Do AI answer engines like ChatGPT and Gemini still use traditional SEO signals (meta titles, schema, backlinks), or do they rely on something else entirely?
AI answer engines run on traditional SEO signals, with one change. Google states there are no additional requirements to appear in AI Overviews or AI Mode and no special schema.org structured data to add (Google Search Central, September 2026). The change is the unit of retrieval: engines lift passages, not pages, so extractable answer structure decides citation.
The mechanism, stated by the people who run it
Google's own documentation is unusually direct about this. "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary," and "you don't need to create new machine readable files, AI text files, or markup to appear in these features" (per Google Search Central, verified September 2026). The same page says there is no special schema.org structured data required. That sentence quietly kills a large amount of what is currently sold as AI SEO.
What sits underneath the answer is the same ranking stack. Google's AI features run a query fan-out, issue related sub-queries, and assemble an answer from pages that the ranking systems already surface. If your product guide does not rank for the sub-query, it is not in the pool.
ChatGPT works differently in one respect that matters operationally. OpenAI runs three separate crawlers: GPTBot for model training data, OAI-SearchBot "used to surface websites in search results in ChatGPT's search features," and ChatGPT-User, which "may visit a web page" live when a user asks a question (per OpenAI's developer docs, September 2026). Three crawlers, three jobs, one robots.txt file that governs all of them.
Signal by signal, what still earns its budget
| Signal | Still load-bearing for AI answers | Why |
|---|---|---|
| Title tag | Yes, indirectly | It does its normal ranking and click job; no engine quotes it as the answer |
| Backlinks and authority | Yes, indirectly | They feed ranking, and ranking feeds the candidate pool for AI features |
| Product and Article schema | Yes, for product surfaces | Not required for AI Overviews per Google (Sep 2026), but structured data plus a Merchant Center feed "maximizes your eligibility to experiences and helps Google correctly understand and verify your data" (per Google, Sep 2026) |
| Crawlability and render | Yes, hard requirement | An answer engine cannot cite HTML it never received |
| Page speed | Weakly | It is a ranking input and a conversion input, not a citation input |
| llms.txt | No | Google's AI-features guidance names no such file; no engine documents reading one |
| Answer structure on the page | Yes, and this is the new part | Passage-level retrieval means a self-contained 40 to 60 word answer directly under the question is the unit that gets lifted |
What actually changed
The retrieval unit moved from page to passage. That is the whole shift, and it has two practical consequences for a $20M to $500M merchant.
First, a 2,400-word buying guide with the answer buried in paragraph nine competes badly against a 600-word page that answers in the first forty words. The ranking signals are the same; the extractability is not.
Second, your product data has to be complete in markup, not just in the admin. Engines describing a product read the Product node. On Shopify the native structured_data Liquid filter emits Product or ProductGroup for products and Article for articles (per Shopify developer docs, September 2026). Organization, BreadcrumbList, FAQ and review markup are theme work, which is why our Build vs. Buy verdict on structured data is BUILD for any store with a developer, at an estimated $3,000 to $8,000 one-time (Deploi estimate, illustrative, September 2026).
Edge cases worth knowing
- Product discovery inside ChatGPT is not a crawl channel at all. Shopify's Agentic Storefronts push merchant products into ChatGPT, Microsoft Copilot, Google AI Mode and the Gemini app from the Shopify admin, through Shopify Catalog (per Shopify, March 2026). That is a feed, not a citation. Content SEO does not control it.
- An AI-visibility score sold by an app is the app's checklist, not an engine's behavior. No engine publishes a placement product, so nothing can be bought.
- Robots.txt changes take time to propagate. OpenAI notes it "can take ~24 hours from a site's robots.txt update for our systems to adjust" for search results (per OpenAI, September 2026).
The Deploi point of view
Our own position, from building on Shopify. Separate from the facts above.
- Our take: The signals overlap heavily with traditional SEO. What changes is the unit of retrieval, so the work that actually moves citation is extractable answer structure and complete product markup, not title tags. We scored AI answer-engine visibility CUSTOMIZE for exactly this reason: buy the cheap technical hygiene, then build the measurement and the content, because no app can promise that ChatGPT, Gemini or Perplexity will recommend you (Deploi verdict, September 2026).
- What we’ve seen: Teams arrive with a list of AI-specific tactics and leave with a content restructuring project. The most common finding in the first week is not a missing file. It is that the pages answering the highest-intent questions bury the answer under a scroll of brand preamble.
- Times we’ve shipped this: 6 builds delivered.
- What it takes: roughly 96 hours of scoped work for a structured-data-for-AI-answers build (directional Deploi estimate from a small sample of engagements, not a measured average).
- Where we disagree: The category says AI search is a new discipline requiring new artifacts. Google says in writing that it is not, and we agree with Google here. The honest new work is narrow, specific and mostly editorial, and vendors who cannot sell editorial work have an incentive to describe it as infrastructure.
Reviewed by Martin Dejnicki, Director of SEO & AI Search. Facts verified 2026-09-13.
Where we worked this out
Follow-up questions
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.
Does a migration to Shopify ever cause a temporary drop in AI-answer-engine citation share even if traditional Google rankings hold steady?
A Shopify migration costs AI answer-engine citation share through two mechanisms that Google rankings never reveal: changed URLs break the links engines already cite, and every engine re-crawls on its own separate cycle. OpenAI runs a dedicated crawler, OAI-SearchBot, for ChatGPT search results (OpenAI developer docs, September 2026). No engine publishes a re-crawl interval. Budget one quarter.
What's the honest answer on whether AI Overviews and AI answer engines even care about a site's migration history, or only about current content quality?
AI answer engines carry no migration-history signal. Google's published guidance names no ranking or AI-features factor for platform history, and states that AI Overviews and AI Mode need no additional requirements beyond standard SEO (Google Search Central, September 2026). What a migration changes is current content quality: URLs, redirects, rendered markup and page speed. Fix those four and the history stops mattering.