Instinct is a personal AI agent that takes tasks by text message (iMessage, WhatsApp or email) and carries them out on the web: booking an appointment, starting a store return, finding swim lessons. The startup was valued at $10 billion in late September, according to The Verge’s hands-on, even after Meta launched Muse and OpenAI launched Dots, two similar products.
For anyone who sells or advertises online, the number that matters is at the end of the piece: analysts expect these agents, free or cheap today, to end up funded by advertising and product commissions. And the agent reads your site as a robot, not as a person.
What is Instinct and how does it compare to Muse and Dots?
Instinct is an assistant that acts on the user’s behalf, with access to a virtual machine, email and calendar. Muse and Dots do the same job with a big brand behind them. The visible differences are interface and price.
| Instinct | Muse | Dots | |
|---|---|---|---|
| Maker | Startup founded by Noah Shinn | Meta | OpenAI |
| How you talk to it | iMessage, WhatsApp or email, no app | Interface with a bear mascot | Interface with anthropomorphized blobs |
| Price today | Free for invite-only users | Free | Paid subscribers only |
Source: The Verge. Per the report, Instinct reached the market in August, invite-only, with no marketing and barely a website. Shinn worked at Sierra before starting the company. The reviewer was approved off the waitlist within a couple of hours.
These are consumer versions of the agents developers already use, such as Codex and OpenClaw. If you follow the technical side, Cloudflare’s Kitesurf shows the infrastructure that makes it cheap to give every agent a browser.
How does the agent behave on a real site?
Instinct nails admin tasks and stumbles on buried information. The reviewer asked for swim lessons for a preschool-age son. The times and dates the agent returned matched the community center’s site. One detail was missing: the center requires a prerequisite class for that age group. The article sums up the problem: agents are great with clean data, but the internet was made by humans, and sometimes the crucial piece of information is hiding in a pop-up.
When I was in sales, the information that closed or killed a deal was almost never in the brochure. It was in the fine print the customer never saw. AI agents work the same way, except the one missing the fine print is the robot, and the purchase decision can go wrong without anyone noticing.
Who will pay for these agents, and what does that mean for advertisers?
The bill will arrive through advertising and commissions, in the reading of the analysts The Verge spoke to. Avi Greengart of Techsponential says the logic for a startup like this is to grow fast until it becomes worth paying for, and sums it up: if you are not paying directly, you are paying indirectly. Greg Ireland of IDC points out that agents are heavy on transactions and shopping, which makes them a natural fit for ads.
There is one concrete signal: Instinct has already tried “recommended products” on its users. Greengart also floats the hypothesis that stores and services may cut deals for access to the agent of someone who is not paying for the product. That is an analyst’s hypothesis, not a confirmed practice. For anyone buying media, treat it as a scenario to watch, the way the SEO versus paid search debate has moved budgets before.
What should you check on your site on Monday?
Open your highest-converting pages as if you were a robot and check four things.
- Prerequisites, final price and lead time sit in the page text, not inside a pop-up, collapsed tab or image.
- Business data is consistent (hours, address, return policy). The agent reassured the reviewer by pointing to a 2024 confirmation email from the eye doctor; if sources conflict, the recommendation comes out wrong.
- Forms work without tricks: Instinct filled out a store return and an appointment email, so a field that needs odd interaction can stall an agent.
- You show up in AI answers, where the agent will look for candidates. The guide to showing up in ChatGPT and the case of a competitor that won the category in ChatGPT show where to start.
Summary: what to take from the Instinct, Muse and Dots race
- Three consumer agents compete for the same market: Instinct (startup, $10 billion valuation), Muse (Meta, free) and Dots (OpenAI, paid only).
- They handle admin tasks well and fail when the deciding information is hidden on the site.
- The likely business model is advertising and product commissions, per analysts at Techsponential and IDC. None of the three has confirmed that as the final model.
- Immediate action: take critical information out of pop-ups, align business data across all pages, and measure presence in AI answers.