] [PubMed]
- 10. Swartzendruber, A. and Lambert D. N.. 2020. “A Web‐Based Geolocated Directory of Crisis Pregnancy Centers (CPCs) in the United States: Description of CPC Map Methods and Design Features and Analysis of Baseline Data.” JMIR Public Health Surveillance 6(1): e16726. 10.2196/16726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Montoya, M. N. , Judge‐Golden C., and Swartz J. J.. 2022. “The Problems With Crisis Pregnancy Centers: Reviewing the Literature and Identifying New Directions for Future Research.” International Journal of Women's Health 14: 757–63. 10.2147/ijwh.S288861. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Rader, B. , Upadhyay U. D., Sehgal N. K. R., Reis B. Y., Brownstein J. S., and Hswen Y.. 2022. “Estimated Travel Time and Spatial Access to Abortion Facilities in the US Before and After the Dobbs v Jackson Women's Health Decision.” JAMA 328(20): 2041–7. 10.1001/jama.2022.20424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Bryant, A. G. and Levi E. E.. 2012. “Abortion Misinformation From Crisis Pregnancy Centers in North Carolina.” Contraception 86(6): 752–6. 10.1016/j.contraception.2012.06.001. [DOI] [PubMed] [Google Scholar]
- 14. Lim, M. S. C. , Molenaar A., Brennan L., Reid M., and McCaffrey T.. 2022. “Young Adults' Use of Different Social Media Platforms for Health Information: Insights From Web‐Based Conversations.” Journal of Medical Internet Research 24(1): e23656. 10.2196/23656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Moorhead, S. A. , Hazlett D. E., Harrison L., Carroll J. K., Irwin A., and Hoving C.. 2013. “A New Dimension of Health Care: Systematic Review of the Uses, Benefits, and Limitations of Social Media for Health Communication.” Journal of Medical Internet Research 15(4): e85. 10.2196/jmir.1933. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Fuller, S. A. 2023. “Teens Need Access to High‐Quality, Comprehensive Digital Information About Reproductive Health Options.” Adolescent Health & Well‐being: Children's Hospital of Philadelphia Policy Lab. Posted July 6, 2023. Accessed Sept 27, 2026. https://policylab.chop.edu/blog/teens‐need‐access‐high‐quality‐comprehensive‐digital‐information‐about‐reproductive‐health.
- 17. Waring, M. E. , Blackman Carr L. T., and Heersping G. E.. 2023. “Social Media Use Among Parents and Women of Childbearing Age in the US.” Preventing Chronic Disease 20: E07. 10.5888/pcd20.220194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Choose Life Marketing . 2022. Social Media Services for Pregnancy Centers. Accessed August 13, 2026. https://www.chooselifemarketing.com/services/digital‐marketing/social‐media/. [Google Scholar]
- 19. Fu, J. , Li C., Zhou C., Li W., Lai J., Deng S., Zhang Y., et al. 2023. “Methods for Analyzing the Contents of Social Media for Health Care: Scoping Review.” Journal of Medical Internet Research 25: e43349. 10.2196/43349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Stimpson, J. P. , Srivastava A., Tamirisa K., Kaholokula J. K., and Ortega A. N.. 2025. “Crisis Communication About the Maui Wildfires on TikTok: Content Analysis of Engagement With Maui Wildfire‐Related Posts Over 1 Year.” JMIR Formative Research 9: e67515. 10.2196/67515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Loeb, S. and Malik R.. 2022. “Methodologies in Social Media Research: Where We Are and Where We Still Need to Go?” Journal of Clinical Oncology 18(8): 533–5. 10.1200/OP.21.00871. [DOI] [PubMed] [Google Scholar]
- 22. Pretorius, K. 2024. “A Simple and Systematic Approach to Qualitative Data Extraction From Social Media for Novice Health Care Researchers: Tutorial.” JMIR Formative Research 8: e54407. 10.2196/54407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Andreotta, M. , Nugroho R., Hurlstone M. J., Boschetti F., Farrell S., Walker I., and Paris C.. 2019. “Analyzing Social Media Data: A Mixed‐Methods Framework Combining Computational and Qualitative Text Analysis.” Behavior Research Methods 51(4): 1766–81. 10.3758/s13428-019-01202-8. [DOI] [PubMed] [Google Scholar]
- 24. Cohen, J. 1960. “A Coefficient of Agreement for Nominal Scales.” Educational and Psychological Measurement 20(1): 37–46. 10.1177/001316446002000104. [DOI] [Google Scholar]
- 25. Avram, M. , Micallef N., Patil S., and Menczer F.. 2020. “Exposure to Social Engagement Metrics Increases Vulnerability to Misinformation.” The Harvard Kennedy School Misinformation Review 1(5). Accessed Sept 27, 2026. 10.37016/mr-2020-033. [DOI] [Google Scholar]
- 26. Vosoughi, S. , Roy D., and Aral S.. 2018. “The Spread of True and False News Online.” Science 359(6380): 1146–51. 10.1126/science.aap9559. [DOI] [PubMed] [Google Scholar]
- 27. John, J. N. , Gorman S., Scales D., and Gorman J.. 2025. “Online Misleading Information About Women's Reproductive Health: A Narrative Review.” Journal of General Internal Medicine 40(5): 1123–31. 10.1007/s11606-024-09118-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Pennycook, G. , Epstein Z., Mosleh M., Arechar A. A., Eckles D., and Rand D. G.. 2021. “Shifting Attention to Accuracy Can Reduce Misinformation Online.” Nature 592(7855): 590–5. 10.1038/s41586-021-03344-2. [DOI] [PubMed] [Google Scholar]
- 29. Swartzendruber, A. , English A., Greenberg K. B., Murray P. J., Freeman M., Upadhya K., Simpson T., et al. 2019. “Crisis Pregnancy Centers in the United States: Lack of Adherence to Medical and Ethical Practice Standards; a Joint Position Statement of the Society for Adolescent Health and Medicine and the North American Society for Pediatric and Adolescent Gynecology.” Journal of Pediatric and Adolescent Gynecology 32(6): 563–6. 10.1016/j.jpag.2019.10.008. [DOI] [PubMed] [Google Scholar]
- 30. Traberg, C. S. , Harjani T., Roozenbeek J., and van der Linden S.. 2024. “The Persuasive Effects of Social Cues and Source Effects on Misinformation Susceptibility.” Scientific Reports 14(1): 4205. 10.1038/s41598-024-54030-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Borrero, S. , Frietsche S., and Dehlendorf C.. 2019. “Crisis Pregnancy Centers: Faith Centers Operating in Bad Faith.” Journal of General Internal Medicine 34(1): 144–5. 10.1007/s11606-018-4703-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Dehghan, A. , Siuta K., Skorupka A., Dubey A., Betlen A., Miller D., Xu W., et al. 2023. “Detecting Bots in Social‐Networks Using Node and Structural Embeddings.” Journal of Big Data 10(1): 119. 10.1186/s40537-023-00796-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data sets generated or analyzed during this study are available from the corresponding author upon reasonable request.