AI shopping agents have arrived

AI agents are choosing for your clients' customers. Are they choosing your clients?

Meta's Muse, ChatGPT, Perplexity and Gemini now search, compare and pick on the customer's behalf. AIVO Win Rate sends real shopping agents to your clients' customer requests and shows how often they pick your client, and who they pick instead.

Create agency account See an example report

See full sample report →A London seafood restaurant

"Looking for a good fish restaurant in London for 6 people on a Saturday, up to £150 a head."

Found by the agents14 of 15

Picked by the agents1 of 15

ChatGPTFound 5 · Picked 0

GeminiFound 5 · Picked 1 · Beaten by J. Sheekey

PerplexityFound 4 · Picked 0

Visible almost every time. Chosen once. That gap is what Win Rate measures.

The system doesn't just tell you what happened. It tells you why it happened, and how to fix it for your clients.

The shift

The customer no longer scrolls the results. An agent reads them and makes the choice.

Meta launched Muse

On 8 September 2026 Meta released Muse, a personal agent that searches for products, compares them and prepares the purchase for the customer to approve.

Agents shop the whole web

Muse launched with Shopify, Stripe, Walmart, Best Buy, Sephora, Gap and Wayfair. ChatGPT, Perplexity and Gemini already run searches, read pages and recommend.

One pick, not ten links

An agent ends with a choice. If it isn't your client, your client doesn't get the sale, however well they rank.

Why GEO and visibility tools can't tell you

Being mentioned is not being chosen.

GEO and "AI visibility" tools count whether a brand appears in the first answer to a prompt. But agents don't stop at the first answer. They narrow the options, compare them and pick one. A brand can be everywhere at the start and nowhere at the decision.

Win Rate measures the only moment that pays: whether the agent picks your client.

About a thirdof who wins is explained by first-prompt visibility, across 222 brands in our research. Two thirds is decided later in the conversation.

100% known, 0% chosenHertz was recognised in every conversation we ran and chosen in none.

Citations ≠ recommendationsIn our research, how often a brand is cited did not predict whether it was recommended.

How it works

Three steps, per client

Write what the customer asks

Add a client and the requests their customers make: where, when, budget, must-haves. We can suggest briefs from the client's website.

Agents go shopping

ChatGPT, Perplexity, Gemini and Claude each run as a real shopping agent: they search, read pages and make a pick, again and again, because AI answers vary.

See who wins and why

Your client's Win Rate, the rivals who beat them, where they drop out, and a fix list. Retest regularly to see real change, not noise.

Built on research, not guesswork

17 months of research into how AI makes the choice.

Win Rate is built on AIVO's research into how AI assistants recommend and choose: tens of thousands of real decision conversations, studied turn by turn. Our methodology is published openly, so your clients can see exactly how every number is made.

17months of research

27,252AI decision conversations

104,140conversation turns analysed

Our research featured in

FortuneAdAgeBusiness InsiderAmerican Banker

Start with one client today.

Open an agency account, add a client and run a test: see your client's Win Rate, who beats them and what to fix. A typical test (one brief on ChatGPT, Perplexity and Gemini, 5 runs each) costs $65.

$4 a run, pay as you go. Or monthly plans from $3,200 at up to 30% less per run. See pricing

Create agency account → Free to open. You pay only for the tests you run. Already have an account? Sign in