It’s early 2026 and a VP of e-commerce is puzzled by the analytics dashboard in front of them. Organic search traffic is flat and search efficiency is declining, but conversion rates from a small and growing slice of “direct” traffic are anomalously high — six, eight, sometimes 10x the conversion rate of search traffic.
Their analyst traces it and discovers the source of this mysterious traffic: AI agents. These shoppers aren’t arriving to make a decision. They’re arriving to complete one.
The question this VP is sitting with is what every e-commerce leader should be asking right now: “If shoppers are making decisions before they reach my site, what does my brand look like in the place where those decisions are being made?”
For decades, e-commerce relied on a relatively stable equation:
- A shopper searched.
- They clicked.
- They browsed categories.
- They landed on a PDP.
- They evaluated the product.
Every major investment in e-commerce was built around this five-step funnel. SEO drove top-of-funnel discovery, merchandising guided browse behavior, and PDP optimization improved conversion. The website wasn’t just the transaction layer — it was the research environment itself.
Today, the data tells a different story:
- AI-driven traffic to U.S. retail sites grew 4,700% year over year between 2024 and 2025.
- Conversions from AI referrals increased 1,247% in late 2025.
- AI and agents influenced 20% of all U.S. retail sales during the 2025 holiday season.
The analytics dashboard isn’t lying to our VP of e-commerce. Traffic is down, but not because it’s disappearing. It’s relocating to environments outside of the brand’s control as consumers increasingly use AI agents like ChatGPT, Google Gemini, and Amazon Rufus as personal shopping assistants.
A user can now describe what they want in natural language, receive tailored recommendations, compare options, and make a buying decision without ever visiting a brand’s website.
Anomalously high-converting traffic streams are a direct result of this new reality. Users aren’t arriving to research products, but to complete a decision that was made upstream inside the agent conversation itself.
That dynamic is already creating separation between brands. As discovery shifts into agent-driven environments, brands optimized for machine-readable product data, rich visual context, and structured metadata are beginning to capture disproportionate visibility with high-intent shoppers.
For most of the modern internet era, e-commerce teams had a single discovery system to optimize for: Google.
The rules were complicated, but they were stable. Brands invested in SEO because they understood how visibility worked. Search rankings depended on a relatively consistent set of signals, and entire e-commerce organizations were built around improving performance inside that ecosystem.
The landscape that has replaced it is structurally different. There is no longer one agent making discovery decisions. There are many — each with different architectures, different data ingestion mechanisms, different trust signals, and different format requirements.
Some rely heavily on structured commerce feeds, while others lean more heavily on web indexing, merchant authority, customer reviews, or visual interpretation. For example:
- ChatGPT Shopping sources products through merchant feeds, web indexing, and Shopify Catalog integrations.
- Google AI Mode builds on existing Search authority while evaluating increasingly sophisticated product schema and structured data based on Google Merchant Center.
- Amazon Rufus operates mostly within Amazon’s ecosystem, weighing product data, reviews, and rich visual content when surfacing product recommendations.
- Perplexity Snap to Shop blends conversational discovery with computer vision, using product imagery to power visual search and recommendations.
- Microsoft Copilot pulls from Bing Shopping and merchant catalog feeds to surface product recommendations inside conversational workflows.
When agents require varying methods of optimization, a brand that has done everything right for one may be invisible to others. What works for Google may not work for Rufus or Copilot.
This alters the definition of multichannel commerce. Historically it was about managing data across Amazon, Walmart, your website, and a few other channels. Now it means ensuring that your product’s visual and semantic profile can be understood across a growing ecosystem of AI agents.
Not only is that ecosystem growing, but it’s constantly evolving, too. Agents continuously change how they ingest product data, interpret visual assets, and determine what they can recommend with confidence.
Most e-commerce organizations are not built for that reality yet. In fact, many don’t even realize it’s here.
For years, e-commerce visibility was shaped by the search box.
Shoppers translated what they wanted into queries, like “outdoor wedding shoes,” instead of describing the actual situation: “I need something comfortable enough to stand on grass for four hours at my sister’s wedding.”
Brands optimized product titles, descriptions, and metadata to align with those queries. The goal was simple: Create the strongest possible match between what a shopper searched for and what a product page contained.
Search engines have since become more sophisticated. Google can understand context, infer intent, and connect concepts in ways that would have been impossible a decade ago. But the shift to AI agents introduces a different challenge.
Search engines work from the clues shoppers provide. AI agents allow shoppers to describe exactly what they want.
When a shopper asks an AI agent, “What should I wear to my sister’s outdoor wedding in June? I’ll be standing on grass for four hours, my dress is light blue, and I’d like to stay under $150,” the system is not simply retrieving relevant results. It is evaluating products against a highly specific set of requirements and recommending the ones it can justify with confidence.
That changes the optimization challenge for brands.
Years of investment in search visibility do not automatically translate into recommendation visibility. A product that ranks on page one for “casual wedding shoes” is not necessarily the product an AI agent recommends when a shopper describes a specific occasion, budget, dress color, and comfort requirement.
The question is no longer just, “Can this product be found?“
It is, “Can this product be recommended confidently?“
That confidence comes from three things:
- Attribute completeness. Does the product data describe the full range of situations the product can serve?
- Semantic richness. Do images, metadata, and alt text describe what the product actually looks like and how it might be used?
- Visual context. Does the catalog contain imagery that helps an AI system connect the product to the scenario the shopper is imagining?
Consider two nearly identical shoes.
The first has a keyword-optimized title, a generic description, and a standard white-background image.
The second includes occasion-specific attributes, a lifestyle image tagged for outdoor weddings and summer events, and alt text describing an ivory block-heel sandal being worn in an outdoor setting.
Both products may be discoverable, but only one gives the agent enough evidence to confidently answer the shopper’s question.
That is the real shift underway. The optimization target is moving beyond keyword relevance toward recommendation readiness. The brands that win will not be the easiest to find, but the easiest for AI systems to recommend.
There’s a temptation to treat agent-driven discovery as a future problem, but authority in this new landscape is being established now.
AI traffic to retailers rose 393% in Q1 of 2026. And that traffic converts up to 9x higher than traditional search traffic.
These are not casual browsers. They are shoppers arriving with a high degree of purchase intent.
The brands capturing this high-value traffic tend to have one thing in common: Their products are easy for AI systems to understand and recommend. Their catalogs contain the structured attributes, visual context, and semantic signals agents use to build confidence in a recommendation. In other words, they have made it easier for machines to choose their products, not just find them.
The brands that aren’t capturing this traffic face a different challenge: they often don’t know they’re missing it.
Unlike a Google ranking drop, which shows up clearly in analytics, agent invisibility is silent. When an AI system excludes a product from a recommendation set, the shopper never visits the site. The traffic never arrives. The brand may not even realize it was considered and rejected.
That makes the problem deceptively easy to ignore. Lost visibility in traditional search creates obvious signals. Lost visibility in agent-driven discovery creates absence. The opportunity doesn’t show up as declining traffic — it simply goes somewhere else.
It also makes waiting expensive.
The brands investing in machine-readable catalogs and intent-rich visual assets today are beginning to establish an advantage that looks familiar. Just as early SEO investment created authority that took years for competitors to overcome, early investments in agent visibility are beginning to shape which products AI systems understand, trust, and recommend.
At its core, the problem is surprisingly simple. Most product catalogs were built for human shoppers arriving through search engine, not machine readers. The products are there. The images are there. What is often missing is the AI translation layer — the metadata, attributes, and contextual signals that help systems understand products — as well as the intent-ready visuals that help those systems recommend them confidently.
The good news is that most brands already have much of the foundation in place.
The products are in the catalog. The images are sitting in a DAM. The commerce infrastructure is already in place. The challenge is less about starting from scratch and more about enriching what already exists while building the intent-aware visual experiences that many catalogs still lack.
What is often missing is the connective layer that brings those pieces together and makes them legible to the systems increasingly mediating discovery.
In practical terms, that layer needs to do four things:
Every product image should describe what it actually shows — not just the product name, but the color, material, angle, setting, occasion, and use case. Product videos need transcripts and chapter markers. Metadata needs to be structured, accessible, and continuously maintained as catalogs evolve.
AI agents are increasingly asked occasion-based questions.
- “What should I wear to an outdoor wedding?”
- “What’s a good gift for a new parent?”
- “Which sofa works in a small apartment?”
Answering those questions requires lifestyle imagery that connects products to real-world situations and makes those situations machine-readable. Brands should be building a library of imagery tied to the occasions, use cases, and contexts that drive purchase decisions.
Different agents prioritize different formats and signals. The same product may need to be represented differently for Amazon Rufus, ChatGPT, Google AI Mode, or Perplexity. Managing those variations manually at catalog scale is not realistic. Brands need systems that can automatically enrich, transform, and distribute product content based on the requirements of each discovery environment.
AI agents can’t recommend products they can’t access. As discovery shifts beyond websites and search results, brands need a reliable way to expose catalog data, product attributes, and visual assets to the systems mediating commerce. The specific protocols will evolve, but the requirements will endure: Product information must be structured, accessible, and ready to move wherever discovery occurs.
The goal is not to rebuild your commerce stack.
It is to make the assets you already own understandable, explainable, and recommendable to the systems shaping how people shop.
Somewhere, that VP of e-commerce is still looking at the same dashboard.
The difference is that what once looked like an anomaly now looks like a signal — a signal that some of the most valuable traffic in e-commerce is no longer being won on the website itself, but in the systems mediating discovery and recommendation.
There is a useful precedent here. The brands that recognized the strategic importance of SEO in the early 2000s built advantages that compounded for years. The brands that embraced mobile commerce early benefited from a similar dynamic. Agent-driven discovery is beginning to show the same characteristics.
The encouraging part is that most brands do not need to start from scratch. The infrastructure largely exists.
What is missing is the combination of AI-ready product intelligence and intent-driven media that allows products to be understood, adapted, and recommended consistently across a growing ecosystem of discovery channels.
The gap between “not ready” and “ready” is probably smaller than it appears. And the window to build that advantage is still open. It just won’t stay open for long.
