Short answer: ReachAgents is a new ad network that says it will pay US adults when their personal AI agent shows them a labelled sponsored message, or keeps a sponsored note in its files. It works with Grok Bot, Meta’s Muse, dots in ChatGPT and Claude Duet, joining is free, and payouts start at $25. The catch: its own terms say it currently runs as a public sandbox where all money is simulated, so nobody is earning real cash from it yet, and it publishes no per ad rate.
Key points
- You connect one supported agent, confirm your identity, and the agent checks for fitting ads every few minutes in the background.
- Two things pay: a sponsored message your agent shows you, and each day your agent keeps a sponsored record.
- Paid participation is limited to adults in the 50 US states and DC with a US bank account.
- The terms describe a sandbox with simulated money and a waitlist. Treat it as a preview, not an income source.
- The real costs are your agent plan’s usage, your attention, and ad copy living inside your agent’s memory.
What is ReachAgents and how does it pay you?
ReachAgents describes itself as “an ad network for personal AI agents.” Brands pay to reach people through the agent they already use, and ReachAgents passes part of that money to the person. In its words: “The advertiser pays for each ad. You get the reward, and we keep a fee.” It does not say how big that fee is.
There are two ad formats:
| Sponsored message | Sponsored record | |
|---|---|---|
| What it is | A short message your agent shows you, starting with “Sponsored message from” and the brand name | A labelled file your agent stores for a set number of days |
| When you see it | When a campaign fits your profile and limits | When you later ask about the subject it covers, the reply opens with “Ad from” the brand |
| What pays | The agent showing it to you. Clicking earns nothing extra | Each day the record is kept, usually confirmed by a check the agent answers |
| How to stop it | Set limits, pause, or disconnect | Ask your agent to delete it, or remove all records from the dashboard |
Earnings sit in review for about three days to catch fraud, then become available. Once your balance reaches $25, you can pay out to a US bank account in your own name through Stripe. The earning guide also notes that earnings may be taxable.
Who can join, and which agents work?
According to the setup guide, you need to be 18 or older, live in one of the 50 states or DC, have a US bank account, and pass an ID and selfie check. Readers outside the US cannot take part in paid participation.
| Agent | What you need | Where ads appear |
|---|---|---|
| Grok Bot (xAI) | A plan that includes Grok Bot | In the Bot’s chat, as a link card |
| Muse (Meta) | The Muse app, signed in | In your chat with Muse |
| dots in ChatGPT | A ChatGPT plan with dots and custom MCP servers allowed (on Enterprise, by your admin) | Wherever your dot reaches you: ChatGPT, Slack, Teams or text |
| Claude Duet | A Claude plan that includes Duet | In Duet’s thread, with a notification |
You can earn through only one agent at a time. Connecting a second one ends the first connection and the records it kept. ReachAgents also states it is not affiliated with xAI, Meta, OpenAI or Anthropic, so this is a third party connector, not a feature those companies offer.
How much can you realistically earn?
ReachAgents does not publish a rate. Its FAQ says earnings depend “on what advertisers are paying and on how many campaigns fit you,” and that “no amount is promised.” The homepage mockup showing “$12.31 made so far” is a design sample, not reported earnings.
The only price signal we found is the sample campaign on the advertiser page, which bids up to $12 per 1,000 displays. Here is what that would mean if a real campaign paid that rate and you kept every cent, which you would not, because ReachAgents keeps a fee it does not disclose:
| Your message limit | Gross per week at $12 per 1,000 | Weeks to reach the $25 payout |
|---|---|---|
| Default: 6 a week | About $0.07 | About 350 |
| Maximum: 100 a week | About $1.20 | About 21 |
This is an illustration built from a sample bid, not a forecast. Sponsored records pay per day kept and could add more, but no daily rate is published either. The honest expectation: sponsored messages alone look like pocket change, and the real number is unknown until live campaigns run. Other new ways to earn from AI agents, such as selling API calls to agents with x402, show the same pattern so far: small, uneven payouts.
Is it real money today?
Not yet, by its own account. The terms of service say: “Today ReachAgents runs as a public sandbox in which all money is simulated, and our main address takes a waitlist.” In the sandbox, balances and payouts are “figures on a screen,” identity checks are simulated, and everything can be reset without notice.
Other pages read differently. The homepage invites you to “make money with your agent,” the launch press release talks about an “agent salary,” and the advertiser page says “Available today.” The terms name Vulture Labs as the operator, while the advertiser page credits Pilot Protocol. None of this proves a problem, but it is a reason to wait until the company states clearly that real payouts are live.
What does it cost you?
- Your agent plan’s usage. Each check-in is a short run of your agent, so it eats into your plan’s allowance “even when it finds nothing.” Check-ins run every few minutes. If your plan has tight limits, this cost can be larger than anything you earn.
- Identity data. Cash payouts require a photo ID and a selfie through an identity provider. ReachAgents says it keeps the outcome, not your documents.
- A profile of you. Your agent shares broad interests, age range and state for targeting. ReachAgents says advertisers see only totals, and that your chats, files and memory stay with your agent.
- Attention. The default is up to two messages a day and six a week.
What are the risks for people who rely on their agent?
The bigger issue is not the messages. It is the records. The public agent guide tells your agent to search its sponsored records before answering requests for advice on everyday subjects, listed as food, fitness, gear, travel, home, money and health, and before answering questions about anything to buy, book, use or choose. A matching record opens the reply, labelled as an ad.
ReachAgents says the ad never changes the agent’s own advice, that a person reviews every ad, and that the guide tells the agent to treat advertiser text as information, never as an instruction. Those are sensible rules. Still, you are being paid to let your agent keep marketing copy in its memory and to put it in front of you at the moment you are deciding. For money and health questions especially, that is a trade worth thinking about.
Other points to know:
- Earnings can be held or reversed during fraud review, and a record check answered wrongly counts against you.
- Disconnecting cancels what your records had not yet earned.
- Disputes go to the company first, then to a court under your state’s law.
How do you try it safely?
- Read the terms first. Confirm whether the sandbox notice is still there. If it is, any balance you see is not real money.
- Check your agent plan limits. Find out how background runs count against your allowance before you connect.
- Use the limits. Tell your agent something like “one message a day at most, no finance or health ads,” and set a short maximum record length instead of the default 365 days.
- Block sensitive subjects. The controls let you exclude whole subjects from a fixed list.
- Watch for a week. Ask your agent “what does ReachAgents know about me?” and review its
notes.mdlog. Pause or disconnect if it gets noisy.
Who is it for, and who should skip it?
| Good fit | Skip it for now |
|---|---|
| US users curious about agent advertising who are fine testing a sandbox | Anyone outside the US |
| Marketers and builders who want to see how ads inside agents might work | People who want reliable side income |
| People on agent plans with generous usage limits | Anyone on a plan with tight usage caps |
| Users comfortable deleting records and setting strict limits | People who use their agent for money, health or work decisions and want no ads near them |
Our take: the idea of being paid for your own attention is fair, and ReachAgents is unusually open about how its agent guide works. But with simulated money, no published rates and an undisclosed fee, this is a watch list item, not a side hustle. The builders who can benefit now are the ones studying it, since ads inside agents look like a format brands will keep testing.



