If you’ve checked your analytics lately and noticed a referral source you can’t quite place, “chatgpt.com” or “perplexity.ai” next to Google and Instagram, you’re not imagining it. Shoppers are starting to ask AI assistants what to buy before they touch a search engine, and that shift already shows up in conversion data, not just headlines.
Our finding: shoppable video today still gets discovered overwhelmingly through PDP and ad placements. But the merchants asking us about AI agents aren’t asking “should I care” anymore, they’re asking “what do I need to fix first.” This guide answers that directly: what AI shopping agents actually do right now, what they read to recommend a product, and where your video content fits, backed by the data points below.
The shift
How Are AI Shopping Agents Actually Being Used Right Now?
In 2026, AI-referred traffic to US retail sites is up 138% year-over-year, and 1,324% since Adobe began tracking it in October 2024 (Adobe, via Digital Commerce 360, 2026). That’s not a rounding error, it’s a new acquisition channel showing up whether you’ve built for it or not.
The more interesting number isn’t traffic volume, it’s what that traffic does once it lands. Shoppers referred by AI assistants now convert 31% better than other traffic sources, spend 45% more time on-site, and view 13% more pages per visit (Adobe Analytics, 2026). AI-driven revenue per visit is up 254% year-to-date this holiday season, which tells you the agents themselves got better at matching intent to product, not just at driving clicks.
Salesforce’s analysis of over 1.5 billion shoppers found AI influenced 20% of global online sales during the 2025 holiday season, worth an estimated $262 billion, and retailers running their own shopper-facing AI agents grew sales 59% faster than retailers still on the sidelines (Salesforce, 2026). AI-referred traffic converted at 8x the rate of social referral traffic in the same analysis.
Shopify’s own numbers
What Does This Look Like Specifically for Shopify Merchants?
Shopify’s own data shows AI-referral sessions to Shopify storefronts grew more than 8x year-over-year, with AI-referred orders up nearly 13x in Q1 2026, versus organic search sessions growing only about 5% over the same window (Shopify, 2026). That gap alone should reorder your priority list if AI referral isn’t on it yet.
The behavior pattern differs too. AI-referred visitors convert at nearly 50% higher rates than organic search visitors on product pages, carry a 14% higher average order value, and more than half of AI-referred sessions land directly on a product detail page, versus about 20% for organic search (Shopify, 2026). Organic search shoppers browse and compare; AI-referred shoppers arrive with the agent’s homework already done.
That has a direct implication for shoppable video: when an AI agent sends a shopper straight to a single PDP instead of a category page, the video content on that exact PDP, and the data describing it, matters more than your homepage or collection-page strategy. See our guide to winning the product detail page for what a high-converting PDP needs beyond video.
Where the agents stop
Are AI Agents Actually Completing Purchases, or Just Recommending Products?
Right now, mostly the latter. OpenAI discontinued ChatGPT’s Instant Checkout in March 2026 after weak merchant uptake, a signal in-chat checkout hasn’t found product-market fit yet, even with the infrastructure built (trade press coverage, 2026). Consumers agree: Accenture surveyed 25,590 consumers across 16 countries and found 74% would trust a personal AI agent more than their best friend to help pick a purchase, but only 32% would let an agent choose within a set budget and brand, and just 9% are open to a fully autonomous purchase (Accenture, 2026).
That gap, trust the recommendation, hesitate at the transaction, is the whole story right now. Would you let an assistant spend your money without a final look? Most shoppers wouldn’t either, and only 12% are open to letting an agent make the purchase decision itself (Accenture, 2026). Payment networks are building for a future where that changes anyway, Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay both formalized agentic-payment infrastructure through 2025-2026, with Stripe expanding to support both network-led protocols (Digital Commerce 360, 2026), but that’s infrastructure for a shift that hasn’t landed, not a reason to deprioritize discovery today.
The practical takeaway: don’t spend your next quarter building agentic checkout. Spend it making sure your product data, including video, is something an agent can actually recommend accurately when a shopper asks.
What agents actually read
What Do AI Shopping Agents Actually Read to Recommend a Product?
Structured, machine-readable data, not marketing copy. Ahrefs studied 1,885 pages that added JSON-LD schema markup against a control set drawn from a 6-million-URL dataset and found pages already cited by AI tools were about 3x more likely to carry schema markup than non-cited pages (Ahrefs, via Search Engine Journal, 2026). Worth being precise: that’s a strong correlation, not proven causation. Experimentally adding schema to pages that lacked it produced no statistically significant citation lift. Structured data looks like a marker of a well-organized, agent-friendly site, not a lever you pull in isolation.
That distinction matters for how you prioritize. If schema alone doesn’t move citations, the underlying signal AI agents seem to reward is a clean, consistent, machine-readable product record across your whole catalog: accurate pricing, real-time inventory, structured attributes, and video tagged and described the same way. Shopify activated “Agentic Storefronts” by default for eligible merchants on March 24, 2026, making products from 5.6 million stores discoverable across ChatGPT, Microsoft Copilot, Google AI Mode, and the Gemini app (Shopify, via industry coverage, 2026). If you’re on Shopify, some of this groundwork may already be live; verify your own storefront’s eligibility rather than assuming.
| Surface | What it currently does well | Checkout status (mid-2026) |
|---|---|---|
| ChatGPT (OpenAI) | Conversational product recommendations, comparison shopping | Instant Checkout discontinued March 2026 |
| Google AI Mode / Gemini | Search-integrated shopping results, product comparison | Discovery-focused; checkout still routes to retailer sites |
| Amazon Alexa for Shopping (formerly Rufus) | In-marketplace product Q&A and recommendations | Native Amazon checkout only |
| Microsoft Copilot Shopping | Cross-retailer comparison inside Copilot sessions | Discovery-focused; routes to retailer sites |
| Shopify Agentic Storefronts | Structured catalog exposure to third-party AI agents | Merchant-controlled; checkout stays on-site |
The video angle
Where Does Shoppable Video Fit Into AI-Driven Product Discovery?
Here’s the honest state of the evidence: no rigorous, published study yet isolates video content specifically as a ranking or citation signal for AI shopping agents, the way Ahrefs isolated schema markup. That’s a real gap in the public research, not something this article is going to paper over with an invented number.
What we do see, working with Shopify and ecommerce brands running shoppable video at scale, is that the same discipline driving AI citation for text and schema applies directly to video. An agent parsing your catalog can’t watch your product video the way a shopper can. It can read the transcript, the alt text, the product tags attached to the video, and the structured data describing what’s shown. If that layer is missing, your video is invisible to the exact systems now driving higher-converting traffic to Shopify stores.
This is why Whatmore’s AI Discovery feature exists as a distinct capability rather than an afterthought: it structures the product-tag and metadata layer around your shoppable video, the same underlying discipline the Ahrefs data points to for text content. Pair it with a clean product catalog feed and smart product matching so the video-to-product relationship is unambiguous to your own site search and to any agent reading your storefront.
Brands already treating video metadata as a first-class asset are seeing it pay off in conventional conversion terms, which is the leading indicator worth tracking while agent-specific data catches up. Nish Hair hit 62x ROI after adding structured, tagged shoppable video to its site, the kind of clean product-video relationship that also happens to be exactly what an AI agent needs to parse correctly.
Do this next
How Do You Prepare Your Shoppable Video Content for AI Shopping Agents?
By the end of this checklist, your product and video data should be in a shape both human shoppers and AI agents can actually use.
- Audit your product feed for completeness. Missing prices, stale inventory, and inconsistent variant naming are the fastest way to get skipped by any agent reading your catalog.
- Confirm your PDPs carry accurate, current Product schema markup. It won’t guarantee citation alone, but it’s a strong marker of the kind of site AI tools already trust.
- Add descriptive alt text and transcripts to every shoppable video, not just top sellers. An agent can’t infer what it can’t read.
- Tag products consistently within video content so the video-to-SKU relationship is unambiguous, not inferred from page proximity.
- Check your analytics for AI-referral traffic now (chatgpt.com, perplexity.ai, and similar referrers) so you have a baseline before you invest further.
- Prioritize your highest-traffic PDPs first. With more than half of AI-referred sessions landing directly on a product page, that’s where returns show up fastest.
Measuring it
How Do You Know If This Is Actually Working?
Watch AI-referral traffic as its own segment, not lumped into “direct” or “other.” Most analytics platforms misclassify chatgpt.com and perplexity.ai referrals unless you build a dedicated segment for them. Then watch conversion rate and AOV on that segment specifically, not just volume: given that AI-referred traffic already converts 31% better than other traffic sources industry-wide (Adobe Analytics, 2026), a segment converting meaningfully below that benchmark signals your product or video data isn’t giving agents what they need.
PYMNTS surveyed US adults in January 2026 and found interest in using AI for shopping-adjacent tasks ranged from 54% for school and parenting purchases up to 71% for health and wellness management (PYMNTS, 2026). Treat that as a baseline to build from now, not a channel to revisit “later.”
Common mistakes
Mistakes to Avoid When Optimizing for AI Shopping Agents
Building an agentic checkout integration before fixing your product data. Checkout adoption is stalled industry-wide, OpenAI’s own Instant Checkout shutdown proves it. Discovery is where the traffic and conversion lift already exist.
Treating schema markup as a silver bullet. The Ahrefs data is clear that adding schema alone didn’t move citation rates in a controlled test. It’s a marker of a clean, agent-readable site, not a standalone fix.
Leaving video out of the structured-data conversation. If your PDP’s video isn’t tagged and transcribed consistently, it’s the least legible part of your page to an agent, even if it’s the most persuasive part to a human shopper.
The most overlooked mistake: assuming AI-referred traffic is too small to matter yet. At 138% year-over-year growth and already converting better than most other channels (Adobe, 2026), “too small to matter” has a short shelf life.
Next steps
Next Steps
AI shopping agents are already sending higher-converting traffic to Shopify stores, landing overwhelmingly on product detail pages, exactly where your shoppable video lives. The near-term opportunity isn’t a checkout integration; it’s making your product feed, schema, and video metadata clean and consistent enough for an agent to read accurately.
If you haven’t audited your AI-referral traffic yet, start there this week. If your video content isn’t tagged and transcribed consistently, that’s next. For a broader look at video in the current commerce landscape, see our guide to video commerce in 2026, and for benchmark data on shoppable video performance, see the 2026 shoppable video benchmarks report.
Ready to install? Try Whatmore free on your Shopify store — no credit card required.
Book a Demo →
