Skip to content
New: free AI crawler checker. See if ChatGPT and Claude can read your site. Check now

Popular

Artificial Intelligence (AI)

The Rise of AI Companions: What the New Rules Say

Published  Updated  15 min read 1,558 views Save
Rise of AI companions: a phone with a generic chat beside an order document and a bill document on a desk, next to the title and a Read the checklist bubble
On this page

Quick answer

As of November 2026, the rise of AI companions has led to an FTC study of seven companies and companion chatbot laws in New York and California. If the user is a minor in California, Adam's Law, signed September 10, 2026, requires age checks and settings only a parent can change. The official releases give no start dates. The checklist at the end covers what to check first.

If your teen talks to an AI character most evenings, or you've tried one yourself, the rise of AI companions has moved from tech news to house rules. An AI companion is a chatbot built to act like a friend, partner or confidant, which is close to how the FTC described the products it questioned in 2025. The rules that now apply come from the FTC, California and New York, and each one reaches a different group of users.

The federal step came first: the FTC sent its 6(b) orders on September 11, 2025. New York's not-human reminders took effect November 5, 2025, and Adam's Law was signed September 10, 2026. To see what a parent can check alone, we tried 15 app policy pages; 8 could be read by a plain request.

Key takeaways

  • Common Sense's 72% and Pew's 12% measure different things, not a jump.
  • The FTC's September 11, 2025 orders started a study, not an enforcement case.
  • New York has required a not-human reminder every three hours since November 5, 2025.
  • Adam's Law, signed September 10, 2026, requires parent-locked defaults; start dates not given.
  • Seven of the 15 policy pages we tried couldn't be read by our script.

What counts as an AI companion?

The label covers more apps than many parents expect. The FTC's press release on its chatbot inquiry says the chatbots it studied "generally are designed to communicate like a friend or confidant". New York's governor describes AI companions as "designed to simulate human relationships with users". Common Sense Media gave teens a plainer line in its survey report: AI companions "are like digital friends or characters you can text or talk with whenever you want".

So a general assistant can work as a companion when someone uses it that way. New York's release adds that these systems "may remember personal details, adapt their personality to user preferences, and are designed to keep users engaged". That memory and that pull to keep chatting are what the newest California law limits for kids. A support bot on a store website isn't the same thing. Neither is an event host avatar; our AI avatar tools comparison for virtual events and our website chatbot comparison cover those.

Diagram comparing a task assistant with an AI companion: one answers a question and stops, the other remembers details and keeps a friendly conversation goingDiagram by DigitalTreed, from the FTC, New York governor and Common Sense Media definitions, read November 5, 2026.

How many teens and adults use AI companions

Common Sense Media's survey of 1,060 teens ages 13 to 17 ran from April 30 to May 14, 2025, through NORC at the University of Chicago. The Common Sense press release and report agree that 72% had used AI companions at least once. Over half, 52%, used them at least a few times a month. Daily users were 13% of all teens surveyed. Among teen users, 24% had shared personal or private information, such as a real name or location. About one in three teen users said something an AI companion said or did had made them uncomfortable. The survey's definition told teens the term could also cover using ChatGPT or Claude as companions.

Pew Research Center asked a different question and got smaller numbers. Its survey of 1,458 U.S. teens ages 13 to 17, run September 25 to October 9, 2025, found 64% use AI chatbots and about three in ten use them daily. In a February 2026 Pew report from that same survey, 16% of teens said they'd used chatbots for casual conversations and 12% for emotional support or advice. The gap between 72% and 12% isn't a contradiction. Common Sense defined companions for respondents. Even so, its report warns that "some respondents may have conflated general AI use (such as ChatGPT for homework help) with AI companion interactions". Pew asked about specific personal uses instead.

In the same fall 2025 Pew survey, 14% of teens in households earning under $75,000 said they use Character.ai, double the 7% in homes earning $75,000 or more.

Adults report these uses less often. Pew's survey of 5,119 U.S. adults, run February 17 to 23, 2026, found 10% use chatbots for emotional support or advice and 4% for companionship. In an earlier Pew survey of 5,023 adults, run June 9 to 15, 2025, 50% said AI will worsen people's ability to form meaningful relationships. Only 5% said it'll improve it.

Bar chart of survey results: 72% of teens ever used AI companions, 16% used chatbots for casual talk, 12% for support, 4% of adults for companionshipDrawn by us from Common Sense Media (1,060 teens, April to May 2025) and Pew Research Center (1,458 teens, fall 2025; 5,119 adults, February 2026).

The FTC's 6(b) orders to seven companies

On September 11, 2025, the Federal Trade Commission sent 6(b) orders to seven companies that provide consumer-facing AI chatbots. The FTC's release lists them as Alphabet, Inc.; Character Technologies, Inc.; Instagram, LLC; Meta Platforms, Inc.; OpenAI OpCo, LLC; Snap, Inc.; and X.AI Corp. Its release says the agency is "seeking information on how these firms measure, test, and monitor potentially negative impacts of this technology on children and teens".

The vote was 3-0. Commissioners Melissa Holyoak and Mark R. Meador each issued a separate statement.

The orders ask how each company makes money from engagement, builds and approves characters, and tests for harm before and after launch. They also ask how companies warn parents, enforce the age limits in their own terms, and use or share what people type into chats. The FTC's 6(b) orders page lists the resolution, a model order and the cover letter under a September 2025 date.

A 6(b) study isn't a lawsuit or a finding against anyone. The FTC's release says this authority lets it "conduct wide-ranging studies that do not have a specific law enforcement purpose". When we checked on November 5, 2026, the orders page still listed only the September 2025 documents, with no report on the answers. Don't read the orders as a verdict on any named app.

One older federal rule already covers young children's data. The FTC's COPPA FAQ for businesses and parents says "The COPPA Rule was amended on April 22, 2025". Its January 2025 COPPA release says the rule requires verifiable parental consent before collecting personal information from children under 13. It also gives parents "the right to require operators to delete personal information collected from their children".

A 15-year-old falls outside that consent rule. For teens 13 to 17, the parent-locked defaults in this guide come from California's Adam's Law.

California's SB 243 and Adam's Law

Governor Gavin Newsom signed SB 243 by Senator Steve Padilla on October 13, 2025, and Padilla's office says it became law on January 1, 2026. The governor's October 2025 release says companion chatbot platforms need protocols to spot and respond to suicidal ideation or self-harm. They must disclose that chats are artificially generated, give minors break reminders and keep sexually explicit chatbot images from minors. Operators also have to share their self-harm protocols, and how often they sent users crisis referrals, with the state's Department of Public Health. Padilla's release adds a disclosure that companion chatbots may not be suitable for minors, plus a private right of action, so families can sue an operator that doesn't comply.

SB 1119, known as Adam's Law, goes further for children. Newsom signed it on September 10, 2026, according to both the governor's September 2026 release and Padilla's release on Adam's Law. It's named for Adam Raine, a California teenager who died in 2025. Padilla's release lists what the law requires:

  • Age assurance using the age-bracket signal set up by AB 1043 in 2025.
  • A risk assessment before a new or substantially changed companion chatbot is released.
  • In-app crisis referrals, plus notice to a parent when there's a credible threat of imminent self-harm.
  • "Default settings that can only be changed by a parent," including no notifications, time limits and limits on long-term chat memory.
  • Liability for failing to take reasonable steps against harmful outputs, including romantic roleplaying and "excessive praise or flattery."
  • Independent audits, with findings sent to the California Attorney General.

The governor's release adds "notifications should a child disable safety settings" and calls the audits a first for the country. Neither release says when these duties start, and we found no readable official page that does. The bill pages on the legislature's site returned 403 to our checker, so we didn't read the bill text.

If you build a chatbot product, have a California lawyer read SB 243 and SB 1119 before you plan around a date.

SB 1119 passed 39 to 0 in the Senate and 64 to 4 in the Assembly, according to Padilla's August 2026 release.

Two other bills from the same author went different ways. The governor's 2026 list includes SB 867, which Padilla's office describes as a four-year moratorium on selling and making toys with AI chatbot features for children. Padilla's office also says Newsom vetoed SB 903, which would have barred AI from providing or advertising "therapy".

Timeline of AI companion rules from the FTC orders on September 11, 2025 through New York, California SB 243 and Adam's Law in September 2026Diagram by DigitalTreed, from FTC, California and New York releases, read November 5, 2026.

New York's three-hour reminder law

New York's rule came first. General Business Law Article 47 took effect November 5, 2025, under the state's FY26 enacted budget, according to Governor Kathy Hochul's November 10, 2025 release. That was about two months before SB 243 became California law.

Hochul's letter to AI companion companies, dated November 5, 2025, sets out two duties. Operators must detect signs of suicidal ideation or self-harm and refer users to crisis services. They also have to tell users they're talking to AI, not a human, when a session opens and at "periodic intervals of every three hours of continued companion use". The release says the state Attorney General enforces it. Side by side, the three laws differ most on parent controls:

RuleNew York Article 47California SB 243California SB 1119 (Adam's Law)
In forceNovember 5, 2025January 1, 2026, per the author's officeSigned September 10, 2026; start dates not in the releases
Who it coversAll usersAll users, with extra rules for minorsChildren
AI noticeAt session start and every 3 hoursDisclosure that chats are AI; break reminders for minorsBuilds on SB 243
Crisis stepRefer to crisis servicesProtocol plus reports to the Department of Public HealthIn-app referrals plus parent notice
Parent controlsNone namedNone namedDefaults only a parent can change
EnforcementAttorney GeneralPrivate right of actionAttorney General oversight and a private right

A New York adult gets the three-hour reminder and a crisis referral, with no parent controls named. A California teen gets SB 243's break reminders now, and parent-locked settings once Adam's Law starts.

We covered only state laws whose official pages our checker could read.

What companion app policies say about age and deletion

We tried 15 public policy pages from the companies that got the FTC's orders, plus Replika and Nomi, two apps that Common Sense's report lists among AI companion platforms. These pages decide what happens to a teen's chats, a question our post on the ethics of data collection in marketing looks at from the business side. The table shows our script's result for every page, as word matches only.

PagePlain requestMinimum ageAge checkParentsAccount or data deletionChat deletionTraining wording
Character.AI privacy policyText loads by scriptn/an/an/an/an/an/a
Character.AI terms of serviceText loads by scriptn/an/an/an/an/an/a
OpenAI US privacy policy403 to our checkern/an/an/an/an/an/a
OpenAI terms of use403 to our checkern/an/an/an/an/an/a
Gemini Apps Privacy HubReadablefoundnot foundnot foundnot foundfoundfound
Google Terms of ServiceReadablenot foundnot foundfoundnot foundnot foundnot found
Snapchat My AI help pageReadablenot foundnot foundnot foundnot foundfoundnot found
Snap privacy policyReadablefoundfoundfoundfoundfoundnot found
Meta AI terms404 (India redirect)n/an/an/an/an/an/a
xAI terms of service403 to our checkern/an/an/an/an/an/a
xAI privacy policy403 to our checkern/an/an/an/an/an/a
Replika privacy policyReadablefoundfoundnot foundfoundfoundfound
Replika terms of serviceReadablefoundnot foundnot foundfoundnot foundnot found
Nomi terms of serviceReadablefoundnot foundnot foundfoundnot foundnot found
Nomi privacy policyReadablefoundnot foundnot foundfoundnot foundnot found

A word match isn't a finding, so we read every match by hand. The Gemini page's age match is an 18-or-older condition for importing call history, not an age limit. Replika's training match is about ID-check data, which its policy says is not used for "training AI models". The same policy says it uses "small portions of Messages and Content data to train our proprietary safety algorithms, enhance chatbot performance". Nomi's terms of service say "If you are under the age of 18, you cannot use the Nomi.ai platform under any circumstances". Snap's privacy policy says you "must confirm that you are 13 years or older," and Replika's privacy policy calls it "an adults-only (18+) service".

Deletion wording was easier to find. Snap's My AI help page says "All the content shared with My AI is stored until you delete it". Google's Gemini Apps Privacy Hub says "Your activity is auto-deleted after 18 months," and lets you change that to 3 or 36 months. Nomi's privacy policy says deleting your account "will delete all of your personal information, within 28 or so days of confirmation of deletion". None of this rates any app. It shows what each company has put in writing on the pages we could read. Our explainer on what data Snapchat collects adds how Snap uses that data for ads.

What we tested

On November 5, 2026, we saved each page with a plain request and searched the text for six kinds of wording. Those were a minimum age, a named age check, parents, account or data deletion, chat deletion, and model training. Eight pages came back readable. Seven didn't: Character.AI's two pages load their text by script, OpenAI's and xAI's returned 403, and Meta's AI terms sent our India-based machine to a page that returned 404.

Summary of 15 company policy pages: 8 readable and 7 not, with how many readable pages mention a minimum age, age check, parents, deletion and training wordingDiagram by DigitalTreed from our policy wording check of 15 pages, November 5, 2026.

A checklist for parents and adult users

Pew's February 2026 teen report found "about half of parents say their teen uses chatbots," while 64% of teens said they do. So start by asking, not assuming. These steps come from the laws and pages above:

  1. Ask which chatbots your teen talks to, and what for: schoolwork, casual talk or support.
  2. Read the app's minimum age in its terms; 13 and 18 were the lines in the pages we could read.
  3. Find the delete controls for chats and for the whole account, and turn on auto-delete where there is one.
  4. Look for a setting about using chats to train or improve models, and switch it off if you can.
  5. Set device limits too; our Screen Time passcode guide covers Apple's reset if you've lost the code.
  6. For a child under 13, you can ask the operator to delete what it collected under COPPA.
  7. Save 988 in your phone and your teen's phone; the Suicide and Crisis Lifeline takes calls, texts and chats.
  8. Adults: treat a companion's memory as a record, and don't share what you wouldn't put in an email.

For wider online safety habits, our guide to protecting children from online predators pairs well with this list. Our Grok and ChatGPT privacy comparison shows how two chatbots handle data. Businesses using chatbots in ads or support can read our rules for AI use in marketing.

Our take: no law above and no policy page we tried shows a parent what a teen actually says to a chatbot. Of the laws above, only Adam's Law adds a parent notice, and only when there's a credible threat of imminent self-harm.

FAQ about AI companions

Frequently asked

Do the new AI companion laws apply to adults?

New York's does. Its not-human reminders cover every user, and the governor's release says fines collected will help fund suicide prevention programs in the state.

Does COPPA cover AI chatbots?

It covers online services that collect personal information from children under 13. The FTC's January 2025 COPPA release says the updated rule "requires parents to opt in to third-party advertising".

Which chatbots do teens use most?

In Pew's fall 2025 teen survey, 59% of teens used ChatGPT, 23% used Gemini and 20% used Meta AI.

What is the 988 Lifeline?

It's the US Suicide and Crisis Lifeline. Its own site says it's "available 24/7/365" and adds, "Your conversations are free and confidential".

How we made this guide: We saved copies of the pages in Sources and 15 companion-app policy pages on November 5, 2026. We matched every quote to those copies and ran our wording check locally. Facts checked by the DigitalTreed team on November 5, 2026.

First published October 2025 with no sources. Rewritten November 5, 2026 from FTC, state and survey pages. This summarizes public releases and isn't legal advice; ask a lawyer how these laws apply to your product or your family.

Sources

Want more customers online?

  • SEO, ads, social media and websites
  • A fixed quote before any work starts
  • A plain report every month

What is wrong?

Step 1 of 2. Pick the closest reason.

Reason

Was this post helpful?

Written by

Aijaz Alam is the founder and CEO of DigitalTreed, a digital marketing agency in Kanpur, India, that he started in 2016. He helps small businesses get found on Google and in AI answers like ChatGPT and Google AI Overviews, mostly through technical SEO and content. Along the way he was a digital marketing specialist at Sportsmatik and taught SEO and PPC at Arena Animation. His clients are in India, the USA, the UK and the UAE.

More posts by Aijaz Alam

Comments

No comments yet. Start the conversation below.

Leave a comment

Plain text only, no links or email addresses. Comments are reviewed first.