E-commerce
August 19, 2026
Daydream is a fashion discovery map, not a purchase rail. The directory promises 10,000 boutiques, natural language search, and a single click to the retailer's site. No gross merchandise value (GMV), no accuracy rates, no UPCs. Nor does this article. Daydream, as contemporary texts describe it, sends the buyer to the merchant and takes a commission on the sale. In 2026, a second product installs the same engine on your site, over your assortment. Two business models. One logo. The 10,000 brands is a publisher-provided figure, not an audit.
TechCrunch covered Daydream's public beta on June 25, 2025: fashion chatbot, public beta, registration (name, date of birth, budget, brands). No integrated checkout funnel: click-through to the merchant site, commission on the sale. More than 8,000 brands at launch, free merchant registration. Daydream announced Powered by Daydream on July 29, 2026: Powered by Daydream, pilot with STAUD, Alice + Olivia, Couper, Cult Mia, Hampden Clothing, plus over 25 signings (Anine Bing, Mansur Gavriel, ba&sh, Sandro, Maje, MESHKI, Ramy Brook). The company claims 1.5 million shoppers, 325 retailers, over 10,000 brands, 3 million products, exiting beta. Figures from a press release. Not an independent census.
If the concept of a software shopper is new: what agentic commerce is. The 2026 timeline: what has been delivered. Other files: AI shopping agents. Official totals remain human shopping journeys: e-commerce statistics 2026.
Summary
Two Daydream products, one 10,000+ capsule
Two Daydream products. Many articles only show one.
Product | Where it happens | Who gets paid |
|---|---|---|
Consumer app / chat | daydream.ing, a multi-brand graph | The merchant, after a click. Daydream takes a commission (TechCrunch) |
Powered by Daydream | The search bar on your site, your catalog | You. The customer does not leave, says the press release |
We sometimes hear that a brand "cannot buy" its way into discovery. This is the 2025 narrative of the affiliate app: free registration, visibility based on data, not a Google Shopping auction. In 2026, Powered by Daydream is precisely a B2B product you sign up for on your own site. Do not read "you cannot buy your way in" as "Daydream has no merchant offer."
This is not "all merchants" fashion in the Operator sense. It is a fashion index. Zara, Levi's, Gucci, Nordstrom mentioned elsewhere are examples of a directory. The sourced names for the 2026 pilot are a different list. Do not merge them in a partner slide.
A non-fashion merchant who "chooses Daydream" like a payment provider has read the 10,000 stores headline, not the catalog. Julie Bornstein and the press release repeat: exclusively fashion. Your tooling SKU will not fit in.
The pills (search, price, wishlist, no automatic purchase) are consistent with an outbound click. They are not Nordstrom coverage audits.
Convert over 2,000 customers on average per month with Qstomy.
The world’s 1st Shopify AI dedicated to customer conversion
Empowering 200+ e-commerce merchants
It is neither UCP nor ACP nor a payment tunnel
UCP and ACP are checkout rails. Daydream, in June 2025, does not have an integrated checkout tunnel. In July 2026, Powered by Daydream returns results from the site's assortment, not an ACP session. Do not paste a protocol integration. The rails: how protocols differ. ChatGPT has ACP. Gemini has UCP. Copilot has MMC and, in the US, Copilot Checkout. Daydream has a fashion graph and, for some, an on-site search engine.
The three platform shifts (readable catalog, bots, delegated checkout) apply differently. Here, the readable catalog is mostly style attributes (cut, drape, silhouette, occasion) that the CTO describes to TechCrunch, plus a feed that Daydream can ingest. The application firewall matters if their crawler visits. Delegated checkout does not: the click opens your checkout tunnel. Framework: what platforms need to change.
MCP is not a Daydream rail either. Nothing in TechCrunch or in the press release discloses a tool server that you plug in. If an integrator sells "Daydream MCP", they are selling a home-grown layer. Ask for the contract, not the word agent.
Vocabulary (discovery, affiliate, merchant of record): agentic commerce glossary. Here, the merchant of record is you, as soon as the cart is yours.
A Daydream buyer landing on a broken theme abandons on your site, not on Daydream. The storefront still matters: a Shopify storefront readable by an LLM.
The 2025 route: long prompt, click, commission
TechCrunch describes the 2025 roadmap: long query or image plus text, collections, "Say More" button to change color or style, style passport, daily inspiration. The pattern is the same: describe, refine, click to buy from the merchant.
Language: Bornstein: fashion search has trained people to look for "a red dress." ChatGPT has expanded the prompt (occasion, need, detail). Daydream is betting on that.
Attributes: Belousova: embroidery, silhouette, who would wear the dress (bride or guest), image recognition. This is not keyword matching.
Output: no Daydream checkout. Redirection. Commission on sales. You deliver, you refund, you handle 3-D Secure.
The directory search filters are the right bucket: agents for product search, not agents for the checkout tunnel in a mandate sense.
If the price or stock in the Daydream index differs from your product page, the human clicks, then sees something else. The order, if placed, is yours. Accuracy: where agentic commerce still breaks.
2025 roadmap cited by TechCrunch (explicit feedback, matching with an already owned item, social sharing): product intentions, not delivered features measured here. We do not count them as delivered.
An affiliate redirection is also a new session for you: often without the Daydream cookie, sometimes with a utm parameter. If your anti-fraud system sees a "new" buyer who already knows their size, calibrate. This is not fraud. It is a graph that has already filtered. The pixel sees an entry. It does not see the prompt "wedding guest, Sicily, no heels."
Powered by Daydream: the engine in your home
Powered by Daydream, July 29, 2026, changes the merchant contract. The engine runs on the brand's site. The results come from its assortment. The press release says it retains customers, data, and site experience. It's the opposite of affiliate marketing: no more commission on an outbound click, but a search you host.
Pilot named. Signatures named. It's not 10,000 integrated B2B boutiques. It's a handful in production plus a queue. A marketing director who reads "10,000+" on an agent sheet and orders Powered by to "be among the 10,000" is confusing the consumer index with the on-premise software.
Lisa Yamner: partners wanted conversational search on their own sites; Daydream "handles the technical side." This is a sales pitch. Ask for the service level agreement, the data processor, the fallback if the chat is unavailable. The press release doesn't provide these.
STAUD (Alicia Carbone) values seeing how preferences evolve in conversation, a context missing from a classic search. Useful. This is not a conversion rate. A pilot case is not your category.
"Easy" integration with existing catalog and infrastructure: a launch phrase. Your keyword taxonomy, as Bornstein says, doesn't talk like the customer. Powered by doesn't fix an empty PIM. It better interprets an already rich catalog: what a machine needs to compare.
Keeping the customer on the site also means keeping the dispute, the cookie, the loyalty program. Powered by does not outsource after-sales service. It changes the search bar. If your internal search is already an in-house vector engine, the pilot is a product comparison, not a strategic requirement. If your search is a broken color filter, the pilot might be worth the time. Measure both using the same references.
Being in the index is not signing the software
To appear in the consumer app, TechCrunch says registration is free upon opening. "Free" does not mean "you control the ranking". Daydream chooses what rises to the top. It is not an auction.
You still need a catalog that their engine can read: images, silhouette, material, occasion, price, stock. A header visual without attributes is invisible in a prompt like "wedding guest dress Sicily". The work is merchandising plus feed, not media.
The incoming traffic is an affiliate referrer. Label it. Otherwise, the SEO and Daydream debate will have no data. The pixel sees a visit. The prompt, however, was in their chat.
Do not build a "Daydream UCP connector". In these two texts, they do not document an open protocol API. The 2025 deliverable: be in the index. The 2026 deliverable: an on-site engine. Two tickets.
If you are already in the index and you sign up for "Powered by", you have two Daydream surfaces: one that sends you traffic, one that searches your stock. Two possible managers (acquisition and on-site search). A single "Daydream owner" mixes them up.
An entry in the application is not a guarantee of freshness. If Daydream re-indexes late, the buyer sees a price that your product page has already changed. You will pay for customer service. They will take, according to the 2025 model, a commission on the sale that actually goes through. Align the stock before celebrating indexation.
Brands, boutiques, buyers: publisher totals
The totals in the press release (1.5 million buyers, 325 banners, 10,000+ brands, 3 million products) and the 8,000+ from June 2025 are company statements. The gap from 8,000 to 10,000 in one year is possible. It is not audited here. Moreover, the "10,000+ shops" label confuses brands and shops: the press release says brands and banners, two separate accounts.
50 million seed funding, Forerunner, Index, GV, True Ventures: the press release repeats this. It is not an income statement. This is not your addressable market.
"First AI-powered search built exclusively for fashion" is an About page tagline. Other fashion tools exist, TechCrunch cites some (Deft, Cherry). We are not declaring a winner.
The marketing director evaluating "where discovery will go" has a more down-to-earth question: is our visual catalog describable without the photo shoot? If the answer is no, neither Daydream nor Gemini will extract silk from a blurry JPEG. Start with the fields. The directory can wait.
Neither text provides a click-through rate to your site, a conversion rate after redirection, a return rate, nor a country map beyond the implicit US launch of a New York / San Francisco startup. Some listings are labeled US and global. Global is not sourced here. We are not inventing a coverage map.
Daydream is not Copilot, not Gemini, not ChatGPT
Four discoveries, four contracts.
Surface | What it indexes | Where we pay |
|---|---|---|
Daydream App | A multi-brand fashion graph | At your place, after affiliate click |
Powered by Daydream | Your assortment | At your place, without leaving |
Copilot / Bing | MMC, potentially Copilot Checkout in the United States | Click or inside the Microsoft chat |
Gemini / ChatGPT | The UCP or ACP depending on the integration | Depending on the rail, not Daydream |
A "fashion AI" pitch can sell the index, Powered by, a theme, or a directory listing. Four deliverables. The directory is the fourth.
Daydream is not a browser agent. It does not bypass your application firewall like Claude. It sends a human, or it searches your HTML via their module. Two technical paths.
Copilot Checkout collects payment in the chat, for those eligible in the United States. Daydream, in the two cited texts, does not collect payment. Gemini via the UCP can, if you have plugged it in, keep a session on the Google side. Lumping these three surfaces into an "AI purchase" metric erases the only thing that matters to finance: where the payment takes place, and who is the merchant of record for the transaction.
The app's wishlist is not a shopping cart. An item pinned in Daydream has not reserved your stock. Treat it as a cold intent, not as a pre-order. If you build an "abandoned Daydream" retargeting campaign, you are retargeting a graph you cannot see.
Three steps, not a slide, 10,000 stores
This week, three actions. Not a slide with 10,000 shops.
Name the business: be in the consumer index, install "Powered by", or both? Acquisition and on-site search.
Fashion attributes: silhouette, material, occasion, size, price. A long prompt doesn't save an empty PIM.
Measurement: Daydream affiliate tag. Price parity between index and product page. Do not budget a cost-per-acquisition without attributed orders.
The single test: would this work still be useful if Daydream shut down the app tomorrow and only kept "Powered by"? A descriptive catalog, yes. A makeshift solution "for their chat window", no.
If you only have one hour: open five main references, note if a human can describe them like a wedding guest would. If not, Daydream won't find them any better than your internal search.
Name what you will not do: no media budget based on 1.5 million buyers, no "Daydream UCP" project, no merging of index and "Powered by" into a single indicator. The directory doesn't need your "not-to-do" list. Your committee does.
You do not turn on Daydream like a channel
You don't turn Daydream on like a Shopify Agentic channel. For the app: you are in the index, or you aren't. For Powered by: you sign a pilot, or you don't. Buyers then come according to their graph, not according to the directory.
If a vendor promises to "be in Daydream like in ChatGPT," ask: affiliate signup, on-site widget, or directory listing. Three deliverables. One word.
The owner is not "growth AI fashion." It is merchandising (attributes) plus acquisition (affiliate) plus, if applicable, on-site search. A single owner stops at the first incomplete PIM.
Bornstein is right on one narrow point: keyword fashion search is a poor interface for a wedding guest. She is wrong if we infer that Daydream replaces Google, Copilot, and the UCP rail. It is just one more graph. You can be absent from it and still sell.
After the click, it's your checkout funnel. 3-D Secure, returns, customer service. Daydream is no longer in the funnel. Don't attribute a dispute to them that they didn't collect.
One more graph, not the end of Google
Fashion discovery is fragmenting. Google is not "dead" because a slide says so. Daydream is just another graph, affiliated, exclusively fashion. Powered by is just another on-site search. Neither replaces a clean feed or a UCP or ACP rail if you need them elsewhere.
Three starting questions: are you in their index, do you want their search on your site, is your catalog describable in guest language? This is not a ranking.
The marketing director evaluating "where discovery will go" has a second, more operational question: who tags the referrer, who maintains the silhouette PIM, who answers the buyer whose price changed between the chat and the product page. If these three roles do not have names, Daydream is not a project. It is a logo in a presentation.
You can ignore the app and still test Powered by. You can be in the index and refuse the software. You can refuse both and still be sellable on Google, Copilot, and your own site. The directory does not decide. Your attributes do.
Frequently Asked Questions
Does Daydream process payments?
Not in June 2025: redirection, commission on sales. Powered by 2026 searches within your site; the payment funnel remains yours. Neither the UCP nor the ACP in these texts.
Are the 10,000 stores an addressable market?
The press release says more than 10,000 brands and 325 stores. TechCrunch said more than 8,000 brands at launch. Publisher figures. Not your addressable market. Not 10,000 Powered by contracts.
Can you buy your visibility?
The 2025 application: free registration, no described bidding. Powered by 2026: a B2B program you sign up for for your own search. Two answers. One word "Daydream".
Do you need a special feed?
You need style attributes and real stock. Two files that lie remain two prices. The index will take its own, your product page the other.
Is Daydream Copilot or ChatGPT?
No. It's a fashion graph plus, potentially, an on-site search engine. Other integrations, other tracks.
What to do this week?
Separate index and Powered by. Enrich the fashion PIM. Label the affiliate. Do not budget a channel for 1.5 million buyers. Postpone the protocol.
Should Daydream be blocked at the application firewall?
The application sends a human on click. Powered by runs on your end. A crawler, if it exists, is a separate robots policy. Do not merge the three into an "AI fashion" allowlist. Test what actually hits your logs.
Sources
Enzo
August 19, 2026
