Greg Bernhardt is the senior SEO specialist at Shopify.

The launch of ChatGPT and other large language models brought a major shift in how consumers discover products. Today, generative AI tools are responsible for referring an estimated two billion site visits per month, creating a new challenge for marketers: How can your brand be included in an AI response? This is what answer engine optimization, or AEO, is all about.

In this article, I’ll explain what AEO is and why it matters for brands today, as well as look at some top strategies for how to get your brand and content discovered by large language models (LLMs).

What is AEO?

Answer engine optimization (AEO), also called generative answer optimization (GAO), 

involves strategies to get content featured by new technologies that use generative AI as part of search engine retrieval. This includes AI answer engines like ChatGPT, as well as search engine features like Google’s AI overview and hybrid search-answer platforms like Google’s AI Mode.

With AEO, there are two different ways to show up in an answer: 

1. The base generation coming from the training model (the answer itself) 

2. The web search component that has become part of many AI-generated answers (the citation list) 

Each requires different optimization techniques, which we’ll cover in depth below.

Why AEO matters

Right now, AI search engines represent about 2% to 3% of search engine usage. That doesn’t sound like much, but a couple of years ago, that number was zero. It’s actually astronomical, considering Google had almost complete market share for 20 years. 

When you factor in the AI features of traditional search engines, AI search may have captured 50% or more of search use. We don’t know for sure, because Google and other search engines combine traffic data from both traditional and AI results.

AEO vs. SEO

AEO and search engine optimization (SEO) are converging as GenAI engines incorporate more web search results, and traditional search engines push more AI-generated summaries. In the near future, it’s possible there won’t be much of a difference between AEO and SEO.

Here’s a breakdown of the differences between the two:

In general, the process of optimizing for AI answer engines is fairly similar to that of optimizing for traditional search engines, Google in particular. 

How to measure success with AEO

One of the biggest challenges with AEO is that success is hard to measure. If you’re used to traditional SEO, you can measure your position in search results, along with clicks from the SERP. But with AI answer engines, it’s a little more complicated. To measure success with AEO, you need to break AI-generated responses into two parts: the generative answer portion and the citations list.

Optimizing for generative answers

For the generative portion of an answer, there are two main metrics you can track:

• Mentions. Is your brand name mentioned in this response? This is a simple yes or no.

• Sentiment. If your brand name is mentioned, use sentiment analysis to determine if it’s positive or negative.

Perform a lexical or semantic analysis to go deeper: What keywords were closest to your brand mention? Are competitors mentioned? If so, are they mentioned positively or negatively? Where in the answer are you and your competitors mentioned? Are you at the top or the bottom of the answer?

Within the generative answer, a brand mention won’t lead directly to a click to your website. Answer engines aim to answer the user’s query within the platform, so the user doesn’t need to click on any links at all. To earn that click to your site, you need to be a cited source, and the user must visit your site to fulfill their search journey.

Optimizing for AEO citations

To generate organic search traffic from AEO, your website needs to be linked in an answer’s citations. Think of the citation list as being ranked, like traditional search engine results. Strategies for achieving this tend to be the same as those for ranking at the top of traditional SERPs. In fact, the sources cited by ChatGPT and other AI-powered answer engines tend to rank in the top positions in traditional Google search. That’s because ChatGPT now answers most queries by first searching Google, then generating an answer. 

It didn’t always work this way, though. Early on, when ChatGPT introduced the citation feature, it was through a partnership with Bing. ChatGPT would generate an answer to a user’s prompt based on its training, then verify its answer with a Bing search.

In July 2025, French SEO consultant Alexis Rylko discovered that ChatGPT had quietly ditched Bing in favor of Google search results, and its citations were even in the same order as the SERP.

Aside from tracking your position in a citation list, you can also measure success by looking at referral traffic. The good news is you can easily see how much traffic AI search engines are driving to your site. Just use Google Analytics, Shopify Analytics, or your preferred analytics tool and check the referring domain, such as chatgpt.com or perplexity.ai. The bad news is you won’t be able to tell what percentage of your traffic from Google is coming from its AI Overview or AI Mode. All Google search traffic is combined in reporting.

Top AEO strategies

• Optimize around fan-out queries
• Use brand marketing and PR to appear in training data
• Think about how you structure information
• Focus on unique value
• Consider information density
• Track the right prompts

There’s been a lot of discussion around whether optimizing for AI requires a whole new skill set. Right now, we’re seeing—due to all of these platforms scraping Google—that if you’re doing well in Google, you’re doing well in AI search. 

But there are a few strategies you can use to stay on top of AI-powered search:

Optimize for fan-out queries

When an answer engine like ChatGPT scrapes Google for search results, it doesn’t just pop your prompt into Google. It translates your prompt into a number of queries that “fan out” from your prompt.

If your goal is for your URL to show up in the citation list, you need to rank for those fan-out queries. But how do you know which fan-out queries to optimize for? 

It’s pretty technical to figure out the actual fan-out queries, but you can try to do it yourself by translating a prompt into five to 10 queries you would search for in Google. Take a look at your prompt and ask: What variations of this prompt would I ask to get the topical information I need in the answer? Something we see a lot are adding different qualifiers, like “best” or a temporal qualifier like the year. 

For example, “What are some dropshipping apps for Shopify?” could turn into “best dropshipping apps Shopify” and “Shopify dropshipping apps 2026.”

Use brand marketing and PR to appear in training data

Aside from appearing in the AI’s citation list, you may also want your brand to show up in the generative portion of an AI answer—that is, the part based on its training data. It can take about eight months for training data to update, so this is a long-term strategy.

The way to ensure your brand becomes part of AI training data is to get your message out there and get people talking about it. This comes down to classic brand marketing and public relations (PR). 

Large language models (LLM) are processing billions of documents. To decide what’s relevant, they look at the authority of the source (The New York Times is going to be looked at a little bit differently than a hobbyist blog) and consensus. If a model sees a fact once or twice, that learning probably won’t stick. If they see a fact many times across the web, then it becomes a learning experience.

Think about how you structure information

There is an age-old struggle in SEO around what’s best for the reader versus what’s best for search engines. It’s the same with AI-powered search engines.

Search engines have limitations. They’re built to parse. They take everything you have and break it into components, rather than reading top to bottom, like a human would. They’re programmed to understand explicit relationships like heading-paragraph and heading-list.

With ChatGPT and Google AI Overviews the citations are connected to a snippet of text. What this tells us is that if you want to be cited, you need to have a snippable, machine-digestable portion of text that’s relevant to a heading.

This doesn’t mean you should disregard the user experience and start structuring your content just for machines. Instead, try to find a balance between human narrative and machine digestibility. 

Focus on unique value

In the early days of SEO, everyone published an ultimate guide to XYZ and the content was all the same. The ones that were able to rise up and rank in search engine results pages (SERP) had unique or proprietary information, insights, or data.

AI search detects the same thing. Imagine an AI answer engine deciding which website to cite. Five websites could back up the AI’s claim, and they are all similar. If your content is 75% the same as your competitors’ but has 25% new material, the AI is more likely to cite your content because of the information gain.

Consider information density

AI search engines and crawlers want you to get to the point. For any given snippet, paragraph, or section, be concise and reduce verbosity. 

To have the best chance of appearing in AI results, you also need to prioritize information density. You’ll gain trust from content richness or density, like a date, statistic or entity (notable person, place, or thing). If you’re specific and get to your point sooner, you’ll have a higher chance of winning a citation.

This shouldn’t come at the expense of the humans reading your content. Focus on the sections where it makes sense to have a high density of information, like a definition or product description.

Track the right prompts

There are two different ways to track prompts for AEO. A number of third-party tools are available. Traditional SEO reporting tools like Ahrefs and Semrush have expanded their offerings to include AI prompt tracking, and there are newer tools like Profound. Each tool works slightly differently, but the idea is that you can input your website and the tool will come up with relevant prompts to track. These tools make prompt tracking easier, but they can be expensive.

If you have a small website and limited budget, you can start tracking manually. For each of your top URLs, come up with three to five prompts. When synthesizing prompts, don’t aim for perfection or exactness, because that’s impossible. If you focus on creating prompts that are concise, mimic human grammar and natural language, and dense with topicality and intent.

Create a spreadsheet with your prompts and check each prompt on ChatGPT daily, just by typing the prompt into the AI. Add fields for each dimension you want to track, such as brand mentions or citations. You can report with a simple yes or no for brand mentions and a number ranking for citations.

My recommendation is to start by prioritizing your top 20 highest-converting URLs, so you don’t get overwhelmed with data entry.

What’s challenging about AEO? 

Some of the most challenging aspects of AEO include:

Variety of prompts

The nice thing about Google and traditional SEO is that over the past 20 years, we developed a standard format of three to eight words per query. This made it fairly easy to track queries related to your niche.

With AI answer engines and natural language processing (NLP), the typical prompt is closer to 20 or 30 words long. To make it more complicated, two people can have the exact same intent but use completely different prompts. This makes it virtually impossible to know exactly which prompts to track. The best we can do is an approximation.

Variety of answers

With AI answer engines, you can ask the same question 10 times in a row and each time the AI will provide a slightly different result. We can’t work around that—it’s a consequence of the technology. This makes measurement challenging. The best thing to do is to look at the bigger picture and notice overall trends.

Training time

If your goal is to not just get cited by AI but actually become part of its training data, it can take a long time—at least eight months—to see results. With such a long delay, it can be hard to tell whether your efforts are working.

The future of AEO

AEO is a super-fast moving discipline. The technology itself is rapidly changing, and I would encourage anyone interested in how AI is changing search to take an hour each week to research and read about updates.

But you don’t need to panic and upend your current strategy. Keep doing what you’re doing, for the most part. No one is going to lose all of their traffic overnight because of AI search. A good place to start is with prompt tracking.

If you have an ecommerce store, keep an eye on specific AEO tactics for ecommerce and the available integrations. Instead of citations, you might be more focused on the in-app shopping experience. For example, Shopify has a partnership with OpenAI that makes it easier for users to purchase directly from Shopify merchants within ChatGPT.