A standing box for agent shopping volume, dated as of November 2, 2026. Each line carries its own source, because these numbers do not measure the same thing:
- Meta, per Wired, October 31, 2026: more than 10 million people in the US and Canada have activated Muse, and over 25 percent of them have used it for shopping. Mark Zuckerberg gave those figures on Meta's third quarter earnings call, and Wired reports shopping as the most popular way people use the app. Activation and shopping use are measures from Meta's side of the funnel. Neither one means an order was placed.
- DoorDash, per TechCrunch, October 30, 2026: "several hundred" orders in total from third party agents, on DoorDash's third quarter earnings call. Muse is one of the agents DoorDash named. Instinct and Fo are others. This is an all agent total at one merchant, not a Muse only count.
- Hilton, per Wired, October 31, 2026: reservations made through agents were "insignificant."
- eBay, per Wired, October 31, 2026: agent activity was "very, very small."
- Next update to this box: when DoorDash, Meta, or a named merchant publishes a dated order or booking volume in its own name. The update runs even if the next count is still in the hundreds.
Agent Shopping Check, the reader report for this watch: has Muse actually placed an order for you? Not searched, not shortlisted. Placed, with a merchant name and a date. Reports that meet that bar will be counted in the next update to this page. Searches will not be.
The short version
Two statements that both sound true now sit side by side. They do not add up to a conversion rate.
On Meta's side, the funnel is wide. More than 10 million activations in the US and Canada, over a quarter of those people having used Muse for shopping, per the figures Wired took from Meta's third quarter earnings call. On the merchant side, the till is quiet. Wired's piece draws on executives at roughly 50 companies, across travel, internet services, food delivery, and self storage. It is a reporting sweep, not a questionnaire: some answers came on earnings calls, some in conversation, and they do not all arrive in the same format. Company after company used small words: hundreds, insignificant, quite low, nothing significant, close to zero.
Both can be true because they count different events. Activating an app is one event. Using it for shopping, in whatever sense Meta counts that, is a second and looser one. A completed order at a named merchant is a third, and it is the only one a merchant can see. This post keeps those three events separate. The moment they get blended into one rate, a quarter of 10 million starts looking like two and a half million buyers, and no source checked for this post supports that number. Nobody has published it, because on the merchant evidence, it is not what happened.
The DoorDash count, read carefully
DoorDash is the one merchant in this story with a number, so the number deserves a slow read.
TechCrunch reported on October 30 that DoorDash used its third quarter earnings call to put agent ordering in the hundreds. DoorDash CEO Tony Xu, answering an analyst, said: "if you look at where AI agents are today, we're seeing maybe several hundred orders in total from some of these third party agents." TechCrunch named the agents in DoorDash's framing: Muse, Instinct, Fo, and others.
Three limits sit inside that quote, and each one matters.
First, it is a total across agents. Muse's share of several hundred could be most of it or a small slice of it. DoorDash did not split it, and no source checked splits it either. Any headline that turns this into a Muse order count has added a fact that is not in the record.
Second, Xu gave no period for the count in the remarks TechCrunch reported. "Where AI agents are today" is a state, not a date range. Muse launched on September 8, 2026, so the plausible window is weeks, but plausible is not published. This page will not print a per week rate built on a window nobody stated.
Third, DoorDash is one merchant, and a merchant agents reach in a particular way. Set against it, TechCrunch noted DoorDash's own 2025 figure of an average of 3 million deliveries a day. Several hundred agent orders against that daily volume is a rounding error today. Xu also said customers still place orders predominantly through DoorDash's own app, and that DoorDash wants a healthy relationship with third party agent makers. Read the whole answer and it is not a dismissal. It is a big merchant saying the new channel exists, is tiny, and is worth learning about anyway.
Do not quote the DoorDash figure as "Muse placed several hundred orders on DoorDash." That sentence is wrong in two places. The count covers several third party agents combined, and DoorDash never assigned Muse a share of it. The accurate sentence is longer and less exciting: DoorDash has seen several hundred orders in total from third party agents, Muse among them.
What the other companies told Wired
Hilton's word was insignificant, for reservations made through agents. eBay's was very, very small, for agent activity. Other names carried the same register. That is per Wired's reporting as banked in this site's October 31 sweep, and per coverage of the piece the same day. T-Mobile had seen nothing significant from agents. Ally Financial called the activity quite low. Delivery company Eternal, in India, put commerce through agents at close to zero.
One answer in the coverage of Wired's survey measured a fear instead of a sale. Analysts have wondered whether people would set agents loose on subscriptions and memberships they no longer use. Life Time Group said it had seen zero cancellations from agents. Public Storage and Hartford Insurance Group reported similar results, per the same coverage. Zero is a number, and here it cuts against the disruption story in both directions: agents are not yet buying much for people, and they are not yet cancelling much on them either.
Coverage of Wired's piece also reported the response from Meta. Meta declined to comment beyond its earnings call remarks, and pointed to partnerships with 10 retailers, including Walmart, Best Buy, and Gap, meant to simplify shopping through Muse. A partnership is a door being open. It is not traffic through the door, and Meta offered Wired no order count of its own to set against DoorDash's.
Notice what kind of evidence each side brought. Meta brought a count of its own users, which only Meta can see. The merchants brought adjectives, which anyone can say. DoorDash brought the only count from the receiving end, and it is also the vaguest kind of count, an order of magnitude rather than a figure. Until a merchant publishes an exact, dated volume, the receiving end of agent shopping is described better than it is measured.
The timing problem in the survey
There is a fair objection to Wired's picture, and the coverage of Wired's piece contains it. Many of the retailers partnered with Muse last reported earnings before Muse launched in September, so they have not yet had to disclose anything about its effect on their business. The survey catches the first eight weeks or so of a product that launched on September 8. Eight weeks is early. It is also the whole life of the product so far, which is why the counts are worth writing down now, dated, instead of reconstructed later from memory.
The doors agents can walk through are also still being decided. Amazon blocked Muse on September 20, and this site's merchant tracker has followed the refusals and welcomes since (https://museexplained.com/guide/whos-blocking-muse/). A shopper whose agent cannot enter the largest store in the country will place fewer orders through that agent. That is context for the small numbers. It is not a reason to round them up.
And some companies are doing work that only makes sense if they expect the numbers to grow. Chipotle's chief digital officer, Arlie Sisson, described agents struggling to complete purchases when a loyalty discount was involved, or when an order arrived outside business hours, in the coverage of Wired's piece. Chipotle has seen agent orders rise by a single digit percentage since Muse's debut, per that coverage, and is cleaning up its website code so agents can finish orders more reliably. That is the clearest early effect in the whole survey: not sales, preparation.
A note on the 25 percent
One figure in this story needs a warning label of its own. The scoreboard at the top follows Wired: over 25 percent of the people who activated Muse have used it for shopping. That sentence could not be verified verbatim from Wired's own text in the checks run for this post. It rests on summaries of Wired's piece and on the AIBriefs card this site banked on October 31, and other outlets, Marketing Brew and Analytics Insight among them, attach the same figure to shopping.
TechCrunch's earnings coverage breaks the same call down differently. In its account, about 25 percent of people who set up Muse use it daily, and over a quarter of Muse's engaged users who were shopping through the agent had an average purchase size over $200. Daily use and shopping use are different claims, and a quarter in one is not a quarter in the other.
This post does not merge them. If the daily use version is the right one, Meta's shopping share is smaller than the scoreboard line suggests, and the gap between Meta's funnel and the merchants' tills is wider, not narrower. Either way, neither version is a count of completed orders, and the merchant side counts in this post do not depend on which version Meta meant.
Why this matters for the fee
This site covered Meta's money plan in October: no ads inside Muse, and a small fee on transactions Muse completes, paid by the merchant (https://museexplained.com/guide/muse-transaction-fees/). That plan has a missing half, and this survey is it. Meta has never published the fee percentage. Now it turns out the other half of the multiplication, the transaction volume, is also unpublished by Meta and, at the one merchant that counted, sits in the hundreds across all agents.
So no honest revenue estimate exists yet. Anyone who multiplies a guessed fee by a guessed volume is guessing twice. What can be said plainly is narrower: at several hundred orders on a platform that handles millions of deliveries a day, transaction fees on agent orders are not a business yet, for DoorDash or for Meta. The fee story stays what it was in October, a keynote promise waiting for two numbers. This survey supplied a rough size for one of them, and the size is small.
What would count as a harder number
This page will move when any of these appear, in a named source, with a date on it:
- A Muse only order count, from Meta or from a merchant, that says which agents it covers and which it does not.
- A period. Orders since a stated date, or orders in a stated quarter, so a rate can be checked instead of vibes.
- A merchant that is not DoorDash. One merchant's till is a datapoint. Two, in different categories, would be the start of a pattern that can be checked.
- A booking count from a travel company to set beside Hilton's adjective. Insignificant is a position. A number is evidence.
Until then, the scoreboard at the top of this page stands as written, and the next update runs even if the new count is still in the hundreds.
The Agent Shopping Check at the top of this page holds the same line when answers come in: a completed order you can name counts, a shopping search does not.
Quick answers
How many orders has Muse placed on DoorDash?
Nobody has published a Muse only figure. DoorDash CEO Tony Xu said on the company's third quarter earnings call that DoorDash had seen "maybe several hundred orders in total from some of these third party agents," per TechCrunch on October 30, 2026. Muse was one of several agents named, alongside Instinct and Fo.
What did Wired's survey find?
Wired, on October 31, 2026, drew on executives at roughly 50 companies about Muse and other shopping agents, from earnings calls and its own conversations. The answers were small across sectors: Hilton called agent reservations insignificant, eBay called agent activity very, very small, T-Mobile had seen nothing significant, and Ally Financial called activity quite low, as banked in this site's sweep of the piece and same day coverage of it.
How many people use Muse for shopping?
Per Wired, Mark Zuckerberg said on Meta's third quarter earnings call that more than 10 million people in the US and Canada had activated Muse. Over 25 percent of them had used it for shopping, per that account, and shopping was the most popular way people use the app. Meta has not published what counts as using Muse for shopping. That figure is not a count of completed orders. See the note on the 25 percent above: TechCrunch's earnings coverage attaches about 25 percent to daily use instead, and that conflict is unresolved in the sources checked.
Does several hundred orders mean agent shopping has failed?
No. It means completed orders, at the one merchant that gave a count, are tiny in the first weeks since the September 8, 2026 launch, against DoorDash's own average of 3 million deliveries a day in its 2025 reporting. Several companies in Wired's survey are preparing for agents anyway, and many partner retailers have not yet reported earnings covering the launch period at all.
Why can the 25 percent and the several hundred both be right?
Because they count different events at different ends of the same funnel. Activation and shopping use are counted by Meta, on its side. Completed orders are counted by merchants, on theirs. This post does not blend them into a conversion rate, because neither source publishes the definitions that a rate would need.
When will there be a better count?
When a merchant or Meta publishes one: dated, named, and clear about which agents it covers. The next earnings round from Muse's retail partners is the obvious place, since many of them last reported before the September launch. This page updates when one lands, even if the count is still small.
Frequently Asked Questions
Q: Is the DoorDash count about Muse specifically?
A: No. It is a total for third party agents as a group. TechCrunch, reporting the earnings call on October 30, 2026, listed Muse, Instinct, and Fo among them, and DoorDash published no split by agent. Treating the total as a Muse figure overstates what DoorDash said.
Q: Did Meta dispute the companies' answers?
A: Not in the reporting on Wired's piece. Meta declined to comment beyond its earnings call remarks and pointed to partnerships with 10 retailers, including Walmart, Best Buy, and Gap. Meta has published no completed order count of its own in any source checked for this post.
Q: Are companies ignoring Muse because the orders are small?
A: The survey points the other way. Chipotle is cleaning up its website code so agents can complete orders more reliably, after finding agents stumbling on loyalty discounts and orders placed outside business hours, per coverage of Wired's piece. Wired's broader finding was that most executives it spoke with expected to benefit if agent use grows. Small numbers, real preparation.
Q: What about agents cancelling subscriptions for people?
A: That came up in the same coverage. Life Time Group said it had seen zero cancellations from agents, and Public Storage and Hartford Insurance Group reported similar results. The feared wave of agent driven cancellations has not shown up in these answers either.
Q: How does this connect to Muse's transaction fee?
A: Directly. Meta's stated plan is a small fee on transactions Muse completes, and Meta has never published the percentage. Volume is the other half of that math, and this survey is the first merchant side look at it. Until both numbers exist in public, fee revenue cannot be estimated honestly, and this site will not estimate it.
Q: What should readers report back?
A: One thing: a completed order. If Muse placed an order for you, the useful report is the merchant, the date, and that it completed. Searches, comparisons, and shortlists are shopping use, and Meta may count them that way, but they are not orders, and this scoreboard only moves on orders.
Sources
- Wired, "Meta's Muse Could Upend All Kinds of Businesses. It Hasn't Yet," by Paresh Dave, October 31, 2026 (6:30 AM, Business section, per the Wired page header): executives at roughly 50 companies, drawn from earnings calls and conversations over the past several weeks; Hilton reservations through agents "insignificant"; eBay activity "very, very small"; Ally Financial "quite low"; Zuckerberg's figures as reported there, from Meta's earnings call, more than 10 million people in the US and Canada having activated Muse, with over 25 percent having used it for shopping. The roughly 50 company count and the exact 25 percent shopping sentence rest on summaries of the piece and the AIBriefs card, not on wording verified verbatim from wired.com, and the next section flags a conflicting breakdown of the 25 percent figure in other earnings coverage.
- TechCrunch, "DoorDash said it's only received 'several hundred' orders from AI agents," October 30, 2026: DoorDash third quarter earnings call; CEO Tony Xu, "maybe several hundred orders in total from some of these third party agents," with Muse, Instinct, and Fo named; DoorDash's 2025 average of 3 million deliveries a day as the comparison figure.
- TechCrunch, "Meta says Muse has over 10M users, with shopping a top activity," October 21, 2026, read via Yahoo Finance syndication: a different breakdown of the same earnings call, with about 25 percent of people who set up Muse using it daily, and over a quarter of engaged shoppers averaging over $200 a purchase. Marketing Brew (October 23, 2026) and Analytics Insight (October 22, 2026) attach the over 25 percent figure to shopping use, matching the summaries of Wired.
- This site's demand log sweep, October 31, 2026, banking the AIBriefs card on the Wired piece (1 source: Wired), and the editorial checks run November 1, 2026.
- All Things Muse, "WIRED: Companies Say Muse Hasn't Upended Their Businesses Yet," October 31, 2026, a same day summary of the Wired piece: T-Mobile, Ally Financial, Life Time Group, Public Storage, Hartford, Eternal, the Chipotle preparation detail, and Meta declining to comment beyond its earnings remarks while pointing to 10 retail partnerships. DuckDuckGo snippet checks against wired.com on November 1, 2026 confirmed most of these in Wired's own text, including the Hilton, eBay, Ally, Life Time, Chipotle, and Meta response items; the survey size wording and the exact 25 percent shopping sentence could not be verified verbatim from wired.com.
- Muse Explained, "How Muse will actually make money: the transaction fee plan," October 2026, and "Who's Blocking Muse: The Merchant Refusal Tracker," October 2026: the fee promise with no published percentage, and the Amazon block of September 20, 2026.
Comments