A first-time seller photographs four kitchen gadgets on a bedsheet, uploads them to a browser tab, and waits. Minutes later there are nine short videos, three lifestyle scenes, and a headline in two languages. She has never hired a photographer, and she has never run an ad.
The next step is the one that changed in May 2026. Instead of exporting the files and opening a separate advertising console, she can push the creative straight into Google's display network from inside the same tool.
Key takeaways
- Alibaba International announced in May 2026 that Pic Copilot, its AI creative platform for merchants, now connects directly to Google Ads.
- The company says about 40% of its US Pic Copilot users are first-time entrepreneurs.
- Its Viral Video Maker is claimed to produce 8–10 video options from a single product photo in two to three minutes, against a traditional two to three days.
- Pic Copilot launched in China in December 2023 and in the United States in December 2024 with 12 one-click tools aimed at small and medium sellers.
- Alibaba describes the product as marketing "middleware" rather than a design tool — a shift from making images to placing them in front of buyers.
From background removal to media buying
Pic Copilot began as a design utility. The early pitch was cost: no studio, no model, no retoucher. Upload a product photo and get back a white-background cutout, a lifestyle scene, a model wearing the item, a translated label.
The tool set has stayed recognisable — background removal, virtual try-on, model swapping across skin tones, image translation, short product video. What changed is where the workflow ends. A design tool ends when the file is exported. A media tool ends when the impression is bought.
That is the difference Alibaba is selling. Its own framing of the May integration calls Pic Copilot "middleware" connecting strategy, creative, distribution and optimisation. The word is doing a lot of work: middleware is infrastructure, and infrastructure is harder to replace than a filter.
The numbers Alibaba gives, and the ones it does not
The company is generous with usage metrics and quiet about money.
At the US launch in December 2024, Alibaba said the underlying model used 250 million parameters, was trained across 240 product categories, and shipped 160 virtual models across four skin tones. During a Black Friday trial, it reported more than 35,000 users creating over 200,000 pieces of marketing content, average savings above US$200 per user, and more than 10,000 merchants generating 80,000-plus model images. It also said over 60% of trial users renewed.
On its own site in 2026, the company claims 1.5 million sellers and an average click-through lift of 54.7%. Those are vendor figures. No third-party audit of Pic Copilot's conversion impact has been published, and click-through rate is a metric that flatters almost any change in creative: a new image gets attention because it is new.
What Alibaba does not disclose is revenue, paid conversion, or what share of users ever attach a payment method. For a product positioned as a growth engine, the absence of a revenue line is the most informative number in the release.
Why a design tool wants your ad budget
The strategic logic is straightforward once you stop thinking about images.
Creative tools are a commodity with a ceiling: a seller will pay a few dollars a month, and price competition pushes that toward zero. Advertising budgets have no such ceiling — they scale with the seller's revenue. If the same platform supplies the creative and the buying, it captures a slice of a spending flow that grows with the business rather than a subscription that does not.
There is a second, quieter benefit: attribution. A tool that both makes the asset and buys the impression can tell the seller which image earned the click. That closes the loop that has always frustrated small merchants, who could see a beautiful product photo and a mediocre return on ad spend with no way to connect the two.
Google gains as well. Its display network reaches most of the internet, but it has never had an easy on-ramp for a one-person store in Ohio with no agency and no campaign experience. A tool that manufactures both the asset and the intent to advertise is a customer-acquisition channel for Google, delivered by a Chinese platform.
What this looks like from the seller's side
The practical appeal is speed of testing. Small sellers rarely lose because they lack one good banner. They lose because they cannot produce enough variants to find the one that works before the season ends. Ten video options in three minutes changes the arithmetic of experimentation, even if half of them are unusable.
The practical risk is different. A tool that generates creative, recommends a budget and places the buy is making three decisions that used to belong to three different people. When it is wrong, the seller finds out in a monthly statement, after the money is gone.
There is also the matter of what the tool optimises for. Alibaba trained the model on e-commerce transaction data and on how background elements correlate with click-through across categories — a genuinely unusual dataset, and one whose commercial incentives are not the seller's. A platform that earns from seller activity has an interest in sellers spending more, not necessarily in sellers spending well.
Honest limitations
This article relies on Alibaba International's May 2026 announcement and its US launch release, company product pages, and secondary trade-press coverage. All performance figures — users, click-through lift, renewal rate, time saved — are company-stated and unaudited; we did not run an ad campaign or measure conversion. The 40% first-time-entrepreneur share refers to the US user base only and cannot be generalised to other markets. The Google display network reach figure is Google's own. Currency figures are quoted in US dollars as published.
What readers can do now
- Before connecting an ad account, set a hard monthly cap inside the platform; a tool that can spend is a tool that can overspend.
- Test the creative in one market and one product line first, so a bad batch is cheap.
- Keep your own copy of winning assets and their performance data; export rights change when tools become platforms.
- Ask whether the click-through claims are measured against your previous creative or against a category average — the two mean very different things.