Adobe tracks more than a trillion visits to retail sites in the United States. During the 2025 shopping season, traffic referred by AI jumped 693% year-on-year.
More important than volume is the fact that the conversion behavior changed completely. In January 2025, visitors arriving from ChatGPT converted 49% worse than other sources. By October, they converted 16% better. During the shopping season, traffic from AI converted 31% better than traditional sources.
People stopped using Chat GPT for research and started using it to buy directly.
Salesforce reports that AI influenced 17% of orders during Thanksgiving weekend 2025. That's $13.5 billion in sales. From zero to 17% of all holiday orders in one year.
If your ecommerce strategy continues to optimize only for Google, you're ignoring a channel that's already generating higher conversions.
Microsoft released Copilot Checkout. Google launched Universal Commerce Protocol. OpenAI added native checkout in ChatGPT.
The payment infrastructure works. If someone finds your product in an AI agent, they can complete the purchase without leaving the chat. That's already been resolved.
What's not resolved is how the agent decides to recommend you in the first place.
Being integrated with Shopify or having checkout enabled is table stakes. It assures you that you are not left out. But it doesn't generate discovery. And discovery is where you win or lose.
Traditional SEO for ecommerce hasn't disappeared. It remains critical.
Descriptive URLs, unique meta titles, keyword-rich descriptions, images compressed with alt text. This is still the basis.
Logically organized categories, breadcrumbs, internal linking that connects related products. Google and users need to understand your hierarchy.
Product schema with price, availability, SKU, reviews. FAQ schema for common questions. Review scheme for ratings. Google uses this for rich results.
Loading speed, responsiveness, visual stability. Slow sites lose both in SEO and in conversion.
Buying guides, comparisons, how-tos. They capture search intent early in the funnel.
All of this is still working. But if you're left alone here, you're wasting space in AI searches and Google product feeds.
LLMs don't crawl your site like Googlebot. They don't follow links. They don't calculate PageRank.
They extract information from structured sources, prioritize external validation, and build recommendations based on semantic data.
And the truth is that, although Google continues to crawl its site looking for the usual components, it is increasingly prioritizing structured information.
Google has Merchant Center. ChatGPT has its own merchant program where you upload feeds in JSON, CSV, TSV, or XML.
The quality of the feed determines if you show up. Google Shopping was already penalizing feeds with missing or incorrect information. Now ChatGPT does the same.
Critical Attributes:
Google added dozens of new attributes to Merchant Center designed specifically for conversational commerce. Answers to common product questions, descriptions of use cases, details of materials formatted for extraction.
If your feed only has the bare minimum, you're competing at a disadvantage.
Real-time update
Agents check availability and price in real time. If your feed shows available stock but the product is out of stock, the agent moves on to the next one without informing the user that you existed.
A human could click and discover that there is no stock. An agent simply removes you from the candidate pool.
Schema markup isn't new, but it's more critical now.
LLMs need to pair information quickly. If your specs are buried in marketing paragraphs or inconsistently formatted, the agent may not extract them or misinterpret them.
Someone asks: “cabin suitcase weighing less than 4 pounds”
If the weight is in a structured specs table, the agent extracts it. If you are in a sentence such as “incredibly light at only 3.8 pounds”, the agent has to parse natural language and may not capture it when comparing against competitors.
Essential diagram:
This helped with Google rich results. Now it also helps ChatGPT understand your product.
People don't talk to ChatGPT like they google.
Google: “ergonomic office chair”
ChatGPT: “I need a desk chair for someone who is 1.90m tall with chronic low back pain, maximum budget $400"
Your description needs to answer that type of query.
Wrong: “Premium chair with modern design and high quality finishes” The agent has no useful information.
Good: “Ergonomic design with adjustable lumbar support, weight capacity 136kg, adjustable height range 46-56cm, breathable mesh backrest” The agent can map this to the user's specific query.
What to include in the product copy:
This doesn't replace keyword optimization for Google. It's additional. But without this, ChatGPT can't recommend you when someone asks in a conversational way.
Agents don't just crawl your product pages. They consult the entire web and weigh certain sources more than others.
A product reviewed on Wirecutter or listed in an established buying guide has more authority than the same product described only on your site.
When an agent needs to recommend “the best affordable blender”, they look for sources they consider credible. If you only exist on your own site, you don't have external validation.
Where you need to be:
This is not new. It was always important for SEO. But agentic commerce matters more because agents are explicitly looking for recommendations, not just search results.
Actively pitch to review and comparison sites. Don't wait to be discovered.
Nearly every demo of AI shopping shows agents incorporating review sentiment into recommendations.
“Here are three highly rated options within your budget”
Review volume and ratings aren't just social proof for humans. They are ranking signals for agents.
A product with 500 reviews at 4.5 stars weighs differently than one with 12 reviews at 4.2 stars.
This creates a cold start problem for new products. But it's the reality. Agents use review data as a quality proxy.
If you don't have reviews, start getting them. Early access programs, post-purchase follow-up emails, incentives to leave reviews. Volume matters.
The platforms have not revealed what determines ranking within recommendations.
When an agent shows three options, what makes one rank first vs. third? We don't know.
Are there paid placement opportunities? Google is testing Direct Offers in AI Mode where advertisers can show exclusive discounts. But the model is not clear.
How much does brand recognition matter? Does an agent recommend Nike over a small brand purely by authority, or does he evaluate products objectively? Stranger.
What is the balance between price, quality, availability, and authority in the ranking algorithm? No idea.
This is being defined in real time.
Don't expect official guidance. Focus on what works regardless of how platforms evolve.
Make sure it's complete, accurate, and includes the new conversational attributes that Google added. This is the most direct path to appearing in Google's AI experiences.
Product, FAQ, Review, HowTo schemas give agents structured data to extract.
Pitch a review sites, comparison platforms, and publications in your category. External validation matters for agent recommendations.
Make sure that someone asking about use cases, constraints, or specific problems can find the answer in your copy.
If your feeds show incorrect availability or outdated prices, you're filtered before anyone sees you.
eMarketer projects that AI platforms will represent 1.5% of total ecommerce sales in 2026, approximately $20.9 billion. It's almost quadruple vs. 2025, but it's still small.
Traditional channels, Google, Amazon, paid search, organic, social, still generate the vast majority of discovery.
Merchants who win in AI are generally those who are already earning in traditional channels, because agents are extracting from the same underlying data.
This is not a separate channel that requires a completely new strategy. It's a new interface based on the same product information and signs of authority that already existed.
McKinsey estimates that agentic commerce could redirect $3-5 trillion in retail spend by 2030. The trajectory is real. But we are still in the infrastructure phase. The checkout mechanics work. The discovery and ranking logic is being built in real time.
Keep the fundamentals right, data clean, external validation, accurate inventory, conversational product descriptions, and you'll be positioned for whatever comes.
.png){kind=small}
SEO consulting that scales traffic, grows your brand, increases revenue, and helps you win on Google and AI platforms and LLMs like ChatGPT.
%402x.svg){kind=icon}
