Over the past few months, many of you have probably seen the four abbreviations "GEO," "AIO," "AEO," and "LLMO" pop up one after another. All of them are used to describe SEO strategy for the AI era, but the explanations vary slightly from article to article, making it hard to tell which one is "correct" or which to start with.
To get straight to the point: these four terms refer to almost the same set of activities under different names. The only real differences are which AI services they focus on, or which country or industry the person using the term comes from — 70-80% of the actual work overlaps. Plenty of articles debate the hierarchy between these terms, but there's no objective consensus within the industry. Rather than spending time on terminology debates, what matters in practice is focusing on the fact that Google Search remains the main battleground, and AI-generated answer surfaces have simply been added on top of it, then prioritizing your efforts accordingly.
This article focuses narrowly on comparing the four terms and on practical decision-making material updated as of August 2026. For the basic definition of GEO and fundamental measures on Shopify (such as robots.txt settings and how to implement FAQ structured data), please also see the article below.
Sorting out the four terms: where they came from and how they're used
First, let's look at where each of the four abbreviations originated and where they're used.
GEO (Generative Engine Optimization)
GEO traces back to a paper first published as a preprint in November 2023 and accepted at KDD (an international conference in the data mining field) in 2024. Conducted jointly by researchers from Princeton University, Georgia Institute of Technology, IIT Delhi, and the Allen Institute for AI, it applied nine content strategies to 10,000 queries and measured citation rates within generative AI answers, making it the first empirical study in this field. Its academic origin is what sets it apart from the other three terms.
Source: GEO: Generative Engine Optimization (ACM Digital Library)
LLMO (LLM Optimization)
LLMO is a term widely used within Japan. It owes much of its spread to the book Strong LLMO: A Marketing Guide to Being Chosen in AI Search (強いLLMO AI検索で選ばれるためのマーケティングガイド), published on September 26, 2025 by Keita Takeuchi, CEO of LANY, and it's arguably the most established of the four terms within the Japanese-language sphere. There's a regional skew here: content that overseas literature would call GEO or AEO is often discussed using the term LLMO among practitioners in Japan.
AEO (Answer Engine Optimization)
AEO has become established as a practice-oriented term in English-speaking regions. Its scope tends to be somewhat broader, often discussed as covering not just ChatGPT Search and Perplexity but also Google AI Overviews and even optimization for voice assistants.
AIO (AI Optimization)
AIO is used as an abbreviation for "AI Optimization," but it's also used as an abbreviation for "AI Overview," Google's search results feature — and this shared abbreviation is one source of confusion. When you see the phrase "AIO measures," you need to check the context to know whether it's talking about optimization for AI in general or measures specifically for the Google AI Overviews feature.
Here's how the four terms compare:
| Term | Where it's mainly used | AI it's typically discussed for |
|---|---|---|
| GEO | Academic / global | Generative AI in general, such as ChatGPT and Perplexity |
| LLMO | Within Japan | LLM services such as ChatGPT and Gemini |
| AEO | English-speaking regions / practice | ChatGPT Search, Perplexity, Google AI Overviews, voice assistants |
| AIO | Global (usage varies widely) | Depending on context, either generative AI in general or specifically Google AI Overviews |
Opinions within the industry are divided on which term is the broader umbrella and which is the narrower subset, and there's no objective resolution to that debate. Here, we'll work from the premise that the names differ, but most of what you actually need to do is the same.
The reality of AI search as of August 2026
Let's look at objective figures for just how large AI-driven search and purchasing behavior actually is.
Looking purely at Google's scale, growth has genuinely been rapid. At Google I/O on May 19, 2026, Google announced that Google AI Mode had surpassed 1 billion monthly users and become available in more than 200 countries and regions. Japanese-language support began on September 9, 2025. The earlier AI Overviews has 2.5 billion monthly users, and it's now being integrated with AI Mode. It's a fact that the act of "searching" itself is shifting toward an AI-driven, summarized, conversational experience.
Source: Google Search's I/O 2026 updates (Google's official blog), Google Search's AI Mode launches in Japanese (Mynavi News)
On the other hand, what matters to EC merchants is "how far actual purchasing behavior has shifted toward AI." According to a survey published by ZETA INC. on June 30, 2026, when it comes to gathering information before an EC purchase, "on-site search within the EC store" (63.8%) and "search engines such as Google" (59.4%) remain the two dominant methods.
Meanwhile, only 10.0% of respondents said they use an AI agent such as ChatGPT or Gemini to gather information. That said, 35.1% said they use AI in some form to gather information, so once you include more supplementary uses, the level becomes too significant to ignore.
What these figures tell us is that we haven't yet reached the stage where "AI-driven purchasing has become mainstream" — we're at the stage where "search engines and on-site EC search remain the main battleground, and roughly one in three people now also uses AI alongside them." Pushing existing SEO and product-data upkeep to the back burner in order to focus solely on AI measures doesn't match the currently measured data.
Source: [The Frontline of the Purchasing Experience in the AI Era, 2026] (ZETA INC.)
Three measures Shopify stores should take right now
Based on the measured data, we can prioritize the measures Shopify stores should take.
Priority 1: The basics of existing SEO and building out product data
The main battleground for information-gathering is still search engines and on-site EC search. Building out basic SEO elements such as title tags, product descriptions, and alt attributes, along with first-party information such as a product's materials, sizes, ingredients, and how to use it, is the most cost-effective measure — it pays off for both Google Search and AI answers alike. Generative AI uses text on the web as its source material when learning and referencing information, so if your product pages are thin on information, there's simply no material for AI to cite in the first place. For the basics of SEO strategy, please see the article below.
Priority 2: Use Shopify's own official AI features
On March 24, 2026, Shopify rolled out a feature called "Agentic Storefronts" to all eligible stores. It's a mechanism that automatically distributes your product information to ChatGPT, Microsoft Copilot, Google AI Mode, and the Gemini app. Checkout is processed within Shopify's own infrastructure: in ChatGPT it completes on your own store's checkout, while in Google AI Mode, Gemini, and Copilot it completes via Shopify's checkout inside the AI channel when direct checkout is enabled — and in every case, order and payment data stays tied to your own store. The benefit is that you can secure exposure within AI services by riding on Shopify's standard functionality, without having to build a custom AI-facing implementation yourself. For more on "Agentic Storefronts," please see the article below.
Source: Shopify agentic storefronts (Shopify Help Center)
Shopify also offers official features such as Shopify Sidekick, which supports store operations tasks using AI, and prioritizing the use of these official features is likely to deliver a better return on investment than hand-rolling your own AI measures from scratch. For more on Shopify's AI feature "Sidekick," please see the article below.
Priority 3: Distinguish between bot types in robots.txt
One thing that's surprisingly easy to overlook is controlling crawlers via robots.txt. There are broadly two types of AI-related bots: bots like GPTBot and ClaudeBot that crawl to collect training data, and bots like OAI-SearchBot, Claude-SearchBot, and PerplexityBot that crawl to generate search results — and these serve different roles.
Wanting to block access from training bots is understandable in itself, but if that same setting also blanket-blocks search bots, you end up cutting off your own opportunities for exposure through AI search. When setting up robots.txt, we recommend checking that you haven't conflated these two types.
Fact-checking common claims: llms.txt and FAQ structured data
There's no shortage of talk about measures for AI search, so let's also sort out the ones that are said to be effective but whose actual impact is uncertain. This helps you avoid a situation where excessive effort only piles up operational cost.
llms.txt: no harm in adding it, but it doesn't reach Google, the main battleground
llms.txt is a text file that summarizes a site's content for AI, written in a format aimed at AI. You'll see articles recommending that you adopt it, but there's no backing from any official standards body. Major AI providers — OpenAI, Anthropic, and Perplexity — do not acknowledge llms.txt as a citation signal in their official documentation. Google, for its part, has officially stated that it doesn't reference it at all, comparable to the meta keywords tag (according to comments from Google's John Mueller). The measured data backs this up too: an Ahrefs survey of about 137,000 sites found that 97% of llms.txt files received zero access from AI bots.
Source: Google Says LLMs.Txt Comparable To Keywords Meta Tag (Search Engine Journal), llms.txt adoption rises 8.8x but 97% of files get zero AI requests (PPC Land)
The share of sites that have adopted it is also said to be only around 5-15% overall. That said, from April to May 2026 Shopify began automatically serving llms.txt as standard for all stores, so Shopify store owners essentially no longer need to set up the file themselves. Since the cost of adding one is low, there's no harm in doing so — but the realistic assessment is that it's a measure that has no effect on search engines, which remain the main battleground for EC purchasing behavior.
Source: LLMS.txt Adoption: 8.7% of the Top 1,000 (Rankability)
FAQ structured data
You may come across claims that implementing FAQ structured data (the FAQPage schema) doubles or triples citation rates, but this comes from a specific blog post, and the backing from primary data is weak.
What's more, on May 7, 2026, Google discontinued displaying FAQ rich results in search results altogether. The schema itself hasn't been invalidated, but even before this change, the display benefit of FAQ rich results for a typical website had already largely disappeared. In other words, the motivation for implementing FAQ structured data has effectively narrowed down to a single point: the hope of being cited in AI search.
Source: Google Drops FAQ Rich Results From Search (Search Engine Journal)
There's another operational issue: the Shopify blog editor has a specification where, if you save an article while in visual mode, script tags (including JSON-LD) get automatically stripped out — meaning your structured data risks disappearing every time you edit the article.
What we've learned from this is that you should evaluate not just "how effective a measure is" but also "how likely it is to break in operation." A measure whose effectiveness is uncertain and that also breaks on a regular basis in operation isn't a high priority for your limited operational resources.
Prioritize your measures for AI search
The four terms GEO, AIO, AEO, and LLMO differ only in which AI service they focus on and which country uses them — most of the actual work overlaps.
For your search strategy, prioritize your own store's measured data above all, and confirm that search engines and on-site EC search remain the main battleground while AI-driven purchasing is still a supplementary channel. Also keep in mind the principle of evaluating each measure on both "effectiveness" and "how resistant to breaking it is in operation," and deprioritizing measures whose effectiveness is uncertain and whose operational cost is high.
This information is current as of August 31, 2026. Pricing, scope of support, and specifications are subject to change, so please check each service's official information for the latest details.
