ChatGPT Adds Global Virtual Try-On and Shopping Favorites
OpenAI on Thursday, 1 October 2026, launched ChatGPT virtual try-on for clothing and accessories worldwide, plus a Favorites library for saving products, TechCrunch reported.
PromptCrates Editorial
Staff Writer
OpenAI on Thursday, 1 October 2026, announced the global launch of two ChatGPT shopping features: virtual try-on for clothing and accessories, and a Favorites library for saving products, according to TechCrunch. The company says both updates lean on the newly launched ChatGPT Images 2.5 model, which OpenAI claims produces more natural lighting and richer textures, follows editing instructions more reliably, and reduces image-generation latency. The Try On control appears in ChatGPT shopping results, while Favorites stores discovered products—and try-on images—inside an in-app Library for later reference.
How virtual try-on works in ChatGPT
Virtual try-on starts with a user-uploaded selfie or full-body photo. Shoppers can then visualize how a garment or accessory might look on them, or upload an image of an item such as a web screenshot and ask ChatGPT to try that piece on. TechCrunch reports that OpenAI is also pitching adjacent shopping behaviors: describing a desired style and asking the assistant to gather the pieces that complete the look, or uploading photos of celebrity outfits and requesting purchasable matches. Those flows push ChatGPT deeper into fashion discovery territory long associated with Pinterest and Google, rather than limiting the assistant to text product Q&A.
The feature set arrives after OpenAI had to pivot away from an earlier instant-checkout experiment that underperformed. More recently, TechCrunch notes, agentic AI startup Instinct drew criticism when proactive product pushes felt closer to ads than helpful suggestions. OpenAI’s framing for try-on and Favorites emphasizes user-initiated visualization and save-for-later organization instead of automatic checkouts. That is a narrower commerce bet, but it still places generative imagery in the critical path of retail conversion. Merchants watching assistant-mediated discovery will care whether try-on outputs preserve brand colorways, size cues, and fabric authenticity—or whether they become stylized approximations that create returns.
Google already launched virtual try-on last year, TechCrunch notes, so ChatGPT is entering a contested surface rather than inventing the category. Differentiation will depend on Images 2.5 quality, how smoothly try-on sits beside conversational shopping, and whether Favorites becomes a durable consideration set rather than a disposable gallery. PromptCrates readers comparing creative shopping tooling can also weigh OpenAI’s move against broader agent shopping and assistant packaging covered in stories such as Meta Muse for small business and Shopify and product-adjacent media work like Google’s Diffusion Controller image steering.
Favorites, Images 2.5, and the shopping funnel
Favorites is the organizational half of the launch. Users can save products they discover to a Library and return later, with try-on images stored alongside those items. For retailers, a persistent Library is more commercially interesting than one-off generations because it creates a revisit loop inside ChatGPT rather than bouncing shoppers back to search. For privacy and trust teams, the same loop raises questions about biometric-adjacent selfies, retention of body imagery, and whether saved outfits become training or advertising signals. TechCrunch’s report does not spell out retention defaults; buyers and consumers should read OpenAI’s product privacy disclosures before treating try-on as casual.
OpenAI’s claim that Images 2.5 improves lighting, texture, instruction following, and latency is central to whether try-on feels useful. Fashion visualization fails when sleeves melt into skin, fabric weight looks wrong, or edits ignore “shorten the hem.” If Images 2.5 materially reduces those failure modes, ChatGPT can become a serious consideration tool. If not, users will treat outputs as mood boards and still complete purchases on retailer sites with better fit data. Either outcome still expands the assistant’s role upstream of checkout—the part of the funnel OpenAI previously struggled to own with instant buy.
Competitive pressure is rising from agent platforms that mix shopping with always-on tasking. Teams evaluating assistant platforms should ask a practical question: does shopping visualization increase session value enough to justify image compute costs, and does Favorites reduce abandonment better than browser bookmarks? Those metrics will decide whether this is a durable commerce surface or a demonstration feature.
What retailers and shoppers should test first
Retail partners and brand marketers should run controlled try-ons across diverse body types, skin tones, lighting conditions, and garment categories before promoting the flow in campaigns. Measure color fidelity, occlusion around hands and hair, accessory scale, and whether users understand the output is a visualization rather than a size guarantee. Customer-care teams should prepare macros for “the try-on looked perfect but the size was wrong,” because generative fit is not the same as size recommendation. Shoppers should avoid uploading sensitive photos if they are unsure how long images remain associated with their Library. Teams should also log which garment categories produce reliable try-ons in their own tests, so merchandising decisions rest on measured results rather than on launch demos.
Primary reporting for this article is TechCrunch’s 1 October 2026 story by Sarah Perez on ChatGPT’s global try-on and Favorites launch. Anchored facts include the two-feature announcement, Images 2.5 as the claimed visual engine, selfie or full-body upload paths, Try On in shopping results, screenshot-based try-on, Favorites Library storage with try-on images, style-description and celebrity-outfit shopping examples, the underwhelming prior instant-checkout pivot, Instinct’s recommendation backlash as industry context, and Google’s earlier virtual try-on as competitive backdrop.
