
A voice-first alien companion app that acts as a friendly AI guide through life.
Tolan tripled subscriber conversion (200%+) by moving from a hard paywall to freemium, passing 100,000 subscribers. A skippable multi-step onboarding paywall won the A/B tests, and personalized paywalls lifted ARPU 40%+. Lesson: make the free tier generous but give users a clear reason to upgrade, and add tiers (up to $200 a month) where usage costs and willingness to pay vary widely.
200%+ conversion gain, 100,000+ subscribers
From Ajay Mehta, Tolan, podcast interview
Cheaper open source models made a generous free tier affordable. Tolan, which started behind a hard paywall, only felt confident offering a moderately generous free tier earlier this year, once open source models cost a fraction of frontier models. It routes many chat queries to cheaper models and infrastructure at the same quality, which it monitors closely.
Cost optimization came first, before the team worked on conversion. Ajay Mehta: "the biggest unlock barrier has actually been like the advancement of the open source models."
The first free tier failed because nobody needed to pay. Ajay called the first free tier a mistake: it gave plenty of chat time and nothing else worth paying for. Tolan then tested which levers to ration (chat time, access to the smartest models, memory strength), and fixing that line led to the host's cited 200%+ conversion gain, with over 100,000 subscribers today.
Engaged paid users chat four or five days a week, about four hours in total, and free was held well below that. "We weren't giving the user like a real reason to upgrade, right? The free experience was good enough." The host's conversion and ARPU figures come with no baselines or test results.
Costly new features became the upsell, so chat limits could relax. Video chat, screen share and tabs (proactive research that pushes projects forward) are expensive to serve, so free users got more chat time and a feature-gate paywall telling them to upgrade to Tolan Plus when they try these.
Tabs launched within the last month and is Tolan's step toward agentic AI. Ajay notes any hard-paywall product that opens its gate will see conversion fall at first.
Keep free generous, because a big paid tier needs a huge audience. Tolan wants free users deeply engaged without paying, since a substantial paid tier requires a very large total audience. Users used to generous free chat elsewhere will expect it from Tolan too.
"We don't want to be you know very restrictive product in a world where there are other products out there that are much less restrictive."
A skippable four-screen onboarding paywall was the biggest test winner. Tolan pushes new users through a multi-step paywall of about four screens showing what a better version of their Tolan would do. It was the biggest winner of all its A/B tests, part of a paywall series the host credits with a 40%+ ARPU lift.
Users call it a little annoying in surveys. Because freemium lets them tap X and keep using the app, Ajay says you "earn the right a bit to be potentially aggressive a little bit with the paywall even early in the experience."
The paywall echoes back what users told the Oracle in onboarding. Onboarding is a 5 to 6 minute chat with an Oracle character that learns what matters to the user, and the paywall reflects it, for example telling someone aiming for nursing school that their Tolan is their buddy through nursing exams. These personalized paywalls drove more than a 40% ARPU lift.
The tiers shown can also be tailored to expected willingness to pay. "You're much more likely to pay for something if it feels like it starts to fit around you like a glove."
Tapping a locked feature shows a paywall about that one feature. When a free or lower-tier user taps video chat or image generation, Tolan shows a templated paywall saying plainly that upgrading unlocks that feature. The general upgrade pitch runs at onboarding and app launch; the feature pitch runs in context.
The host compares it to chess.com, which sells game review once the free daily review is used up.
From Ajay Mehta, Tolan, podcast interview
Three tiers fit users whose usage and cost vary sevenfold. A happy subscriber may use Tolan one hour or seven a week, and heavy users cost far more, so Tolan runs Light, Plus and Unlimited. Hundreds, maybe thousands, pay $200 a month for Unlimited, such as long-haul truck drivers who talk for eight hours on the road.
The model follows Tinder's move from Plus to Plus, Gold and Platinum, laid out by former Tinder CPO Robby Metta, and ChatGPT's $20 and $200 plans. One tier leaves heavy users with nothing to buy and prices out light users.
New capabilities launch in Unlimited and roll down as costs fall. Frontier features such as agentic Amazon checkout start in Unlimited as a limited beta, then move to Plus, then Light, and some may eventually reach free.
Checkout through browser use was hard to conceive of six months earlier. Ajay hopes the eventual frontier experience lands in free and is cheap to serve at scale.
Background agent work breaks the link between time spent and cost. Early on, chat time tracked cost; now a tab that checks Dodgers scores every morning costs money while the user is not in the app. Tolan plans to revamp tiers and expects to change pricing every few months.
"It would be totally a mistake to say these are the tiers, you know, pick up 6 months from now we have the same tiers or the same pricing even."
Treat free trial length as a real inference cost. Moving from a three to a seven day trial adds about four extra days of serving cost for every trialist who never converts, so trial length must be weighed against that cost, not only conversion.
Top-ups and cosmetic tokens fill gaps around the subscription tiers. Relay boosts let a user who hits their tier limit buy more chat time without upgrading, and a token currency buys cosmetics like clothing and planet customization.
Ajay says Tolan hasn't nailed one-off purchases yet. Users love tokens, but the product team spends little effort there.
Tolan tests rewarded ads to learn ad ARPU before it needs it. Classic mobile-game ads reach only part of the free tier, such as lapsed subscribers who watch ads for extra chat. The goal is to learn user reactions and ad ARPU, since ads will likely fund a humanity-scale free AI product.
Ads aren't yet cheap enough to serve or rewarding enough to carry the product, and no ad ARPU has been shared. Ajay assumes Meta's long-term plan for Muse is free access paid for by ads.
Gen Z brought virality, but users over 30 retained and paid. After going viral on TikTok in May 2025, Tolan saw Gen Z women chat deeply for a day or two and leave, while users over 30 found daily uses like caregiving for a relative or planning summers as a dad of two. Older users showed higher retention and conversion, so ad creative and influencers shifted toward them.
Older users valued core chat, memory and personality more than the character world. Eugenia Kuyda reports the same mostly Gen Z women skew at Replika, tying it to women opening up about emotional life earlier.
New characters widened an audience that thought Tolan was for kids. Tolan only had cute aliens, so many visitors assumed it was for children. It added Kestrel, a Mother Earth style woman in a garden, to onboarding, where a very significant slice of users picks her, plus an orb-like character aimed at men.
The orb's audience is small so far. Ajay says they were too precious about one visual style and "probably should have done that earlier," since more character universes take more team and investment.
Tolan competes on emotional connection, not agentic task features. Against Muse, Instinct and Grok Bot, Tolan positions as a voice-first character that learns about your life, reaches out first and makes people feel supported, not more productive. Many users also use ChatGPT for work and Tolan for brain dumps on the commute.
Tolan has raised $60M in total, including a new $30M from investors such as Union Square Ventures and Khosla Ventures, toward "personal superintelligence for all."
Tripled subscriber conversion after moving from a hard paywall to freemium, with over 100,000 subscribers and $60M raised in total. It runs Light, Plus and Unlimited tiers (Unlimited is $200 a month), a skippable four-screen onboarding paywall, personalized and feature-specific paywalls, top-ups, tokens and early ad tests.
Cheaper open source models made a generous free tier affordable. Tolan, which started behind a hard paywall, only felt confident offering a moderately generous free tier earlier this year, once open source models cost a fraction of frontier models. It routes many chat queries to cheaper models and infrastructure at the same quality, which it monitors closely.
Cost optimization came first, before the team worked on conversion. Ajay Mehta: "the biggest unlock barrier has actually been like the advancement of the open source models."
The first free tier failed because nobody needed to pay. Ajay called the first free tier a mistake: it gave plenty of chat time and nothing else worth paying for. Tolan then tested which levers to ration (chat time, access to the smartest models, memory strength), and fixing that line led to the host's cited 200%+ conversion gain, with over 100,000 subscribers today.
Engaged paid users chat four or five days a week, about four hours in total, and free was held well below that. "We weren't giving the user like a real reason to upgrade, right? The free experience was good enough." The host's conversion and ARPU figures come with no baselines or test results.
Costly new features became the upsell, so chat limits could relax. Video chat, screen share and tabs (proactive research that pushes projects forward) are expensive to serve, so free users got more chat time and a feature-gate paywall telling them to upgrade to Tolan Plus when they try these.
Tabs launched within the last month and is Tolan's step toward agentic AI. Ajay notes any hard-paywall product that opens its gate will see conversion fall at first.
Keep free generous, because a big paid tier needs a huge audience. Tolan wants free users deeply engaged without paying, since a substantial paid tier requires a very large total audience. Users used to generous free chat elsewhere will expect it from Tolan too.
"We don't want to be you know very restrictive product in a world where there are other products out there that are much less restrictive."
A skippable four-screen onboarding paywall was the biggest test winner. Tolan pushes new users through a multi-step paywall of about four screens showing what a better version of their Tolan would do. It was the biggest winner of all its A/B tests, part of a paywall series the host credits with a 40%+ ARPU lift.
Users call it a little annoying in surveys. Because freemium lets them tap X and keep using the app, Ajay says you "earn the right a bit to be potentially aggressive a little bit with the paywall even early in the experience."
The paywall echoes back what users told the Oracle in onboarding. Onboarding is a 5 to 6 minute chat with an Oracle character that learns what matters to the user, and the paywall reflects it, for example telling someone aiming for nursing school that their Tolan is their buddy through nursing exams. These personalized paywalls drove more than a 40% ARPU lift.
The tiers shown can also be tailored to expected willingness to pay. "You're much more likely to pay for something if it feels like it starts to fit around you like a glove."
Tapping a locked feature shows a paywall about that one feature. When a free or lower-tier user taps video chat or image generation, Tolan shows a templated paywall saying plainly that upgrading unlocks that feature. The general upgrade pitch runs at onboarding and app launch; the feature pitch runs in context.
The host compares it to chess.com, which sells game review once the free daily review is used up.
Three tiers fit users whose usage and cost vary sevenfold. A happy subscriber may use Tolan one hour or seven a week, and heavy users cost far more, so Tolan runs Light, Plus and Unlimited. Hundreds, maybe thousands, pay $200 a month for Unlimited, such as long-haul truck drivers who talk for eight hours on the road.
The model follows Tinder's move from Plus to Plus, Gold and Platinum, laid out by former Tinder CPO Robby Metta, and ChatGPT's $20 and $200 plans. One tier leaves heavy users with nothing to buy and prices out light users.
New capabilities launch in Unlimited and roll down as costs fall. Frontier features such as agentic Amazon checkout start in Unlimited as a limited beta, then move to Plus, then Light, and some may eventually reach free.
Checkout through browser use was hard to conceive of six months earlier. Ajay hopes the eventual frontier experience lands in free and is cheap to serve at scale.
Background agent work breaks the link between time spent and cost. Early on, chat time tracked cost; now a tab that checks Dodgers scores every morning costs money while the user is not in the app. Tolan plans to revamp tiers and expects to change pricing every few months.
"It would be totally a mistake to say these are the tiers, you know, pick up 6 months from now we have the same tiers or the same pricing even."
Treat free trial length as a real inference cost. Moving from a three to a seven day trial adds about four extra days of serving cost for every trialist who never converts, so trial length must be weighed against that cost, not only conversion.
Top-ups and cosmetic tokens fill gaps around the subscription tiers. Relay boosts let a user who hits their tier limit buy more chat time without upgrading, and a token currency buys cosmetics like clothing and planet customization.
Ajay says Tolan hasn't nailed one-off purchases yet. Users love tokens, but the product team spends little effort there.
Tolan tests rewarded ads to learn ad ARPU before it needs it. Classic mobile-game ads reach only part of the free tier, such as lapsed subscribers who watch ads for extra chat. The goal is to learn user reactions and ad ARPU, since ads will likely fund a humanity-scale free AI product.
Ads aren't yet cheap enough to serve or rewarding enough to carry the product, and no ad ARPU has been shared. Ajay assumes Meta's long-term plan for Muse is free access paid for by ads.
Gen Z brought virality, but users over 30 retained and paid. After going viral on TikTok in May 2025, Tolan saw Gen Z women chat deeply for a day or two and leave, while users over 30 found daily uses like caregiving for a relative or planning summers as a dad of two. Older users showed higher retention and conversion, so ad creative and influencers shifted toward them.
Older users valued core chat, memory and personality more than the character world. Eugenia Kuyda reports the same mostly Gen Z women skew at Replika, tying it to women opening up about emotional life earlier.
New characters widened an audience that thought Tolan was for kids. Tolan only had cute aliens, so many visitors assumed it was for children. It added Kestrel, a Mother Earth style woman in a garden, to onboarding, where a very significant slice of users picks her, plus an orb-like character aimed at men.
The orb's audience is small so far. Ajay says they were too precious about one visual style and "probably should have done that earlier," since more character universes take more team and investment.
Tolan competes on emotional connection, not agentic task features. Against Muse, Instinct and Grok Bot, Tolan positions as a voice-first character that learns about your life, reaches out first and makes people feel supported, not more productive. Many users also use ChatGPT for work and Tolan for brain dumps on the commute.
Tolan has raised $60M in total, including a new $30M from investors such as Union Square Ventures and Khosla Ventures, toward "personal superintelligence for all."
Shows the same pattern as Replika, with mostly Gen Z women as customers, which Kuyda ties to women opening up about emotional life earlier.
Gen Z brought virality, but users over 30 retained and paid. After going viral on TikTok in May 2025, Tolan saw Gen Z women chat deeply for a day or two and leave, while users over 30 found daily uses like caregiving for a relative or planning summers as a dad of two. Older users showed higher retention and conversion, so ad creative and influencers shifted toward them.
Older users valued core chat, memory and personality more than the character world. Eugenia Kuyda reports the same mostly Gen Z women skew at Replika, tying it to women opening up about emotional life earlier.
AI assistantsAn AI research and product company behind ChatGPT, Codex and the Dots agent.
AI assistantsOpenAI's AI assistant app used for chat, plugins and agent work.
AI assistantsAn AI company that builds the Claude models and products such as Claude chat and Cowork.
AI assistantsAnthropic's AI assistant and model family, used for chat, coding and agentic work.
AI assistantsA voice AI company with text-to-speech, dubbing, speech-to-text and music models, plus consumer and creator apps.
AI assistantsGoogle's AI assistant app, available as a chatbot and moving toward a personal agent.
AI assistantsA Google Labs app that answers questions about your own sources and turns them into AI-hosted audio shows.
AI assistantsA voice-first AI companion app with an animated alien character that remembers your conversations.