Could 'LLM SEO' or AI-search optimization become its own discovery channel for ecommerce, separate from Google SEO?
AI search is already a separate channel for Shopify merchants on one axis and the same channel on the other. Shopify's Agentic Storefronts put products into ChatGPT, Microsoft Copilot, Google AI Mode and Gemini from one admin, with order attribution (Shopify, March 2026). That is feed work with its own owner. The answer content is still SEO.
Split the question in two, because it is two questions
"Does AI search deserve its own budget line and owner" gets a different answer depending on which half of AI search you mean. One half is a distribution surface with its own plumbing, its own attribution and its own product-data requirements. The other half is your content ranking well enough to be quoted. Treating them as one line item is why the budget conversation stalls.
| Agentic / feed surface | Answer / citation surface | |
|---|---|---|
| What it is | Products syndicated into ChatGPT, Copilot, Google AI Mode, Gemini via Shopify Catalog | Your pages quoted inside AI answers |
| How it works | A catalog feed managed from Shopify admin (per Shopify, Mar 2026) | Crawl, rank, extract |
| Google's position | A commerce integration built on the Universal Commerce Protocol co-developed with Google (per Shopify, Mar 2026) | "The best practices for SEO remain relevant for AI features" (per Google, Sep 2026) |
| Attribution | Orders land in the Shopify admin with channel attribution (per Shopify, Mar 2026) | No engine reports referrals reliably; needs an owned prompt panel |
| Deserves its own owner | Yes | No, it belongs to whoever owns content |
| Deserves its own budget line | Yes, mostly product-data and merchandising work | Only for measurement |
The feed half is genuinely new, and it is not marketing work
Shopify announced in March 2026 that merchants can sell to ChatGPT users through Agentic Storefronts, with "no separate integrations, no apps, and no transaction fees beyond standard processing rates," and with orders arriving in the admin carrying ChatGPT attribution (per Shopify, March 2026, confirmed on a second Shopify newsroom post, January 2026). Microsoft Copilot, Google AI Mode and the Gemini app are managed from the same place.
That means the levers are product titles, attributes, variant structure, images and inventory accuracy. It is merchandising and data hygiene. The person who should own it is whoever owns the product catalog, not whoever owns the blog.
The answer half is not a new channel, and the vendors selling it as one have a tell
Google's guidance is explicit that no additional requirements or special optimizations exist for AI Overviews and AI Mode, and that no new machine-readable files or AI text files are needed (per Google Search Central, September 2026). Every product that promises AI placement is therefore selling something no engine offers. The tell is the phrase "AI visibility score": it measures the vendor's checklist, and it rises when you complete the checklist whether or not any engine changes its answer.
What is real on this side is measurement. You cannot see AI citations in analytics, so if you want a trend you have to run a prompt panel yourself and log the results over time.
What we would actually put in a 2027 plan
- A named owner for the catalog feed, sitting in merchandising or ecommerce ops, with product-data completeness as the KPI.
- A prompt panel, quarterly, built from real customer language rather than from the marketing team's phrasing. Our decision record prices the panel plus the first citable-content sprint at $10,000 to $30,000 (Deploi estimate, illustrative, September 2026), and puts the first useful trend one quarter out.
- No separate content team. The content work is the SEO work, restructured for passage extraction.
When this answer flips
If an engine ever publishes a placement product or a documented ingestion format that is not a feed, the answer half becomes a channel overnight. It has not happened. llms.txt was the candidate, and Google's AI-features documentation names no such file.
The Deploi point of view
Our own position, from building on Shopify. Separate from the facts above.
- Our take: Give the feed its own owner and its own line. Do not give the content half a separate team, a separate strategy deck or a separate agency. We scored AI answer-engine visibility CUSTOMIZE for this split: free-to-cheap technical hygiene plus native agentic surfaces, then an owned prompt panel and citable content (Deploi verdict, September 2026).
- What we’ve seen: The budget request usually arrives shaped like a Google SEO retainer, because that is the shape the requester knows. The work that pays is shaped like a product-data cleanup plus an editorial restructure, and neither of those looks like a retainer.
- Where we disagree: The consensus framing is "AI search is the new channel, staff accordingly." Our read is narrower: the commerce surface is a new channel with real plumbing, and the citation surface is SEO wearing a new name. Staffing for two channels when one of them is your existing content team is how this budget gets wasted.
- What this page adds: that the question splits into a feed surface with native Shopify attribution and a citation surface with none, and that only the first one justifies a separate owner.
Reviewed by Martin Dejnicki, Director of SEO & AI Search. Facts verified 2026-09-13.
Where we worked this out
Our decision records