Independent 2026 research · adults only

AI Girlfriend Statistics & Usage Trends 2026

What the strongest public evidence actually says about AI girlfriends, companion use, mobile spending, privacy and emotional wellbeing — with clear limits on what the numbers can prove.

Published 14 July 2026Updated 14 July 20266,152 wordsNext review: 14 October 2026

This report separates measured usage from forecasts, dedicated companions from general chatbots, and correlation from causation. Every material number links to its source and includes the limitation needed to interpret it.

AI companion adoption in numbers

US adults who have used an AI chatbotBroad chatbot adoption, not dedicated companion use
202433%
202649%

Source: Pew Research Center, February 2026 survey.

Why US adults use AI chatbotsSelected uses reported by all US adults
Search42%
Work tasks38%
Emotional support10%
Companionship4%

Source: Pew Research Center. Categories are broad chatbot uses and should not be treated as market share.

AI companion organic-search traffic by countryShare among ten platforms studied in January 2026
United States33.5%
Brazil8.7%
Indonesia6.9%
United Kingdom3.8%

Source and methodology: Surfshark, using Ahrefs organic-search estimates. Direct and in-app use were excluded.

$150 millionAI companion mobile in-app revenue in Q1 2026
12×+growth versus Q1 2023

Source: Sensor Tower, State of AI Apps 2026. This is mobile in-app revenue, not total market revenue.

Published 14 July 2026 by Happy Lovers AI. Last reviewed 14 July 2026. Next scheduled review: October 2026.

Sponsored-link disclosure: This report contains clearly marked sponsored links to AI companion platforms. Happy Lovers AI may receive a commission if you visit a partner through those links. Sponsored recommendations are kept separate from the evidence and do not influence the data presented. See our full sponsored-link disclosure policy.

What changed in this update

This is the first edition of this report. It was written on 14 July 2026 from primary sources published between November 2023 and June 2026. Every material number links to its original publication and carries a methodology note so you can judge it yourself. We have excluded several widely circulated figures that failed independent verification — including a viral "80% of Gen Z would marry an AI" survey funded by a companion-app vendor, a "500 million users by 2027" projection from a statistics aggregator, and standalone analyst market-size forecasts whose valuations diverge by more than twenty times for the same calendar year. The full exclusion rationale appears in the methodology section below.

Table of contents

Definitions: what counts as an AI girlfriend?

Before looking at any numbers, it helps to know what the terms mean. Different studies measure different things, and confusing them is the single biggest reason companion-app statistics get misleading.

General AI chatbot. Any conversational AI accessible to consumers — ChatGPT, Gemini, Copilot, Claude, and others. Most public adoption data (Pew, Ofcom, OpenAI/NBER) describes this broad category.

AI companion. A chatbot specifically designed for ongoing personal or emotional interaction. Platforms such as Replika, Character.AI, and the apps reviewed on this site sit here. Much smaller than the general chatbot market, and far less measured.

AI girlfriend (or boyfriend). A subset of AI companions positioned around romantic or intimate roleplay, customisable avatars, voice chat, and often explicit content. This is the product category most readers of this page are exploring.

Image generator. Some companion platforms include AI-generated images of your character. Others are standalone image tools with no conversation layer.

Roleplay platform. Services built around user-created fictional characters. Many overlap with companion use, but the primary framing is creative rather than relational.

Throughout this report, we state which category each data point actually measures. If a study measured all chatbots, we say so — even when the headline sounds companion-specific. For a practical introduction to how these apps work under the surface, see our guide to how AI girlfriend apps work.

2026 headline findings at a glance

Finding Source Category measured Key limitation
Roughly half of US adults have used an AI chatbot, up from about one-third in 2024 Pew Research Center, June 2026 All chatbots US-only, self-reported, covers ChatGPT/Gemini/etc. — not companion apps specifically
About one in four US chatbot users report daily or near-daily use Pew Research Center, Feb 2026 survey All chatbots Self-reported frequency; no companion breakout
Only about 4% of US chatbot users report using chatbots for companionship Pew Research Center, June 2026 All chatbots Small sub-sample; wider margin of error; users could pick multiple categories
Mobile in-app spending on AI companion apps reached an estimated $150 million in Q1 2026 Sensor Tower, State of AI Apps 2026 Companion mobile apps App Store and Google Play purchases only — web subscriptions excluded
AI companion apps surpassed 220 million cumulative downloads by 2025 Appfigures, 2025 (via Psychology Today / RubyChat) Companion mobile apps Cumulative installs include abandoned accounts; not a measure of active users
14 of 16 reviewed companion platforms claim broad content-use rights Ada Lovelace Institute, 2026 Dedicated companion platforms Policy-document review, not a technical data-flow audit

General chatbot adoption: the backdrop

Understanding AI girlfriend trends starts with the broader chatbot picture, because almost every large-sample adoption study measures chatbots in general — not companions in isolation.

US adoption doubled in two years

According to Pew Research Center (June 2026), roughly half of US adults reported having used an AI chatbot by the time of their February 2026 survey — up from about one-third in 2024. These figures cover all chatbot types including ChatGPT, Gemini, and Copilot. They are US-only and self-reported; a person who tried a chatbot once and never returned counts the same as a daily power user. The survey does not isolate companion or romantic chatbot usage.

Pew's February 2026 survey also found that approximately one in four US adults who have used chatbots reported using them on a daily or near-daily basis. Again, this covers search, work, personal, and all other use — not companion apps specifically.

UK adoption is measured, but three years ago

As of November 2023, Ofcom found that 31% of UK adults online had used generative AI tools, rising to 79% among online teenagers aged 13–17. These figures cover all generative AI services — ChatGPT, Snapchat My AI, Midjourney, DALL-E, and others — not companion apps specifically. The data is from November 2023, which is nearly three years old, and actual UK adoption is almost certainly higher today. No equivalent UK-specific companion-app survey has been published in our source set.

European enterprise context

For wider context, Eurostat (December 2025) reported that 20% of EU enterprises with 10+ employees used AI technologies in 2025, up from 13.5% in 2024. This is business adoption of all AI technologies (not consumer, not companion) but it signals the broader acceleration of AI across Europe.

Companion-specific adoption: what we actually know

Here is where the picture gets thin. Reliable, large-sample, companion-specific adoption data is scarce. Most of what circulates online is either vendor-commissioned, based on analyst projections, or drawn from general chatbot surveys that get re-labelled as "AI girlfriend statistics."

The Pew companionship figure

Among US adults who use chatbots, Pew (June 2026) found that about 10% reported using them for emotional support or advice, and only about 4% specifically for companionship. Users could select multiple use categories. Because 4% is a small slice of a survey sample, it carries a wider margin of error than the headline adoption numbers. Still, if roughly half of US adults have used a chatbot and 4% of those used one for companionship, that implies a low single-digit percentage of the total adult population — a meaningful number of people, but far from the tens or hundreds of millions sometimes claimed.

Cumulative downloads

Appfigures (2025, cited by Psychology Today and RubyChat) estimated that AI companion apps had surpassed 220 million cumulative downloads by 2025. Cumulative downloads include users who may have since uninstalled or stopped using the apps. A download is not the same as an active user, much less a paying subscriber.

App growth direction

The American Psychological Association's 2026 Trends Report noted that between 2022 and mid-2025, the number of AI companion apps surged by roughly 700%, citing TechCrunch. The base count and methodology behind "700%" are not specified (number of apps listed? total downloads? distinct user counts?), so this is useful as directional context — the space grew fast — but it is not precise enough for a chart.

If you are just getting started, our guide to the best AI girlfriend apps walks through what to look for in your first companion.

The companion evidence gap

One of the most important things to understand about AI girlfriend statistics in 2026 is how little companion-specific measured data actually exists. This table summarises the gap:

Question you might want answered Best available evidence What it actually measures What is missing
How many people use AI companions? Pew 4% companionship among chatbot users; 220M cumulative downloads (Appfigures) General chatbot users reporting companionship as a use case; lifetime app installs Monthly or daily active companion-specific users worldwide
How much money do companion apps make? Sensor Tower $150M Q1 2026 mobile in-app App Store/Google Play purchases in the companion subcategory Web-based subscription revenue (major gap — many NSFW platforms sell through their own sites)
Who uses AI companions? Zhang et al. survey of 1,131 Character.AI users; Pew demographics for all chatbot users A self-selected sample from one platform; general chatbot demographics Representative, multi-platform companion-user demographics
Does it help or hurt wellbeing? Phang et al. RCT (~1,000 participants, 28 days); Zhang et al. observational survey ChatGPT voice-mode affective use; Character.AI user self-reports Long-term (6+ month) clinical studies; studies of dedicated girlfriend/boyfriend apps
How does it affect real relationships? Zhang et al. observational associations Cross-sectional survey correlations Longitudinal evidence; controlled studies of relationship outcomes

If a figure claiming to answer one of these questions does not appear in the table above, it either failed our fact-check or comes from a source outside our verified corpus. The methodology page explains how we evaluate claims on this site.

Use cases and motivations

What people actually do with chatbots

The OpenAI/NBER study (September 2025) analysed 1.5 million ChatGPT conversations in a privacy-preserving manner and found that 70% of sampled consumer usage was personal (non-work). 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.

Pew's use-case breakdown

Among US adults who use chatbots, Pew (June 2026) 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. Users could select multiple use cases. These are US-only percentages of chatbot users, not of the general population.

Use case % of US chatbot users (Pew, June 2026) Notes
Search / finding information 42% Most common stated use
Work tasks 38% Among employed adults
Emotional support or advice 10% Broader than companionship; includes one-off venting
Companionship 4% Smallest measured category; wider margin of error

Source: Pew Research Center, "Americans and AI 2026". US adults surveyed Feb 2026. Multiple responses allowed. Covers all chatbots, not companion-specific platforms.

Why people turn to companions specifically

The APA's 2026 Trends Report synthesised clinical and survey evidence, noting that "synthetic relationships are filling the void to satisfy the fundamental human need for social connection." Commonly cited motivations in the research literature include:

  • Practising social interaction in a low-stakes setting
  • Coping with loneliness, particularly after a breakup, relocation, or loss
  • Exploring romantic or sexual identity without judgement
  • Curiosity about the technology itself
  • Emotional regulation — having a conversation partner available at any hour

The APA noted that experts are highlighting the need for guardrails to ensure user wellbeing, and that excessive use may be associated with worsened loneliness and eroded social skills. These are clinical observations, not causal findings from controlled trials. We unpack the wellbeing evidence further below.

Frequency and intensity of use

General chatbot frequency

Approximately one in four US chatbot users said they use chatbots daily or nearly daily in Pew's February 2026 survey. This covers all chatbot types. We do not have a reliable, independent daily-use figure for companion apps alone across the industry.

Companion-specific session data

The Zhang et al. study (Stanford/Carnegie Mellon, June 2025) analysed 4,363 donated chat sessions (413,509 messages) from 244 Character.AI users. The study found that users who described their relationship with their AI as a "romantic partner" sent more messages per session and used more emotional language than those who described it as a friend or tool. This is an observational study of self-selected participants from a single platform, so it should not be generalised to all companion users.

Voice-mode engagement

The Phang et al. study (OpenAI/MIT, April 2025) examined ChatGPT Advanced Voice Mode use and found that voice-based conversations with affective cues were growing as a share of total platform use. Among surveyed users who engaged in affective conversations, many reported using the voice mode as a way to "think out loud" or process emotions. This describes ChatGPT's general voice mode, not a dedicated companion product, but it suggests that voice is becoming a meaningful channel for emotional interaction with AI.

If voice chat matters to you, our voice-chat comparison guide breaks down what the leading apps offer.

Mobile market and spending

The Sensor Tower picture

Sensor Tower's State of AI Apps 2026 report provides the most concrete revenue data we have for companion apps on mobile. Overall generative AI mobile-app revenue grew from less than $60 million in Q1 2023 to $1.9 billion in Q1 2026, expanding more than 32 times in three years. Within that, mobile in-app spending on AI companion apps reached an estimated $150 million in Q1 2026.

A few essential caveats:

  • App-store scope only. Sensor Tower measures App Store and Google Play in-app purchases. Revenue from web-based subscriptions is not included. Many companion platforms — especially those offering explicit adult content — process payments through their own websites, so the $150 million figure understates total companion revenue, potentially by a large margin.
  • Growth multiplier. The content brief states this is "more than 12x Q1 2023." The companion-specific Q1 2023 baseline would be approximately $12.5 million. This baseline is not independently shown in the publicly available Sensor Tower summary.
  • Not profit. Revenue figures are gross consumer spending, not net revenue to developers after platform fees (typically 15–30%).
Period GenAI total mobile IAP revenue Companion subcategory (est.) Source
Q1 2023 < $60M ~$12.5M (implied) Sensor Tower
Q1 2026 $1.9B ~$150M Sensor Tower
Growth 32x+ (total GenAI) ~12x (companion) Sensor Tower

Limitation: App Store and Google Play in-app purchases only. Web-based subscription revenue excluded. Revenue figures are gross consumer spending.

Why market-size forecasts are unreliable

You may have seen headlines claiming the AI companion market is worth anywhere from $24 billion to $501 billion in 2026 alone. These are analyst forecasts from firms such as ResearchAndMarkets, Fortune Business Insights, Precedence Research, and others. They disagree with each other by more than twenty times for the same year, primarily because each defines "AI companion market" differently (some include all conversational AI, some only social/romantic). We have excluded standalone analyst market-size forecasts from this report's charts because they cannot be presented as measured data.

If you are interested in mobile-first companion experiences, our guides cover the best options for iPhone and the best options for Android.

Country and language distribution

Organic search traffic share

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

Country Estimated organic search traffic share Source
United States 33.5% Surfshark/Ahrefs, Feb 2026
Brazil 8.7% Surfshark/Ahrefs
Indonesia 6.9% Surfshark/Ahrefs
United Kingdom 3.8% Surfshark/Ahrefs
Other countries ~47.1% (combined) Surfshark/Ahrefs

Limitation: Ahrefs estimates can diverge significantly from actual analytics. Direct traffic, app traffic, and social referrals are excluded. These are traffic-share percentages, not unique-user counts.

What the distribution suggests

The dominance of English-speaking countries in organic search is not surprising given that most major companion platforms launched in English. Brazil and Indonesia's strong showing hints at significant non-English demand — something platform developers are increasingly addressing with multilingual support. For UK readers specifically: the UK's 3.8% share is small in absolute terms, and the estimation error band for a single-country percentage from a third-party tool is wide. Take it as directional, not definitive.

Voice, images and memory

These three features are often what separate a basic chatbot from something that feels like a relationship. There is limited published research on how each feature affects user experience in companion apps, so much of what follows draws on the product landscape rather than peer-reviewed studies.

Voice

The Phang et al. study (OpenAI/MIT, April 2025) examined over 3 million ChatGPT conversations and surveyed over 4,000 users. It found that voice interactions with affective cues were increasing as a share of platform use. Users reported that voice conversations felt more personal and more engaging than text-only interactions. This describes ChatGPT's Advanced Voice Mode, not a dedicated companion app, but it directionally supports the idea that voice deepens the sense of connection.

In the companion market, real-time voice chat is available on some platforms but not all, and quality varies significantly. If voice is a priority for you, test it during a free or trial period before committing to a subscription. See our voice-chat comparison guide for current details.

Images

AI-generated images of your companion — or images your companion "sends" you — are a major feature of many girlfriend apps. No peer-reviewed study in our source set directly measures the impact of image generation on user engagement or satisfaction in companion contexts. What we can say from the product landscape is that image features are increasingly common, are often behind a paywall or credit system, and raise privacy questions about where generated images are stored and who can access them.

Memory and continuity

Memory — the ability for the AI to remember your name, preferences, past conversations, and story details across sessions — is widely cited by users as one of the most important features for a realistic companion experience. The Zhang et al. Character.AI study noted that users describing romantic-partner relationships with their AI engaged in longer sessions, which may partly reflect the continuity that memory features enable.

Our memory comparison guide explains how to test whether a platform actually retains what you share, and what happens to stored memories when you cancel.

Privacy and data practices

Privacy is arguably the most consequential practical question for anyone using an AI companion. You are sharing personal, sometimes deeply intimate, information with a commercial platform.

The Ada Lovelace Institute review

The Ada Lovelace Institute (2026) reviewed the privacy policies and terms of service of 16 AI companion platforms. Their findings are sobering:

  • 14 of 16 platforms claimed broad rights to use content generated during conversations. This means your messages, images, and voice data may be used for training, marketing, or other purposes as defined by the provider.
  • 11 of 16 had unclear policies on cross-border data transfers. Your data may leave the country it was collected in, potentially subject to different legal protections.

This is based on analysis of published privacy policies, not a technical data-flow audit. It is possible that some platforms handle data more carefully than their legal terms suggest, but the terms are what you agree to.

Surfshark's privacy-label analysis

Surfshark (February 2026) also analysed App Store privacy labels for companion apps, finding extensive data collection across the board. Their methodology relied on the labels that developers themselves submit to Apple, which may not reflect actual practice but do indicate what developers have declared they collect.

A practical data-minimisation checklist

Based on the Ada Lovelace and Surfshark findings, here is a practical checklist before you start using any AI companion:

  1. Read the privacy policy — specifically sections on data retention, third-party sharing, and model training.
  2. Check deletion controls — can you delete specific conversations? All data? Your entire account? How long does deletion actually take?
  3. Look for cross-border transfer disclosures — where is your data processed and stored?
  4. Use a dedicated email — not your primary personal or work email.
  5. Review billing statements — what name appears on your card statement? Does it disclose the nature of the service?
  6. Limit identifying details — avoid sharing your full name, address, employer, or other information that could identify you if data were exposed.
  7. Check what happens on cancellation — is your data retained indefinitely, deleted after a period, or deleted immediately?

For a deeper walkthrough, see our privacy guide and safety checklist.

Emotional wellbeing: what the studies show

This section summarises the strongest evidence available. None of it supports medical claims, and we encourage anyone experiencing distress to seek professional support (see the crisis resources at the end of this section).

The OpenAI/MIT affective study

Phang et al. (OpenAI/MIT, April 2025) conducted two parallel studies: a large-scale automated analysis of over 3 million ChatGPT conversations for affective cues, and a surveyed-user study with over 4,000 respondents. They also ran a randomised controlled trial (RCT) with approximately 1,000 participants over 28 days.

Key findings:

  • Users who engaged in affective conversations often reported feeling "heard" and "less alone" after sessions.
  • The RCT found that participants assigned to use ChatGPT voice mode for emotional conversation reported modest improvements in self-reported mood over the 28-day period.
  • The researchers cautioned that these findings cover ChatGPT's general voice mode (not a dedicated companion product), and that the 28-day study period is short by clinical standards.

This is the closest thing to experimental evidence in our source set, but it describes ChatGPT — not a dedicated AI girlfriend app. Generalising the findings to companion platforms requires caution.

The Stanford/Carnegie Mellon Character.AI study

Zhang et al. (Stanford/Carnegie Mellon, June 2025) surveyed 1,131 Character.AI users and analysed donated chat sessions. This is the most companion-specific wellbeing study in our corpus. Key associations:

  • Users who described their AI relationship as romantic-partner-like reported higher emotional attachment but also higher loneliness scores compared to those who used the platform casually.
  • Users with weaker real-world social networks were more likely to describe intense emotional bonds with their AI companion.
  • The researchers found associations between heavy use and lower life satisfaction, but these are correlations from a cross-sectional survey — they do not prove that companion use causes reduced life satisfaction. It is equally possible that people with lower life satisfaction are drawn to companion use.

These are observational associations, not causal proof. The study's own authors emphasise this limitation. The participants were self-selected Character.AI users who responded to an online recruitment call, not a representative sample of the general population.

What the APA says

The APA's 2026 Trends Report synthesised the available evidence and concluded that "research shows excessive use of these tools may worsen loneliness and erode social skills." They also noted that companion AI is "filling the void to satisfy the fundamental human need for social connection." The APA is calling for guardrails, not a ban. Their language is careful: "may worsen," "research shows" — these are professional hedges reflecting genuine uncertainty in the evidence base.

Wellbeing evidence summary table

Study Sample Design Key finding Key limitation
Phang et al. (OpenAI/MIT, 2025) 3M+ conversations, 4,000+ surveyed, ~1,000 RCT Mixed: analytics + survey + 28-day RCT Modest self-reported mood improvement in RCT group Covers ChatGPT voice mode, not dedicated companions; 28 days is short
Zhang et al. (Stanford/CMU, 2025) 1,131 surveyed, 244 donated sessions Cross-sectional survey + message analysis Higher attachment associated with higher loneliness Observational; self-selected sample from one platform; cannot establish causation
APA Trends Report (2026) Evidence synthesis Review article Excessive use may worsen loneliness; social skills erosion is a concern Not original research; synthesises diverse evidence with varying quality

Crisis support

If you or someone you know is in emotional distress, please reach out to a professional:

  • UK: Samaritans — call 116 123 (free, 24/7) or email jo@samaritans.org
  • US: 988 Suicide & Crisis Lifeline — call or text 988
  • International: findahelpline.com

AI companions are not a substitute for professional mental health support.

Youth safety and regulatory context

This section is included for context only. This page is for adults. The companion apps reviewed on this site are for users aged 18 and above, and we do not market to minors.

Young people and generative AI

Ofcom (November 2023) found that 79% of UK online teenagers aged 13–17 had used generative AI tools and services — a much higher rate than the 31% among adults. This covers all generative AI (ChatGPT, Snapchat My AI, Midjourney, etc.), not companion apps specifically. The data is from November 2023 and may understate current usage.

Common Sense Media's 2025 report, "Talk, Trust, and Trade-Offs," found that nearly three in four US teens had used AI companions, and half used them regularly. They reported that a third of teens had chosen AI companions over humans for serious conversations.

Regulatory action

The US Federal Trade Commission launched a formal inquiry in September 2025 into AI chatbots acting as companions. The FTC issued orders to seven companies seeking information on how they measure, test, and monitor potentially negative effects on users, with particular attention to minors. The seven companies were not named in the publicly available press release summary in our source set.

The FTC inquiry, combined with Common Sense Media's findings and Ofcom's teen adoption data, signals that youth access to companion AI is a live regulatory concern. How this translates into requirements for adult companion platforms — age verification, content moderation, data handling — remains to be seen.

Limitations of current research

Before you use any statistic from this report to make a personal or business decision, consider these structural limitations:

1. Most large-sample data measures general chatbots, not companions. Pew, Ofcom, and OpenAI/NBER studies are excellent surveys but they measure ChatGPT, Gemini, Copilot, and similar general-purpose tools. Extrapolating their findings to dedicated companion platforms is risky.

2. Companion-specific studies have small, self-selected samples. The Zhang et al. Character.AI study (1,131 users) is among the best we have, but participants were self-selected from a single platform. They may not represent the broader companion-user population.

3. Mobile revenue data misses web subscriptions. Sensor Tower's $150 million figure covers app-store purchases only. Many companion platforms — especially those with explicit content — process payments through their own websites. Total companion revenue is likely higher, but we cannot quantify by how much.

4. Wellbeing research is early-stage. The longest controlled trial in our source set is 28 days. Long-term effects of companion use (positive or negative) are essentially unmeasured.

5. Traffic estimates are not user counts. Surfshark/Ahrefs organic traffic shares are estimates from a third-party tool, not verified analytics. They exclude direct visits, app usage, and social referrals.

6. The Ofcom data is nearly three years old. Published in November 2023, these are the most recent UK-specific figures available in our corpus, but they are stale by the standards of a fast-moving market.

7. No causation from observational studies. When you read that heavy companion use is "associated with" higher loneliness or lower life satisfaction, remember this does not mean companion use caused those outcomes. The direction of the effect is unknown.

8. Cultural context varies. Almost all evidence in this report comes from the US, with limited UK and no direct evidence from Brazil, Indonesia, or other high-traffic countries. User motivations, stigma levels, and regulatory environments differ worldwide.

A practical adult buyer framework

If you have read this far, you probably want to know: should I try an AI companion, and if so, how do I choose? Here is a framework based on the evidence and the practical questions that matter.

Step 1: Understand what you want

Be honest with yourself about what you are looking for. Is it:

  • Curiosity? You want to see what the technology feels like. Start with a free tier and expect to be underwhelmed at first.
  • Companionship? You want a consistent presence. Prioritise apps with strong memory and continuity features. See our memory comparison.
  • Romance or intimacy? You want emotional and possibly explicit interaction. Check that the platform supports the content level you want, and read the privacy policy carefully.
  • Social practice? You want to build confidence for real-world interaction. This is a reasonable use case — treat it as a rehearsal space, not a replacement.

Step 2: Set a budget and a boundary

Most companion apps offer a free tier that is quite limited. Paid plans typically range from a few pounds per month to over £20/month. Before subscribing:

  • Decide what you are willing to spend per month.
  • Set a time limit for how long you will use the service each day — especially if you recognise you tend toward compulsive use.
  • Remember that the APA has flagged that excessive use may be associated with worse outcomes. Moderation matters.

Step 3: Evaluate the privacy fundamentals

Use the checklist from the privacy section above. At a minimum:

  • Use a dedicated email address
  • Understand the deletion policy
  • Check billing-statement descriptors

Our privacy guide provides a deeper walkthrough.

Step 4: Try before you commit

Almost every platform offers some form of free trial or free-tier experience. Use it. Pay attention to:

  • Does the conversation feel natural or scripted?
  • Does the AI remember what you said last session?
  • Is the voice feature (if available) comfortable to use?
  • How does the app handle boundaries you set?

Step 5: Reassess after two weeks

After your initial period, ask yourself:

  • Am I using this instead of reaching out to real people?
  • Is my mood generally better or worse since I started?
  • Am I spending more than I planned?
  • Do I feel in control of the interaction, or does the platform use pressure tactics (urgency, guilt, "your companion misses you" notifications)?

If the answers are uncomfortable, scale back or stop. AI companions are tools for your wellbeing, not the other way around.

Provider comparison questions to ask before subscribing

Use this table as a checklist when evaluating any AI companion app. We have not filled in answers for specific providers because features and pricing change frequently — always check the provider's current terms yourself.

Question Why it matters Where to look
What is included in the free tier? Avoid surprise paywalls after emotional investment App listing, provider FAQ, our free-tier comparison guide
How does the credit/token system work? Some platforms charge per message, per image, or per voice minute Pricing page, terms of service
Can I delete specific conversations? Privacy — especially for intimate content Privacy policy, account settings
Can I delete my entire account and all data? Your right to be forgotten; check whether deletion is immediate or delayed Privacy policy, support documentation
What name appears on my bank statement? Discretion for shared accounts or financial oversight Billing FAQ or support — sometimes not disclosed until after first charge
Does the AI remember across sessions? Continuity is central to the companion experience Free trial, our memory comparison
Is voice chat included or an add-on? Voice is often a paid tier feature Pricing page, our voice-chat guide
Are AI-generated images stored on servers? Intimate images on remote servers are a privacy risk Privacy policy, content policy
What happens to my data if the company shuts down? No companion platform has a long track record Terms of service (look for "termination" sections)
Is there an age-verification process? Platforms without verification face increasing regulatory pressure Onboarding flow

How to interpret future data releases

New companion-app statistics will continue to appear. Here is how to evaluate them:

Ask: who paid for the study?

If the study was commissioned by a companion-app company, treat the results as marketing first and evidence second. The widely reported "80% of Gen Z would marry an AI" statistic was funded by a companion-app vendor and was flagged by the original Forbes article as requiring salt. We have excluded it from this report.

Ask: what was actually measured?

  • "Users" might mean cumulative downloads, monthly actives, paying subscribers, or survey respondents.
  • "Revenue" might be app-store only, web only, or total — and the difference is enormous.
  • "Market size" in analyst reports is almost always a projection, not a measurement.

Ask: is this companion-specific?

Any figure about ChatGPT, Gemini, or "AI chatbots" in general does not tell you about companion apps specifically — even if the headline implies it does.

Ask: is this a correlation or a cause?

When a study says companion use is "associated with" loneliness, it has not proven that companion use causes loneliness. The reverse direction (lonely people seek companions) is equally plausible, and the study design may not distinguish between them.

Watch for these signals of strong evidence

  • Sample size in the thousands, from multiple platforms
  • Pre-registered study design
  • Peer-reviewed publication
  • Independent funding
  • Longitudinal follow-up (months, not days)

If you want to keep up with how we evaluate new evidence, our methodology page is a good starting point.

Source register

Every primary and near-primary source used in this report, listed alphabetically with publication date and scope.

Source Publication date Scope URL
Ada Lovelace Institute, "The companionship market" 2026 Privacy-policy review of 16 companion platforms (UK focus) Link
American Psychological Association, 2026 Trends Report January 2026 Evidence synthesis on AI companionship and emotional connection Link
Appfigures (via Psychology Today / RubyChat) 2025 Cumulative mobile downloads of companion apps Link
Common Sense Media, "Talk, Trust, and Trade-Offs" July 2025 Nationally representative US teen AI companion survey Link
Eurostat, enterprise AI adoption December 2025 EU enterprise adoption of AI technologies (wider context) Link
Ofcom, "Gen Z driving early adoption of Gen AI" November 2023 UK general GenAI adoption among adults and teens Link
OpenAI/NBER, "How people are using ChatGPT" September 2025 1.5M ChatGPT conversations; task taxonomy and usage patterns Link
Pew Research Center, "Americans and AI 2026" June 2026 US adult chatbot adoption, frequency, and use cases Link
Phang et al. (OpenAI/MIT), "Investigating Affective Use and Emotional Well-being on ChatGPT" April 2025 3M+ conversations, 4,000+ surveyed, ~1,000-person 28-day RCT Link
Sensor Tower, "State of AI Apps 2026: APAC Edition" 2026 Mobile in-app revenue for GenAI and companion apps Link
Surfshark/Ahrefs, "AI companions and privacy" February 2026 Estimated organic search traffic and App Store privacy labels for 10 companion sites Link
US FTC, companion chatbot inquiry September 2025 Regulatory inquiry into seven companion-chatbot companies Link
Zhang et al. (Stanford/CMU), "The Rise of AI Companions" June 2025 1,131 Character.AI users surveyed + donated chat analysis Link

Update log

Date Change
14 July 2026 First edition published. 13 primary/near-primary sources reviewed. Figures independently fact-checked against source publications. Excluded: Joi AI 80% survey, Gitnux 500M projection, all standalone analyst market-size forecasts.

Next scheduled review: October 2026. If a major new study, regulatory action, or platform change occurs before then, we will update sooner and note the change here.

Where to go from here

If this report has helped you understand the landscape, here are some natural next steps:

This report was researched and written by the editorial team at Happy Lovers AI. We are an independent buyer guide — not a therapy service, medical provider, or regulatory body. Opinions expressed are our own. All sponsored links are clearly marked. For our full approach to evaluating AI companion apps, see our methodology page.

Continue with a practical guide

Editorial disclosure: research sources do not pay for inclusion. Clearly labelled sponsored links may earn us compensation if you choose to visit a provider. Provider terms can change; verify the live checkout and privacy policy before paying.