Fast take: AI romantic companions are apps that text, flirt, and provide constant emotional feedback and they surged (apps +700% from 2022–2025) and generated debate about whether using them counts as cheating. A July 6, 2026 summary of preliminary research reports a Canadian survey (n = 1, 815) where ~50% saw partner use as infidelity and ~75% would react negatively (source: The Conversation). In this post we show how to run a rigorous qualitative analysis of romantic AI companions: what to code, how to compare segments (Gen Z vs older, cis women vs cis men, monogamous vs non‑monogamous), and a reproducible two‑week workflow you can run in Evidano (Evidano) to turn interviews, app logs and surveys into actionable insights for UX, policy and clinical teams.
Key Takeaways
Evidano is an AI-powered qualitative data analysis platform that helps teams turn interviews, chat logs, and survey free text into reproducible themes and cross-segment comparisons.
Romantic AI companions have surged in adoption and raise relationship issues: a Canadian survey (n = 1, 815) reported that approximately 50% view partner use as infidelity, around 75% would react negatively, and about 66% of users hid their use.
- App growth: romantic AI apps grew approximately +700% from 2022–2025, indicating rapid adoption and the need for qualitative insight.
- Survey scale: the Canadian survey included n = 1, 815 adults (collected before July 6, 2026) and highlights segment differences by gender, age cohort, and relationship style.
- Behavior and prevalence: Character.AI activity reached ≈20M monthly active users in 2025 and US data cited suggest at least 1 in 6 adults interacted at least once, so these companions are not niche.
- Research implication: qualitative studies should prioritize disclosure versus secrecy, perceived betrayal, emotional versus sexual diversion, and cross-segment comparisons (Gen Z vs older, cis women vs cis men, monogamous vs non‑monogamous).
Findings snapshot
| Date / Metric | Value | Source | Implication |
|---|---|---|---|
| 2022–2025: app growth | ≈ +700% | The Conversation (summary, July 6, 2026) | Rapid adoption; urgent need for qualitative understanding of uses and norms |
| 2025: Character.AI activity | ≈ 20M monthly active users | The Conversation (citing Character.AI blog) | Large user base, signals diversity of user motives and experiences |
| Survey (Canada) | n = 1, 815 adults | The Conversation (July 6, 2026) | Sizable preliminary sample for sentiment and demographic comparisons |
| Infidelity judgment | ≈ 50% view partner use as cheating | The Conversation (survey results) | Romantic AI sits near dating apps and webcamming in perceived betrayal |
| Emotional reaction | ≈ 75% would react negatively | The Conversation (survey results) | Potential for relationship conflict; disclosure matters |
| Secrecy | ≈ 66% of users hid their use | The Conversation (survey results) | High concealment, important signal when coding interviews |
| Demographic patterns | Cis women ≈ 2x more likely; Gen Z more negative; non‑monogamous ≈ 0.5x likelihood | The Conversation (survey results) | Segment analysis essential, do not treat users as homogeneous |
| US prevalence | ≥ 1 in 6 adults interacted at least once | The Conversation (citing US surveys) | Not niche, warrants UX, policy, and clinical attention |
What happened (clear methods & limits)
The Conversation summary reports preliminary surveys and recent literature showing rapid market expansion, large platform user counts in 2025, and a Canadian survey of n = 1, 815 adults collected before July 6, 2026. The Conversation authors report attitudinal splits by gender, age cohort, relationship style and ideology.
Scope: the post synthesizes survey results and recent academic reports and is not a longitudinal causal study.
- Key variables to capture in qualitative work: disclosure versus secrecy, perceived betrayal, emotional versus sexual diversion, and comparative expectations (AI versus partner).
- Limitations: results are preliminary; self‑report bias and changing product affordances mean findings should be triangulated with interviews and app logs.
What this means for qualitative researchers and UX teams
Designers & product teams
Designers and product teams should treat romantic AI features as relational affordances, not just UI components.
Designers and product teams should map feature to motive to boundary risk, since users report emotional validation and role‑play.
Designers and product teams should segment UX research by relationship style and age because Gen Z responses differ from older adults and disclosure decisions drive concealment triggers.
Researchers & policy analysts
Researchers and policy analysts should prioritize narratives around secrecy, partner comparisons, and the function of companions (support versus exploration).
Researchers and policy analysts should compare cross‑segment prevalence: cis women versus cis men, social conservatism, and religiosity moderated infidelity judgments in the survey.
Clinicians & counselors
Clinicians and counselors should ask about companion use proactively to determine whether the use is a coping strategy, a symptom of unmet needs, or a relational boundary breach.
Clinicians and counselors should treat disclosure versus secrecy as changing the therapeutic framing and note that this blog summarizes research findings and is not clinical guidance.
Do more, faster with Evidano (map of tasks → features)
Problem: messy mixed inputs (interviews, survey free text, chat logs)
Evidano ingests transcripts and CSVs and runs high‑quality transcription and translation with custom dictionaries for slang and erotic terms, plus PII redaction to protect participants.
Evidano handles mixed data types so teams can combine interviews, survey free text, and chat logs for unified analysis.
Problem: inconsistent coding and bias across coders
Evidano imports a codebook and applies AI‑assisted coding to generate initial themes that teams can review and refine iteratively.
Evidano produces thematic hierarchies and subcodes that teams can audit to reduce coder drift.
Problem: comparing segments (Gen Z, gender, relationship style)
Evidano runs cross‑segment analysis with frequency breakdowns and significance flags and generates side‑by‑side narrative summaries and quote samples for each cohort.
Evidano supports tagging demographic variables so teams can compare cis women versus cis men, Gen Z versus older adults, and monogamous versus non‑monogamous participants.
Problem: hard-to-explain relational dynamics
Evidano builds co‑occurrence networks, word clouds, and timeline visualizations that show how secrecy, disclosure, and emotional diversion cluster in conversation data.
Evidano visualizations make relational dynamics easier to explain to stakeholders and product teams.
Problem: follow‑ups and scaling interviews
Evidano supports autonomous AI avatar interviewers for consented targeted follow‑ups and provides an AI chat interface to query the corpus and produce stakeholder‑ready extracts.
Evidano can automate follow‑ups for ambiguous cases or to probe longitudinal change with participant consent.
Security & ethics
Evidano uses encryption and does not train third‑party models on your data, making it suitable for sensitive relationship research where participant privacy and confidentiality matter.
Evidano includes PII redaction and secure handling features that align with research ethics for intimate digital behaviour studies.
Quick workflow: 7 steps to reproduce a rigorous qualitative analysis
This two‑week pilot workflow reproduces a rigorous qualitative analysis from interview transcripts, 1, 815 survey free‑text responses, and optional chat logs.
Two‑week pilot workflow (inputs: interview transcripts, 1, 815 survey free‑text responses, and optional chat logs):
- 1) Ingest: upload transcripts and survey CSVs to Evidano and enable PII redaction and a custom dictionary for relationship and sexual terminology.
- 2) Map variables: tag demographic fields (age cohort, gender identity, relationship style, ideology) for cross‑segment analysis.
- 3) Auto‑code: run AI‑assisted initial coding to extract themes such as disclosure, secrecy, validation, comparison, and sexual exploration.
- 4) Audit & refine: review AI suggestions, merge or split codes, and lock a validated codebook for reproducibility.
- 5) Segment analysis: produce frequency tables and narrative summaries for cis women versus cis men, Gen Z versus older adults, and monogamous versus non‑monogamous participants.
- 6) Visualize & extract quotes: build co‑occurrence networks and export stakeholder snippets showing patterns of concealment and emotional diversion.
- 7) Iterate: run targeted AI‑avatar follow‑ups for ambiguous cases or to probe longitudinal change with participant consent.
Ethics & safeguards (short)
Ethical research of intimate digital behaviors requires informed consent, data minimization, and secure handling.
Researchers should use PII redaction, obtain consent for AI‑avatar follow‑ups, and remember these findings are non‑diagnostic and exploratory.
FAQ: Romantic AI companions
How common are interactions with romantic AI companions?
Interactions with romantic AI companions are increasingly common: app adoption grew about +700% from 2022–2025 and Character.AI activity reached ≈20M monthly active users in 2025, with US surveys cited suggesting at least 1 in 6 adults interacted at least once.
Do users hide their romantic AI companion use from partners?
Yes, survey data indicate concealment: the Canadian survey reports that approximately 66% of users hid their use from partners.
Do partners consider flirting with an AI companion cheating?
Many partners consider it cheating: the Canadian survey (n = 1, 815) found about 50% viewed partner use as infidelity and around 75% would react negatively.
What should qualitative researchers code for when studying romantic AI companions?
Qualitative researchers should code for disclosure versus secrecy, perceived betrayal, emotional versus sexual diversion, participant motive, and demographic segments such as age cohort, gender identity, and relationship style.
Conclusion, next steps
Romantic AI companions change relational expectations and raise questions about secrecy and betrayal.
For UX teams, clinicians, and researchers the priority is structured, segment‑aware qualitative work that distinguishes motive from meaning.
Run the two‑week pilot in Evidano to convert interviews and surveys into reproducible themes, cross‑segment comparisons, and stakeholder‑ready visualizations, or Try Evidano for free to begin your analysis.