AI Companion Usage Trends 2026: Adult Use, Privacy & Market Data

Published 14 July 2026 by Backdoor Boutique AI. This page contains sponsored links; see our disclosure policy for details.

---

Table of Contents

• Methodology and Limitations
• What Changed in This Update
• Definitions: What Counts as an AI Companion
• 2026 Headline Findings
• Adoption Evidence and Where the Gaps Are
• Adult-Use Motivations: Why People Turn to AI Companions
• Mobile Spending on AI Companion Apps
• Country Distribution of Companion Traffic
• Intimate-Data Privacy: What You Disclose and Who Holds It
• Platform Content Rights: Who Owns Your Conversations
• Cross-Border Data Transfers
• Deletion: Can You Actually Erase Your Data
• Billing Discretion and Payment Privacy
• Voice, Images and Generated Media
• Roleplay Boundaries and Platform Controls
• Heavy Use and Wellbeing: What the Research Actually Shows
• Regulation: FTC, UK Online Safety, and the EU AI Act
• Pre-Payment Privacy Audit: A Practical Checklist
• Provider Comparison: The Questions That Matter
• Future-Measurement Principles
• Source Register
• Update Log

---

Methodology and Limitations

This report draws exclusively on primary and near-primary research published between late 2023 and mid-2026. Every material number is accompanied by its source, measurement method, and scope limitations in the same paragraph or table row.

We applied a strict verification process before including any figure:

• Measured data only. We use survey data, app-store revenue measurements, and documented policy reviews. We do not cite standalone analyst market-size forecasts as facts — they diverge by more than 20x for the same year across different vendors and use incompatible definitions.
• Source-type labelling. Every figure is labelled as survey-measured, app-store-measured, estimated, or policy-review-derived. We distinguish general chatbot evidence from dedicated companion-app evidence every time.
• Excluded claims. We have excluded the Joi AI "80% of Gen Z would marry an AI" survey (commercially commissioned, undisclosed recruitment methodology), the Gitnux "500 million users by 2027" projection (unverifiable), and unsupported price, rating, or user-count claims that circulate in competitor roundups.
• Causal language. No wellbeing association in this report is presented as causal. Observational findings from cross-sectional or convenience-sample studies are identified as such.
• Youth data quarantine. Teen and child data appears only in clearly labelled regulatory and safety contexts. This page is written for adults and does not market to minors.

Where a figure's scope is limited — for example, Ofcom's data covering general generative AI rather than companion apps specifically, or Sensor Tower's revenue covering only app-store transactions — we state the limitation explicitly.

How we evaluate adult AI companions

---

What Changed in This Update

This is the first publication of this report, dated 14 July 2026. It establishes the baseline evidence set we will track and update.

Key editorial decisions for this edition:

• Pew Research Center data (June 2026) provides the freshest US adoption numbers available. We use the "all chatbot types" measure and clearly separate it from companion-specific evidence.
• Ofcom UK data is dated November 2023 and explicitly covers general generative AI use, not companion apps. We include it as the most recent UK-representative survey available, with the expectation that updated figures will be substantially higher.
• Sensor Tower Q1 2026 mobile revenue data is used for the companion-app subcategory but noted as app-store-only — web-based subscription revenue is excluded.
• The Ada Lovelace Institute's privacy-policy review is treated as a policy-document analysis, not a technical data-flow audit.
• The Phang et al./OpenAI–MIT affective-use study and the Zhang et al. Stanford/Carnegie Mellon Character.AI study are included for wellbeing context, with full methodology notes.

The next planned review is October 2026, when we expect updated Ofcom and potentially updated Pew data.

---

Definitions: What Counts as an AI Companion

The term "AI companion" gets applied to wildly different products. Before the numbers make sense, the categories need separating.

General AI Chatbot

A large-language-model interface designed primarily for information, productivity, or creative tasks. ChatGPT, Gemini, and Copilot fall here. Most adoption statistics — including the Pew and Ofcom figures in this report — measure this category. They do not isolate romantic or adult use.

AI Companion App

A product explicitly designed around ongoing personal or emotional interaction. Character.AI, Replika, and the platforms reviewed on this site belong here. The user typically creates or selects a persistent character and the interface encourages repeated, relationship-like engagement.

Adult AI Companion

A subset of companion apps that permits or actively supports explicit, sexual, or uncensored conversation and media generation. This is the primary focus of this report and the products we cover. Compare uncensored AI girlfriend apps

AI Girlfriend / AI Boyfriend

Marketing labels for companion apps positioned around a simulated romantic partner. These may or may not include adult content. The label itself tells you little about the actual privacy or content policies.

Image Generator vs Companion

Some products are prompt-based image generators that happen to produce romantic or explicit visuals. They are not companions in the conversational sense. We treat these separately. AI girlfriend image generators explained

The critical point: most public statistics describe general chatbots. Companion-specific measured data is thin. Adult-companion-specific measured data is thinner still. Every number in this report specifies which category it actually covers.

---

2026 Headline Findings

Finding	Source	Type	Limitation
Roughly half of US adults have used an AI chatbot, up from about one-third in 2024	Pew Research Center, June 2026	Survey (self-reported, all chatbot types)	US-only; different survey waves; covers all chatbots, not companions
Approximately one in four US chatbot users reported daily or near-daily use	Pew Research Center, Feb 2026 survey	Survey (self-reported)	All chatbot types including search and work use
Only about 4% of US chatbot users said they use chatbots for companionship	Pew Research Center, June 2026	Survey (multi-select)	Small subgroup; wider margin of error; includes non-romantic companionship
31% of UK adults online had used generative AI; 79% among online teens 13–17	Ofcom, November 2023	Survey	Data from late 2023 — likely understates current UK adoption; covers all generative AI, not companion apps
Mobile in-app spending on AI companion apps reached an estimated $150 million in Q1 2026	Sensor Tower, State of AI Apps 2026	App-store measurement	App Store and Google Play only; web subscriptions excluded
14 of 16 reviewed platforms claimed broad rights to user-generated content	Ada Lovelace Institute, 2026	Privacy-policy review (not a technical audit)	16 platforms sampled; interpretation is legal analysis, not code-level verification
70% of sampled ChatGPT conversations were personal (non-work) use	OpenAI/NBER, September 2025	Platform analytics (1.5M conversations)	ChatGPT only, not dedicated companion apps

These are the strongest verified figures available. Notice the recurring pattern: most measured data describes general chatbot use. The $150 million Sensor Tower figure and the Ada Lovelace policy review are the only two that specifically target companion products.

---

Adoption Evidence and Where the Gaps Are

What We Know: General Chatbot Adoption

According to Pew Research Center (June 2026), roughly half of US adults reported having used an AI chatbot — up from about one-third in 2024. These figures cover all chatbot types, not companion or romantic chatbots specifically. The survey is US-only and based on self-reported "ever used" responses. The 2024 and 2026 figures come from different survey waves with potentially different question wording. (Source)

Pew's February 2026 survey found that approximately one in four US adults who have used chatbots reported using them on a daily or near-daily basis. This covers all chatbot types including search, work, and personal use. (Source)

As of November 2023, Ofcom found 31% of UK adults online had used generative AI tools, rising to 79% among online teens aged 13–17. These figures cover all generative AI services — including ChatGPT, Snapchat My AI, Midjourney, and DALL-E — not companion apps specifically. Given rapid adoption since late 2023, these figures likely understate current UK usage. (Source)

An OpenAI/NBER analysis of 1.5 million ChatGPT conversations found 70% were personal (non-work) use. Messages were classified as Asking (49%), Doing (40%), or Expressing (11%). This covers general ChatGPT use, not dedicated AI companion platforms. The 11% "Expressing" category includes emotional or creative expression but is not equivalent to companion use. (Source)

What We Know: Companion-Specific Adoption

Companion-specific adoption data is remarkably sparse. AI companion apps had surpassed 220 million cumulative downloads by 2025, according to app-analytics firm Appfigures (as cited by multiple secondary sources including Psychology Today). Cumulative downloads include users who may have since uninstalled or stopped using the apps, and download counts should not be read as active-user counts. (Appfigures, via secondary reporting)

What We Do Not Know

• How many adults use companion apps for explicit or adult content. No published survey isolates this with a transparent, independently recruited sample.
• Active daily users of dedicated companion apps. Cumulative downloads and organic traffic estimates are not equivalent to daily use.
• UK-specific companion adoption. The Ofcom data covers all generative AI. There is no published UK-specific companion survey.
• Paying subscriber counts for any individual platform. These are proprietary and unverified.

This is the honest state of evidence. Anyone claiming to know exact companion user numbers with precision is either working from proprietary internal data they have not published, or extrapolating beyond what the underlying sources support.

---

Adult-Use Motivations: Why People Turn to AI Companions

Among US adults who use chatbots, Pew found search (42%) and work tasks (38%) dominate. Only about 4% reported using chatbots specifically for companionship, and about 10% for emotional support or advice (Pew Research Center, June 2026). Note: users could select multiple use cases. These percentages describe all chatbot users, not people who specifically sought out companion products.

For people who do seek companion apps — and particularly adult-oriented ones — the motivations differ from general chatbot use. Based on the research literature and the APA's 2026 evidence synthesis, identifiable motivations include:

Privacy and low-stakes disclosure. The ability to express desires, fantasies, or emotional states without judgement or social consequences. For adults exploring sexuality, kink, or intimacy questions, an AI that has no social network to gossip with can feel genuinely safer than a human confidant. Private AI companion options reviewed

Sexual exploration without interpersonal risk. Adults who are curious about fantasies they are not ready to introduce with a partner, or who are between relationships, use companion apps as a low-consequence space. This is a practical motivation, not a clinical finding.

Companionship during isolation. The APA's 2026 Trends Report notes that synthetic relationships are "filling the void to satisfy the fundamental human need for social connection" (APA, January 2026). This observation applies broadly, not only to adult-content users, and the APA simultaneously warns about dependency risks.

On-demand availability. Unlike human partners, AI companions are available at 3 AM, do not need reciprocal emotional support, and do not require scheduling. For shift workers, people with social anxiety, or anyone in an inconvenient time zone, this is a material feature.

Control over pacing and boundaries. The user sets the pace. A conversation can be paused, redirected, or ended without social negotiation. For people who have experienced coercion or boundary violations in human relationships, this degree of control may be genuinely therapeutic — though no clinical trial evidence supports that specific claim.

What "without a filter" really means

---

Mobile Spending on AI Companion Apps

Mobile in-app spending on AI companion apps reached an estimated $150 million in Q1 2026 (Sensor Tower, State of AI Apps 2026). This covers App Store and Google Play in-app purchases only — revenue from web-based subscriptions is not included. Given that many adult companion platforms deliberately bypass app stores to avoid content restrictions (and the 15–30% commission), the $150 million figure almost certainly understates total companion spending.

For context, overall generative AI mobile-app revenue reached $1.9 billion in Q1 2026, having grown more than 32x from under $60 million in Q1 2023 (Sensor Tower). Companion apps represent roughly 8% of that measured mobile total, a modest share that reflects the category's early stage and the prevalence of web-based billing.

Metric	Value	Source	Scope Limitation
AI companion mobile in-app revenue, Q1 2026	$150M (estimated)	Sensor Tower	App Store + Google Play only; web revenue excluded
Total GenAI mobile-app revenue, Q1 2026	$1.9B	Sensor Tower	App Store + Google Play only
GenAI mobile-app revenue, Q1 2023	Under $60M	Sensor Tower	Baseline for growth comparison
Growth multiple (GenAI apps, Q1 2023 → Q1 2026)	32x+	Sensor Tower	Covers entire GenAI category, not just companions

What the App-Store Gap Means for Adult Users

If you subscribe to an AI companion through its website rather than through an app store, your spending is invisible to Sensor Tower and similar trackers. Many explicitly adult platforms encourage web checkout because:

1. App stores restrict explicit content, so the web version often has fewer content limits.
2. Web subscriptions avoid the app-store commission, sometimes resulting in lower prices.
3. Payment processing through the web may offer different billing-descriptor options.

This makes the $150 million a floor, not a ceiling. The real spending on AI companion products is higher — possibly substantially higher — but no public dataset currently captures the web component. Payment privacy and billing explained

---

Country Distribution of Companion Traffic

An analysis of estimated organic search traffic to ten leading AI companion websites found the US accounted for 33.5% of visits, followed by Brazil (8.7%), Indonesia (6.9%), and the UK (3.8%) (Surfshark/Ahrefs, February 2026). These figures reflect search-engine traffic estimates only and exclude app downloads, direct visits, and in-app usage.

Country	Estimated Organic Traffic Share	Source	Limitation
United States	33.5%	Surfshark/Ahrefs, Feb 2026	Ahrefs estimate; excludes app, direct, social traffic
Brazil	8.7%	Surfshark/Ahrefs, Feb 2026	Same limitation
Indonesia	6.9%	Surfshark/Ahrefs, Feb 2026	Same limitation
United Kingdom	3.8%	Surfshark/Ahrefs, Feb 2026	Same limitation

What These Numbers Do and Do Not Tell You

These are estimated organic search shares — the portion of search-engine visits that Ahrefs' crawler attributes to each country across the ten sampled companion sites. They are not user counts, not app-install figures, and not a measure of how many people in each country actually use companion apps.

Ahrefs estimates can diverge significantly from actual analytics. Countries with high app-store usage and low search behaviour — or where companion use happens primarily through social-media referrals — will be underrepresented. The UK's 3.8% share, for example, likely understates UK usage given that many UK users access companion apps directly rather than through search engines.

For UK readers of this report: the combination of the Ofcom generative-AI figures (31% of adults online in late 2023) and the 3.8% organic traffic share suggests growing UK interest in AI tools generally, with companion-specific adoption still difficult to measure precisely.

---

Intimate-Data Privacy: What You Disclose and Who Holds It

When you use an AI companion — particularly an adult one — you are potentially disclosing:

• Sexual preferences and fantasies
• Emotional vulnerabilities and mental-health states
• Relationship status and personal history
• Voice recordings (if voice chat is enabled)
• Generated images of faces, bodies, or scenarios
• Real names, locations, and identifying details shared in conversation

This is not abstract data. It is the most intimate information many people ever articulate to anyone. And unlike a therapist or a partner, an AI companion platform is a commercial entity whose data practices are governed by terms of service you likely did not read.

The Surfshark Privacy-Label Findings

Surfshark's February 2026 analysis of companion-app privacy labels found significant variation in data collection across platforms. The study examined App Store privacy labels — the self-declared data types each app collects — and found that companion apps as a category tend to collect extensive data including contact information, usage data, identifiers, and in some cases health and fitness information (Surfshark, February 2026). This analysis is based on what apps declare on their store listings, which may differ from actual collection practices.

The Ada Lovelace Institute Findings

A review of 16 AI companion platforms by the Ada Lovelace Institute found that 14 claimed broad rights to use content generated during conversations, and 11 had unclear policies on cross-border data transfers. This is based on analysis of published privacy policies and terms of service, not a technical data-flow audit (Ada Lovelace Institute, 2026).

The practical meaning: when you share a fantasy, generate an image, or have an extended intimate conversation, the platform likely has a legal claim to use that content for training, product improvement, or other purposes described in their terms. Whether they actually exercise that right is a separate question — but the contractual permission exists in most cases.

How to delete your AI chat data

---

Platform Content Rights: Who Owns Your Conversations

The Ada Lovelace Institute's finding that 14 of 16 reviewed companion platforms claim broad content-use rights deserves detailed examination (Ada Lovelace Institute, 2026). "Broad rights" typically means some combination of:

• A licence to use your conversation data for model training
• Permission to create derivative works from your inputs
• The right to aggregate and anonymise your data for research or commercial purposes
• Retention of generated content (images, voice clips) beyond the life of your account

What This Means for Adult Content

If you are generating explicit text, images, or voice content through a companion app, the platform's content licence may allow them to use that material in ways you did not anticipate. Even if the terms say "anonymised" or "aggregated," the intimate nature of companion conversations can make true anonymisation difficult — particularly when conversations contain personally identifying details.

What to Check Before You Start

1. Read the content-licence clause. Search the terms of service for words like "licence," "rights," "content," and "grant." Look for whether the licence is "perpetual," "irrevocable," "worldwide," and "sublicensable."
2. Check whether deletion revokes the licence. Some platforms retain the right to use content even after you delete your account.
3. Ask about training data. Does the platform use your conversations to train models? If so, is there an opt-out mechanism?
4. Check generated-content ownership. Who owns an image the AI generates based on your prompt? The answer varies by platform.

Set up private roleplay properly

---

Cross-Border Data Transfers

The Ada Lovelace Institute found that 11 of 16 reviewed companion platforms had unclear policies on cross-border data transfers (Ada Lovelace Institute, 2026). This is a policy-document review, not a technical audit, and "unclear" reflects the researchers' legal interpretation of the published policies.

Why This Matters for UK and EU Users

If you are in the UK or EU, your data is subject to the UK GDPR or EU GDPR respectively. These regulations require that personal data transferred outside the UK or EU receive an adequate level of protection. If a companion platform is based in the US, hosts data in Singapore, or uses cloud infrastructure that spans multiple jurisdictions, your intimate conversations may be processed in countries with weaker privacy protections.

Practical Questions to Ask

• Where is the company incorporated?
• Where are its servers physically located?
• Does its privacy policy mention Standard Contractual Clauses, adequacy decisions, or other lawful transfer mechanisms?
• If you request data deletion under GDPR, does the deletion propagate to all jurisdictions where copies exist?

Most users will not audit these details. But if you are sharing content that could be embarrassing, professionally damaging, or legally sensitive, understanding where it ends up is not paranoia — it is due diligence.

---

Deletion: Can You Actually Erase Your Data

Account deletion and data deletion are not the same thing. Deleting your account typically removes your login credentials, profile, and visible conversation history. It does not necessarily remove:

• Training data derived from your conversations
• Anonymised aggregates that include your inputs
• Backup copies retained for legal or operational reasons
• Content that has already been shared with third-party processors

The GDPR Right to Erasure

Under UK GDPR and EU GDPR, you have a right to request erasure of your personal data (Article 17). Platforms must comply unless they have an overriding legitimate reason to retain it. In practice:

1. Submit a formal data-deletion request, not just an account-deletion request.
2. Request confirmation that the deletion has been completed.
3. Ask specifically about training-data copies and third-party processors.
4. Set a calendar reminder to follow up if you do not receive confirmation within the statutory 30-day period.

Platform-Specific Deletion Controls

Deletion quality varies enormously. Some platforms offer one-click conversation deletion and account removal. Others require email support requests with identity verification. A few have no documented deletion process at all.

Before subscribing to any companion app, locate the deletion mechanism. If you cannot find clear instructions for how to erase your data, that is a signal about the platform's overall approach to privacy. Step-by-step data deletion guide

---

Billing Discretion and Payment Privacy

For many adults, the billing descriptor on a credit-card or bank statement is as much a privacy concern as the data itself. If your statement reads "NSFW AI CHAT $19.99" and a partner, employer, or family member sees it, the privacy breach is social rather than technical — but no less real.

What to Check

Question	Why It Matters	What Good Looks Like
What appears on your statement?	Descriptors vary from generic to explicit	A neutral company name with no adult or AI reference
Does the platform use a third-party billing provider?	The provider's name may appear instead of the platform's	Confirm the descriptor before paying, not after
Can you pay with a prepaid card or cryptocurrency?	Alternative payment methods reduce statement exposure	At least one non-bank option available
What happens if you dispute a charge?	Chargebacks can trigger account bans and additional disclosure	Clear cancellation and refund processes reduce the need for disputes
Are renewal emails discreet?	Email subject lines like "Your AI Girlfriend Subscription" are not	Neutral email subjects and sender names

Free-Trial and Auto-Renewal Traps

Several companion platforms offer free trials that convert to paid subscriptions automatically. If you forget to cancel, the charge appears on your statement. Check:

• Whether a payment method is required at trial signup
• The exact renewal date and amount
• How to cancel (some platforms make this deliberately difficult to find)
• Whether cancelled accounts retain access through the paid period or cut off immediately

Full billing privacy guide

---

Voice, Images and Generated Media

Voice interaction and image generation have become major features in companion apps. Both raise distinct privacy considerations.

Voice

Voice chat means the platform processes audio recordings of your voice — your actual biometric data. Questions to consider:

• Is voice data stored or processed in real-time only?
• Can voice recordings be used for model training?
• Is the platform's voice model running locally on your device or on remote servers?
• If on remote servers, where are those servers located?

Voice data combined with conversation content creates an unusually complete intimate profile. A text conversation reveals what you said; a voice recording reveals how you said it, including emotional state, hesitation, and arousal.

Generated Images

Many companion apps generate custom images based on text prompts or predefined characters. For adult use, this means the platform may be generating explicit visual content associated with your account. Questions to consider:

• Are generated images stored on the platform's servers after you view them?
• Can you download and then delete server-side copies?
• Are images generated on-device or server-side?
• Does the platform's content licence cover generated images?
• Could generated images be used in platform marketing or training data?

A Practical Media Privacy Approach

1. Assume all voice and image data is server-processed unless the platform explicitly documents on-device processing.
2. Periodically delete stored media if the platform offers that option.
3. Do not include personally identifying elements (your face, real name, real location) in image prompts.
4. If you are concerned about voice biometrics, use text chat instead.

Test what your AI companion remembers about you

---

Roleplay Boundaries and Platform Controls

"Uncensored" and "no filter" are marketing terms, not technical descriptions. Every platform has boundaries, whether enforced by content-moderation systems, model fine-tuning, or terms of service. The question is where those boundaries are, how consistently they are applied, and whether you can adjust them.

What "Uncensored" Typically Means

In practice, "uncensored" usually means the platform permits explicit sexual content between consenting adult characters. It does not mean the platform has no rules at all. Most platforms — even those marketed as uncensored — prohibit:

• Content depicting minors in any sexual context
• Non-consensual violence outside of clearly fictional roleplay frameworks
• Content that violates applicable laws in the platform's jurisdiction
• Use of real people's names and likenesses in explicit scenarios without consent

Adjustable Boundaries

Some platforms let you configure the level of explicit content, the types of scenarios permitted, and the conversational style. This is a useful feature because it means:

• You can start with moderate settings and adjust as you become comfortable.
• You can set hard limits on topics you do not want the AI to introduce.
• You can distinguish between a platform that permits adult content and one that forces it.

Compare adult roleplay chatbot controls

The Boundary-Collapse Problem

A known limitation of current AI companions is inconsistent boundary enforcement. A platform may claim to prohibit certain content but fail to enforce that prohibition consistently. Conversely, a platform marketed as "uncensored" may introduce unexpected refusals mid-conversation. This inconsistency is a product of how large language models are trained and fine-tuned — it is a technical limitation, not a deliberate design choice.

If consistent boundaries (in either direction) matter to you, test them during any free-trial period before committing to a paid subscription. NSFW AI chatbot boundary comparison

---

Heavy Use and Wellbeing: What the Research Actually Shows

Two significant studies published in 2025 examined the relationship between AI companion use and psychological wellbeing. Both are important. Neither proves what headlines might suggest.

The OpenAI–MIT Affective-Use Study (Phang et al., 2025)

This study combined large-scale automated analysis of over 3 million ChatGPT conversations for affective cues, a survey of over 4,000 users on their perceptions, and an approximately 1,000-person 28-day randomised controlled trial (RCT) examining the impact of ChatGPT voice mode on emotional wellbeing. (Source: arXiv:2504.03888)

Key points for the adult-companion context:

• The study covers ChatGPT, not dedicated companion apps. ChatGPT's use case, user base, and interaction patterns differ substantially from companion-specific platforms.
• The RCT component gives this study stronger causal evidence than most in the field, but the sample (approximately 1,000 participants over 28 days) is modest.
• Findings on affective use are directional — they indicate that a meaningful proportion of ChatGPT interactions involve emotional expression, but they do not describe the companion market specifically.

The Stanford/Carnegie Mellon Character.AI Study (Zhang et al., 2025)

This study analysed survey data from 1,131 Character.AI users and 4,363 donated chat sessions (413,509 messages) from 244 participants. It examined how companion use is associated with psychological wellbeing. (Source: arXiv:2506.12605)

Key findings and critical limitations:

• The study found associations between heavy companion use and certain wellbeing patterns, but these are observational associations, not causal findings. People who are already lonely or struggling may seek out companion apps; the app use does not necessarily cause the wellbeing outcome.
• Participants were self-selected Character.AI users who volunteered to share their data — this is a convenience sample, not a representative population study.
• The donated chat sessions came from 244 participants, a small subset of the survey respondents. Generalising from this group to all companion users requires caution.

The APA Synthesis

The American Psychological Association's 2026 Trends Report synthesises the available evidence and notes that "research shows excessive use of these tools may worsen loneliness and erode social skills" (APA, January 2026). The key words are "may" and "excessive" — the APA is flagging a risk pattern observed in correlational research, not declaring a proven causal relationship.

Practical Takeaways

1. Heavy use of AI companions is associated with, but not proven to cause, negative wellbeing outcomes. If you notice increasing time spent with an AI companion and decreasing time with human connections, that pattern is consistent with the risk the APA describes.
2. Moderate use alongside a functioning social life has not been shown to be harmful. The research does not suggest that any use of companion apps is problematic.
3. If you are experiencing loneliness, depression, or suicidal thoughts, an AI companion is not a substitute for professional support. Contact a qualified mental-health professional or a crisis service: - UK: Samaritans — 116 123 (free, 24/7) - US: 988 Suicide and Crisis Lifeline — call or text 988 - International: findahelpline.com

This is not a claim that AI companions cause harm. It is a recognition that the evidence is still emerging and that self-awareness about usage patterns is sensible.

---

Regulation: FTC, UK Online Safety, and the EU AI Act

US: The FTC Inquiry

In September 2025, the Federal Trade Commission launched a formal inquiry into AI chatbots acting as companions, issuing 6(b) orders to seven companies that operate consumer-facing AI chatbots. The inquiry seeks information about how these firms "measure, test, and monitor potentially negative impacts on consumers, including on mental health, and whether they employ deceptive tactics that would violate the FTC Act" (FTC, September 2025).

The seven named companies include major companion and general-chatbot operators. The FTC's focus areas include:

• Data practices and privacy protections
• Impact on children and young people (youth-safety context)
• Anthropomorphisation and deceptive design patterns
• Subscription and billing practices

As of July 2026, the FTC has not published enforcement actions from this inquiry, but the investigation itself signals regulatory attention that may affect product design and data practices.

UK: Online Safety Act and AI Companions

The UK Online Safety Act 2023 places duties on online services to protect users, with particular emphasis on children. The Ada Lovelace Institute's analysis highlights that "nearly 50% of AI companion platforms have either no age-verification systems or rely on self-declaration" (Ada Lovelace Institute, 2026), which may bring them within Ofcom's regulatory scope.

For adult users, the practical regulatory question is whether tighter rules on age verification and content moderation will affect the features available on platforms you use. Stricter age gates could mean more robust account verification, which in turn means more personal data held by the platform — a privacy trade-off.

EU: The AI Act

The EU AI Act, which came into force in stages through 2025–2026, classifies AI systems by risk level. Companion chatbots that simulate emotional relationships could fall under obligations requiring transparency — specifically, the requirement to inform users that they are interacting with an AI system. For adult-companion providers operating in the EU, this may require more prominent disclosures about the synthetic nature of the interaction.

A Note on Youth Safety

While this report is written for adults, the regulatory landscape is primarily driven by youth-safety concerns. Common Sense Media's nationally representative US teen survey found that nearly three in four teens have used AI companions, and half use them regularly, with significant numbers sharing personal information with these platforms (Common Sense Media, July 2025). This data is presented here as regulatory context — it explains why regulations that affect adult users are being enacted — not as marketing-relevant adoption data.

The broader 2025 context of EU enterprise AI adoption — Eurostat found 20% of EU enterprises with 10+ employees used AI technologies in 2025, up from 13.5% in 2024 (Eurostat, December 2025) — demonstrates that AI regulation is responding to rapid adoption across all sectors, not only companion apps.

---

Pre-Payment Privacy Audit: A Practical Checklist

Before paying for any AI companion, work through this checklist. It takes ten to fifteen minutes and can prevent regret.

Account and Identity

• [ ] Can you sign up without a real name?
• [ ] Is email-only registration available (no phone number required)?
• [ ] Can you use a disposable or alias email address?
• [ ] Does the platform require ID verification? If so, what for?

Data and Content

• [ ] Have you read the content-licence section of the terms of service?
• [ ] Does the platform use your conversation data for model training?
• [ ] Is there an opt-out for training-data use?
• [ ] Are generated images stored server-side? Can you delete them?
• [ ] Does the platform retain data after account deletion?

Billing

• [ ] What will the billing descriptor look like on your statement?
• [ ] Can you use a prepaid card?
• [ ] Is auto-renewal clearly disclosed?
• [ ] How do you cancel, and is cancellation accessible without contacting support?

Access and Device

• [ ] Does the app appear by name in your phone's app-download history?
• [ ] Can you use the service via a web browser in private/incognito mode?
• [ ] Does the app send push notifications that could be seen on a lock screen?
• [ ] Are conversations stored locally on your device or only on servers?

Jurisdiction

• [ ] Where is the company based?
• [ ] Where are its servers?
• [ ] Does the privacy policy mention GDPR compliance?
• [ ] Is there a data-protection officer or privacy contact?

Print this list or save it to your device. Work through it for any platform before entering payment details. Detailed deletion walkthroughs

---

Provider Comparison: The Questions That Matter

Rather than listing prices and features that change weekly, this section frames the questions you should apply to any companion platform you consider. Use these questions with the platforms we cover or any other service.

Question	What to Look For	Red Flag
What content does the platform actually allow?	A clear content policy with specific permitted and prohibited categories	Vague "we allow everything" claims with no documented policy
How is conversation data used beyond the current session?	An explicit privacy policy section on data retention, training use, and third-party sharing	No mention of data use, or an all-encompassing blanket licence
What happens to your data if the company is acquired or goes bankrupt?	A data-portability clause or a commitment to delete data on corporate dissolution	Silence on this topic — your intimate data becomes an asset in an acquisition
Can you export your conversation history?	A data-export or download feature	No export option and no response to GDPR Subject Access Requests
How does the billing appear?	A testable billing descriptor (some platforms disclose this in their FAQ)	A descriptor that names the platform or references adult content
What moderation is applied to user inputs?	A transparency report or content-policy document	Moderation that is either absent or completely opaque
Does the platform have a history of data breaches?	Search for "[platform name] data breach" before signing up	A breach history with no public remediation

These questions apply equally to every platform, including the ones we review on this site. Our evaluation method for all platforms

Compare AI sexting apps on the same criteria

Candy AI vs GirlfriendGPT detailed comparison

---

Future-Measurement Principles

The current evidence base is thin. As the companion market matures, better measurement will require:

Independent, Population-Representative Surveys

The most useful future data would come from independently funded (not platform-commissioned) surveys using probability sampling of adult populations. These surveys should distinguish between:

• General chatbot use (search, work, coding)
• Companion-oriented use (emotional support, companionship)
• Romantic or sexual use (explicit content, intimacy simulation)

Currently, only the Pew Research Center provides the first category at scale. The second and third categories have no independently measured equivalent.

Standardised Revenue Reporting

The gap between app-store revenue (measurable via Sensor Tower) and web revenue (unmeasured) makes total market sizing unreliable. Future data should include web-subscription revenue, ideally reported using consistent definitions across providers.

Longitudinal Wellbeing Research

The existing studies are either cross-sectional (Zhang et al.) or short-term (Phang et al., 28-day RCT). Meaningful wellbeing evidence requires:

• Longer study periods (six months to two years)
• Random assignment to companion use vs control (not self-selection)
• Separation of moderate from heavy use
• Pre-registration of study design and outcomes

Until this research exists, any strong claim about AI companions "causing" loneliness, depression, or harm is unsupported — as is any strong claim that they "cure" or "prevent" these conditions.

Privacy-Audit Standards

The Ada Lovelace Institute's policy review is a strong start, but future audits should include technical verification — actually testing what data is transmitted, where it is stored, and whether deletion requests are honoured at the technical level. Policy review alone cannot determine actual data practices.

---

Source Register

All primary and near-primary sources used in this report, with publication dates, types, and access notes.

Source	Publication Date	Type	URL	Access Note
Pew Research Center, "Americans and AI 2026"	June 2026	National probability survey (US adults)	Link	Free access
Ofcom, "Gen Z driving early adoption of Gen AI"	November 2023	National survey (UK internet users)	Link	Free access
Sensor Tower, "State of AI Apps 2026: APAC Edition"	2026	App-store analytics (in-app purchases)	Link	Summary free; full report paywalled
Surfshark, "AI companions and privacy"	February 2026	Third-party analysis (Ahrefs organic-traffic estimates + App Store privacy labels)	Link	Free access
Ada Lovelace Institute, "The companionship market"	2026	Privacy-policy review (16 platforms)	Link	Free access
OpenAI/NBER, "How people are using ChatGPT"	September 2025	Platform analytics (1.5M conversations)	Link	Free access (summary); NBER paper may be gated
Phang et al., "Investigating Affective Use and Emotional Well-being on ChatGPT"	April 2025	Large-scale study (3M+ conversations, 4,000+ survey, ~1,000-person RCT)	Link	Free access (arXiv)
Zhang et al., "The Rise of AI Companions"	June 2025	Observational study (1,131 survey respondents + 244 donated chat sessions)	Link	Free access (arXiv)
US Federal Trade Commission, companion-chatbot inquiry	September 2025	Regulatory action (6(b) orders)	Link	Free access
American Psychological Association, 2026 Trends Report	January 2026	Evidence synthesis and clinical commentary	Link	Free access
Common Sense Media, "Talk, Trust, and Trade-Offs"	July 2025	Nationally representative US teen survey	Link	Free access (summary)
Eurostat, enterprise AI adoption	December 2025	Official EU statistics	Link	Free access
Appfigures, cumulative companion-app downloads	2025 (via secondary citation)	App-analytics estimate	Via Psychology Today	Primary report not independently verified in our corpus

---

Update Log

Date	Change	Author
14 July 2026	Initial publication. Baseline evidence set established from 12 primary/near-primary sources. All figures verified against source documents and independent fact-check memo.	Backdoor Boutique AI editorial
—	Next planned review: October 2026 (expected updated Ofcom data, potential new Pew wave)	—

---

This report is maintained by Backdoor Boutique AI as an evidence-based reference for adults considering private AI companion use. Sponsored links are present in this article and are labelled. They support independent platforms we have reviewed; their presence does not influence the evidence we present. Read our full sponsored-link disclosure.

This page is intended for adults aged 18 and over. If you are under 18, please close this page.

About Backdoor Boutique AI

OurDream vs FantasyGF comparison

Joi AI vs Secrets AI comparison