How do free AI chatbots make money? What you pay with, app by app
The Morse teamPublished 9 min read
As of November 2026, free AI chatbots make money in four ways: ads (ChatGPT, Microsoft Copilot, and Google's AI search show them, and Meta uses your Meta AI chats to pick ads on Facebook and Instagram), training on your chats, upselling you to a subscription, and using you as a way into your workplace's budget, with investor money covering what all that doesn't. Most of what free users pay with is attention and data, and the companies say so on their own pages.
We read each company's ad pages, help centers, pricing pages, announcements, and Anthropic's IPO filing on November 18, 2026. Anthropic deserves credit here. Its prospectus, amended November 17, 2026, spells out in plain terms why it gives Claude away and what that costs.
How each free AI chatbot makes money, app by app
Personal accounts in the US, read November 18, 2026. "Trains on free chats by default" is the setting a new free user starts with.
| App | Free tier | Ads | Trains on free chats by default | Other way it earns |
|---|---|---|---|---|
| ChatGPT | Free, plus Go at $8 a month (OpenAI) | Yes, on Free and Go (OpenAI) | Yes, on Free, Plus, and Pro (OpenAI) | Plus at $20 and Pro at $100 a month (OpenAI); workplace sales, over 40% of revenue (OpenAI) |
| Gemini | Free; Google AI Pro is $19.99 a month (Google) | In Google Search's AI Overviews and AI Mode, not the Gemini app (Google) | No, the training settings "are off by default" (Google) | Existing Search ad campaigns carried into AI answers (Google) |
| Meta AI | Free "for everyday use" (Meta) | On Meta's apps, picked partly by your chats (Meta) | Yes, "to improve AI at Meta" (Meta) | Meta One Core at $7.99 and Premium at $19.99 a month (Meta) |
| Microsoft Copilot | Free with a Microsoft account (Microsoft) | Yes, without Microsoft 365 (Microsoft) | No, in the app updated August 18, 2026 (Microsoft) | Microsoft 365 Personal at $9.99 a month, with more Copilot use (Microsoft) |
| Claude | Free, $0 (Anthropic) | No (Anthropic) | "Opt-out" on Free, Pro, and Max (Anthropic); used if you allow it (Anthropic) | Pay-per-use sales, 79% of revenue in the first nine months of 2026 (SEC filing); Pro at $20 a month (Anthropic) |
| Grok | Free, $0 (xAI) | None stated for Grok (xAI) | It "may use your content" unless you turn off "Improve the model" (xAI) | Plans at $8, $30, $100, and $200 a month, sold with X (xAI) |
| Morse | Free, $0, no card (Morse) | None on any plan (Morse) | Morse doesn't train AI on your chats (Morse) | Pro at $19 and Max at $49 a month (Morse) |
Paying doesn't change the training default at ChatGPT, Claude, or Grok. At all three, the line falls between personal and business accounts.
How does ChatGPT make money from free users?
With ads, and by turning free users into paying workplaces later. OpenAI's ads page says ads "may appear for users on the Free and Go plans," that testing began in the US on February 9, 2026, and gives the reason: "Keeping the Free and Go plans fast and reliable requires significant infrastructure and ongoing investment. Ads can help support broader access" (OpenAI). When it announced ads on January 16, 2026, OpenAI pitched them as a way to use ChatGPT "with fewer usage limits or without having to pay," and promised "a paid tier that's ad-free" (OpenAI).
The ads grew fast. OpenAI said "ChatGPT Ads has reached $1 billion in annualized revenue run rate" "in less than 200 days after launch," and called advertising "one pillar of OpenAI's diversified business model, alongside consumer subscriptions, enterprise offerings, and usage-based APIs," with "an advertising-supported free tier" for "more than 1 billion weekly active users" (OpenAI).
In its March 31, 2026 funding post, OpenAI said ChatGPT had "more than 900 million weekly active users, and over 50 million subscribers," and that it was "generating $2B in revenue per month" (OpenAI). That's about one subscriber for every 18 weekly users. The same post explains what the rest are for: "Our consumer scale becomes the front door for enterprise usage, as familiarity in daily life drives adoption at work." Enterprise, it said, is "more than 40% of our revenue."
Which plans see ads, what the ad system reads, and how to switch it off are in does ChatGPT have ads.
Why are AI chatbots free at all?
To get you, and then your employer, to pay. Anthropic says this outright in its IPO prospectus, filed with the SEC and amended November 17, 2026: "we offer free access in significant part to build awareness of, familiarity with, and engagement with our products, and as a means of reaching individuals engaged in knowledge work outside of a business or organizational context." It adds that "we seek to convert these users into paying customers over time," and lists the risk that free users "may choose to use our free offerings indefinitely" (SEC filing). What the same filing says about your Claude data is in what Anthropic's IPO filing says.
The filing also shows that subscriptions are the smaller business. For the nine months to September 30, 2026, 79% of Anthropic's revenue was pay-per-use, from customers "building applications incorporating Claude or utilizing our first-party applications," and 21% came from subscriptions, which cover "both enterprises and prosumers." Anthropic's own explanation of its business, from its February 4, 2026 post, fits in one line: "we generate revenue through enterprise contracts and paid subscriptions" (Anthropic).
Are free AI chatbots losing money?
Anthropic has been, heavily, and its IPO filing puts the numbers in public. It shows 2025 revenue of $4.585 billion against $7.326 billion spent on compute and infrastructure alone, about 1.6 times revenue, for an operating loss of $8.06 billion. For the first nine months of 2026, revenue reached $33.242 billion and the operating loss narrowed to $1.26 billion. The net loss for those nine months was $104.0 billion, but $94.3 billion of that was a non-cash charge for the changing value of its convertible notes and similar instruments (SEC filing).
OpenAI is private and published no profit or loss figure on the pages we read. What it did publish is the money behind it: a funding round with "$122 billion in committed capital at a post money valuation of $852 billion," anchored by Amazon, NVIDIA, and SoftBank (OpenAI).
Do free AI chatbots train on your chats?
Most do, and paying doesn't change the default at ChatGPT, Claude, or Grok: OpenAI says training is "enabled by default" on Free, Plus, and Pro (OpenAI), Claude's pricing page lists it as "Opt-out" on Free, Pro, and Max (Anthropic), and xAI "may use your content" unless you switch it off (xAI), while all three leave business accounts out. Google's Gemini training settings "are off by default" (Google), and Microsoft says the Copilot app updated August 18, 2026 doesn't use prompts and responses "to train foundation models" (Microsoft). The switch for each app is in which AI companies train on your data.
How does Meta AI make money if it's free?
Through the ads you already see on Facebook and Instagram, which Meta has personalized with your Meta AI chats since December 16, 2025 (Meta); is Meta AI private covers what it uses and what it leaves out. The Meta AI app also has a Shopping mode, "a dedicated shopping concierge for you based on your interests and prompt" (Meta).
Meta started charging for more AI on September 15, 2026, with Meta One. It promises that "Meta AI will still be free for everyday use," and sells Core at $7.99 a month and Premium at $19.99 for more image and video generation. Single-app plans start at $2.99, and Meta says that in early testing "more than half of subscribers engaged with both AI and expression features" (Meta).
How do Copilot and Gemini make money?
From subscriptions, and from ads sold through businesses Microsoft and Google already run. Copilot shows ads to people without a Microsoft 365 subscription (Microsoft); how a chat picks an ad, on ChatGPT's example, is in does ChatGPT have ads. The upsell is Microsoft 365 Personal, at $9.99 a month, which lists "Advanced Copilot chat with higher usage than free" (Microsoft).
Google didn't have to build a new ad business for AI. Its Ads Help says every existing Search, Shopping, and Performance Max campaign is "automatically eligible" to show in AI Overviews and AI Mode, so advertisers already paying for Search reach AI answers without setting up "separate budgets" (Google). The subscription side is Google AI Pro, $19.99 a month on Google's plans page (Google).
Do free AI chatbots sell your data?
They say they don't. OpenAI says your "data and conversations are protected and never sold to advertisers" (OpenAI), and Google's Gemini hub says "Your Gemini Apps chats are not being used to show you ads" and "we do not sell your personal information to anyone" (Google). The limited data OpenAI does share with "select marketing partners" to promote its own products, which its privacy policy says some state laws call "targeted advertising" (OpenAI), is covered in does ChatGPT sell your data.
In the data deals that make the news, AI companies are the buyers. Reddit's February 22, 2024 deal lets Google use Reddit's Data API for "providing more efficient ways to train models" (Reddit). Its May 16, 2024 deal with OpenAI brings Reddit content "to ChatGPT and new products," and makes OpenAI "a Reddit advertising partner" (Reddit). Here the free users paying with their words are Reddit's, which does Reddit sell your data to AI covers in full.
How Morse keeps your free chats out of the business model
Every way of earning from a free user above starts with your conversation on the company's servers, tied to your account, where the ad system can read it, the training pipeline can take it, and the sales team can count you. Morse doesn't show ads on any plan, sell your data, or train AI on your chats, and it runs no advertising trackers or third-party analytics. Your chat history is stored in your browser, encrypted, not on Morse's servers, and Morse's servers don't keep or log message content.
The Free plan is $0 with no card: 15 messages a day with the model picked for you, three images a day, web search, and Anonymous and Private chats. The plans page lists what Morse sells: Pro at $19 a month ($17 billed yearly) and Max at $49.
- Private, for everyday questions you don't want feeding anyone's ad profile. It's the default for a new chat. Open-weight models like DeepSeek, Kimi, and GLM run under zero data retention: Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
- Anonymous, to use GPT or Claude without the provider knowing it's you, on Pro with your choice of model. Morse's proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider still sees the message and may keep it for a limited time to check for abuse, under its own policy, so leave names out of what you type.
- E2EE, for anything no one should read, on Pro. Your browser checks the secure enclave's proof that it's genuine, then encrypts the message so only that enclave can read it, and Morse's servers relay text they can't read. The guarantee rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models with no web search or image generation.
Pro also adds uncensored models, which give answers where other apps refuse. In every mode, clearing your browser or losing the device deletes your chats and Morse can't recover them, so save what you'll need. Because Morse doesn't store chats, it usually has very little to give authorities.
Ask your next question on Morse's Free plan, with no card and no ads, and Morse doesn't train AI on what you type.
Common questions
Does paying for ChatGPT Plus stop OpenAI training on my chats?
No. OpenAI's data controls page says training is enabled by default on Free, Plus, and Pro alike. Only business accounts start with it off.
How much money does ChatGPT make from ads?
OpenAI said in 2026 that ChatGPT Ads had reached a $1 billion annualized revenue run rate 'in less than 200 days after launch.' For scale, OpenAI had put its total revenue at $2 billion a month on March 31, 2026.
Why doesn't Claude have ads?
Anthropic said on February 4, 2026 that Claude has no ads and that it earns money through enterprise contracts and paid subscriptions. Its IPO filing says most of its revenue is pay-per-use, metered by how much customers use.
Do free AI chatbots sell my conversations?
The big ones say they don't. OpenAI says your conversations are 'never sold to advertisers,' and Google's Gemini page says 'we do not sell your personal information to anyone.' The data deals you read about mostly run the other way, with AI companies paying sites like Reddit for content.