### ASOS moves shoppable video into answer enginesASOS has launched an integration that lets shoppers discover and buy its products directly from within ChatGPT using shoppable video. Through the ASOS Stylist experience, users can describe what they are looking for in natural language – for example, “a pink floral summer dress” – and receive a curated set of products from the ASOS catalog. Each result can be viewed as a short-form video, then opened in detail on ASOS.com to complete the purchase. The pilot is being rolled out in the UK and US, and builds on the AI Stylist already available in the ASOS app.The experience is powered by Bambuser’s video commerce technology, which converts ASOS’ product catalog and video assets into structured data that can be retrieved in real time by large language models and rendered as shoppable video. ASOS positions this as an evolution beyond AI shopping tools based on text and static images, aiming to make fashion discovery “faster, easier, and more inspiring” in conversational interfaces such as ChatGPT.### From product search to “answer commerce”The integration illustrates a broader shift in e‑commerce from search- and listing-based journeys to “answer commerce”, where users pose conversational queries and expect a small, highly relevant set of options rather than navigating filters and pagination. In that model, the quality and structure of product data become as critical as price or assortment: the answer engine cannot show what it cannot parse.ASOS’ move is one of the more advanced attempts in the UK market to connect a large, fast‑changing fashion catalog with a general-purpose LLM in a way that preserves merchandising logic. The fact that the experience returns not only product tiles but also video suggests that ASOS has invested in mapping products, attributes, and media assets into an indexable, machine-readable layer that can be queried by natural language rather than rigid search syntax.### Implications for product feeds and content infrastructureTo make a conversational, video-first journey work inside an external interface like ChatGPT, an e‑commerce retailer has to treat its product feed as more than a simple export for ad platforms. Several consequences follow.First, feeds need richer attributes. A prompt such as “pink floral summer dress for a daytime wedding, under £80” can only be answered well if the underlying products are tagged with color families, patterns (floral), occasion, price, and often more subjective dimensions such as “summer” or “wedding guest”. Many merchants still rely on limited feeds optimized for price comparison engines; conversational commerce pushes them toward far more granular attribute taxonomies and controlled vocabularies.Second, feeds need a tighter connection to media. Shoppable video in answer engines requires mapping between SKUs and video segments: which products appear in which video, at which timestamps, and in which context (full look, close‑up, try‑on, styling tip). This is typically not captured in standard product feeds, which focus on images and basic metadata. ASOS and Bambuser are effectively treating video as a first‑class data object linked to the catalog, rather than as a loosely attached marketing asset.Third, feeds must be optimized for real-time retrieval rather than batch export. To remain relevant, recommendations surfaced by an LLM must reflect stock levels, price changes, and assortment updates. This pushes merchants toward APIs and streaming updates instead of nightly feed uploads, and toward architectures where product data, inventory, and content are synchronized tightly enough to be exposed safely to external answer engines.### Catalog standardization and machine-readabilityThe ASOS initiative highlights how catalog standards are becoming an enabling layer for AI-driven shopping. Bambuser’s “Intelligence Layer” is described as transforming the catalog and video library into structured data consumable by LLMs. In practice, this likely involves several steps that are becoming standard across advanced e-commerce stacks:- Normalizing product attributes into consistent schemas, so that “floral”, “flower print”, and “bloom pattern” collapse into a single machine-recognizable concept.- Enriching catalog entries using computer vision and language models to detect color, pattern, silhouette, and even inferred use cases when merchants’ own data is incomplete.- Indexing content for semantic search, so that prompts like “soft, breathable summer fabrics” can be matched with products described as “cotton poplin” or “linen blend” even if the exact text is not present.For the broader market, this indicates a direction of travel: catalogs are moving from human-readable but machine-fragile descriptions to machine-readable structures designed explicitly for AI retrieval. Retailers that maintain legacy, unstandardized catalogs will find it harder to participate in answer engines and conversational commerce environments.### Elevating product detail pages through videoBringing video into ChatGPT is not just a distribution play; it has implications for what a “good” product detail page (PDP) looks like. Short-form video has already proven effective in increasing conversion and decreasing returns in fashion by showing fit, movement, and styling context. Surfacing that same content directly in the discovery layer compresses the traditional funnel: inspiration, evaluation, and product selection happen in one interaction.To support that, the underlying product content has to be deeper and more consistent. Video needs to be available across a high share of SKUs, not just hero products. Size, fit, and fabric details must be accurate enough that customers can act confidently on what they see in video. Incomplete or inconsistent PDPs become more visible weaknesses when the first impression is an AI-curated set of shoppable videos instead of a grid of stills.There is also a feedback loop: user interactions with video (which products are paused, replayed, clicked through) create additional behavioral signals that can be fed back into ranking models and content prioritization. Over time, this can influence which products receive richer content and how merchants invest in video production.### Speed of assortment onboarding and content automationFor large fashion catalogs, one of the main bottlenecks in expanding assortment is content creation: titles, descriptions, attributes, imagery, and increasingly video. Deploying shoppable video in answer engines makes speed even more critical: newly onboarded products need to become discoverable in conversational journeys quickly, not weeks after initial listing.AI and automation become central to solving this. Computer vision can extract attributes from images, while language models can generate initial descriptions based on structured data. Video workflows can be partially automated through templates, motion graphics, and AI-assisted editing, with human teams focusing on brand-critical content and quality control. Merchants that systematize these processes can shorten the time between buying decisions and AI-ready catalog entries, giving them an advantage in dynamic categories.No‑code tools also play a role. Marketing and merchandising teams increasingly rely on interfaces that let them define segments, rules, and content variants without developer involvement. When plugged into AI-based discovery surfaces, such tools allow non-technical users to manage which collections, trends, or campaigns should be prioritized in conversational responses (“party season”, “festival looks”, “back to school”), while central systems ensure that underlying schema and tagging remain consistent.### Answer engines as a new distribution channelTreating ChatGPT and similar systems as answer engines rather than traditional media channels reframes e‑commerce distribution. Instead of buying impressions or keywords, retailers must compete on data quality, relevancy, and the ability to integrate safely with external AI systems. The ASOS integration suggests a model in which:- Catalog and content layers are exposed via secure APIs or widgets that answer engines can call, without giving them direct access to internal systems.- Merchants maintain control over pricing, checkout, and customer data by bringing users back to their own domains for transaction completion.- Personalization can be shared between the answer engine and the merchant’s systems through contextual signals, while complying with privacy constraints.This architecture has implications for analytics and attribution. Traditional web analytics are poorly suited to journeys that begin inside a conversational agent and continue across several surfaces. Merchants piloting such integrations will need to reconcile logs from AI platforms, video commerce layers, and their own web and app analytics to understand impact on conversion, average order value, and return rates.### Strategic significance for SaaS and e‑commerce ecosystemsFrom a SaaS perspective, the ASOS initiative illustrates a pattern: specialist platforms are emerging to sit between retailers’ raw data and general-purpose AI interfaces. These platforms handle ingestion, normalization, enrichment, and exposure of commerce data in forms that LLMs can use safely and efficiently. For e‑commerce businesses, this can reduce the need to build complex AI infrastructure in-house, but it raises the bar for upstream data governance and catalog hygiene.For the e‑commerce sector as a whole, deployments of conversational, video-first shopping inside answer engines accelerate several trends:- Product feeds evolve into real-time, richly attributed knowledge graphs.- Catalog standards and taxonomies become competitive assets rather than back-office concerns.- PDP quality is measured not only by on-site conversion, but by how well products can be represented in external AI-driven environments.- Assortment onboarding speed increasingly depends on AI-assisted content and media production.- No‑code and AI tooling become the operational layer that allows commercial teams to steer AI-driven journeys without constant engineering support.The ASOS pilot is early-stage and its commercial impact is yet to be quantified, but it signals a direction in which fashion e‑commerce is likely to move: from static listings to conversational, media-rich discovery that spans multiple platforms, powered by structured, machine-readable content at the core of the commerce stack.The ASOS integration with ChatGPT underscores the growing importance of structured, high-quality product data. This shift towards "answer commerce" highlights a critical need for efficient catalog management and enrichment. For businesses looking to achieve this, [NotPIM](/blog/product_feed/) provides a robust solution for ensuring product feeds are accurate, consistent, and ready for AI-driven applications. By automating tasks like attribute mapping, content enrichment, and feed transformation, NotPIM helps e-commerce businesses adapt to this evolving landscape, improving product discoverability and facilitating seamless integration with platforms like ChatGPT. This helps our clients stay ahead of the curve by optimizing their data for these new, innovative e-commerce experiences. This is particularly useful for improving [product feeds](/blog/product_feed/), which is crucial for AI integration. With the help of AI, you are able to create [sales-driving product descriptions](/blog/how-to-create-sales-driving-product-descriptions-without-spending-a-fortune/) quickly. Understanding how these systems work allows you not to be afraid of [complex integrations](/blog/xml-data-format-how-one-online-store-stopped-fearing-complex-integrations/). Merchants can also benefit from using a [price list processing program](/blog/price_list_processing_program/).