# How to optimize for AI search: an answer engine optimization (AEO) primer **Published:** September 17, 2026 | **Updated:** September 17, 2026 | **Authors:** Mel Gutierrez | **Categories:** Best practices > How to optimize for AI search (AEO) is changing how AI recommends products. Get the schema and content tactics that earn citations. Read the guide. --- AI has been slowly impeding upon search for years. It’s nothing new. AI overviews officially rolled out in Spring of 2024, and it was in beta before that. However, the hot-button topic of AEO/GEO/LLMO/AIO/AI SEO continues to tread upward as more business leaders conflate it with search results. And the widespread use of LLMs has made search, once again, quite a wild west of a career choice. > **TL;DR:** Answer engine optimization (AEO) isn’t a new discipline. It’s SEO 101 (authoritative content, clean backlinks, first-party data, hygienic coding) applied to a new goal: getting cited by AI, not just ranked by Google or surfaced by Google’s AI Overviews. > For ecommerce, that means structured data (Product, Organization, FAQPage, HowTo schema) and an llms.txt file pointing AI tools to your best content, product feeds that are clean enough for machines to trust, and content that answers real customer questions instead of keyword stuffing headers. Track it through GSC, AI referral traffic in GA4, and use a dedicated visibility tool. And it’s bigger than a technical checklist; customer experience itself is becoming part of how you get discovered. I’ve been doing search engine optimization (SEO) since 2018, late enough to miss the keyword-stuffing era, but early enough to get a good grasp of what makes good content–at least good according to Google’s search results. As a career SEO, the one thing that marks my field is its continually changing parameters. Like language itself, SEO is constantly evolving. However, unlike language, this evolution isn’t natural, it’s driven by Google, which still has the lion’s share of search, and other search engines balancing what Internet users want against what makes them the largest profit. ## What is answer engine optimization (AEO)? Answer engine optimization, or AEO, is molding website content with the intent of that information getting picked up, distributed, and hopefully cited by large language models (LLMs). That header and first line were deliberately crafts for AEO, by the way. One of the best ways to get surfaced by AI overviews, or LLMs like Claude and ChatGPT, is to make your content “chunkable.” If an answer can be simple and succinct, make it so. But what makes it different from SEO? Why are businesses suddenly focused on AEO/GEO instead of SEO? The answer: it’s in the zeitgeist. A year or so ago, my supervisors wanted me to go all in on AEO, and the more I researched and the more I attended webinars on the topic, the more I realized AEO is just SEO on steroids. If you ask most SEO practitioners what makes a website more likely to get cited, you’ll probably get these answers: - Credible and authoritative backlinks - Long-form content written by someone who’s authoritative on the topic ([EEAT](https://developers.google.com/search/docs/fundamentals/creating-helpful-content)) - First-party data If you can focus and improve on those three things, you’re doing well in SEO. And you’re already doing AIO. ## How AI search and assistants surface and summarize product information Now that I’ve explained what your in-house SEO or agency has also tried to explain, let’s get to the meat and potatoes of why you care. What does AEO (or SEO+) have to do with your ecommerce site? According to Adobe, retail traffic from LLM’s is up [393% year over year](https://business.adobe.com/blog/ai-traffic-surge-retail-sites-not-machine-readable). At the same time, ecommerce site traffic, and most other industry site traffic, has likely fallen. This is the result of two things: - More people relying on LLMs for product searches - Questions posed in search engines are being answered by AI Overview, making clicking into a website mostly unnecessary This trend means that people who might be searching for something your ecommerce site provides are getting information about *your* brand and *your* product through a complete third-party means. LLMs pull information on products by using the data you provide on your ecommerce site and by pulling referenced information from sources like Reddit, YouTube, and other sources of product reviews. If your site doesn’t provide sufficient information in a way an LLM can access, it’s going to rely more on third-party sources to provide that data to potential customers. Again, one of the most crucial elements of local SEO is consistency of information. Your brand and product should be described exactly the same way everywhere you can possibly manage, because LLMs are going to be pulling from all possible sources to provide an answer that will satisfy its user. For more information on [how LLMs work](/glossary/large-language-model/), check out our glossary entry on the topic. One more thing to remember, AI search and LLM shopping assistance doesn’t just end at summarizing product information. It’s starting to help shoppers complete the purchase too. Agentic checkout is allowing LLMs to process the purchase of ecommerce products. For more information about how this is changing the ecommerce landscape, [check out our other blog](/blog/seo-to-aao-ai-powered-cx-discovery-channel/) on the topic. ## Structured data, schema markup, and machine-readable content So, the million dollar question is now, how do you make sure the LLMs are pulling what you want to pull? You use another tried and tested SEO tool, structured data. Technical SEO involves ensuring that the backend of your website is readable and well-structured for crawl bots, which includes making sure each page has the appropriate structured data. Structured data is formatting code in a way that provides clear facts to machines that might be reading a web page, like a search engine crawler. The code itself is known as schema markup. You can use [Schema.org](http://schema.org) to find out what kinds of schemas are available. For example, this blog post has Article schema on the backend, and any crawler that visits will know that the title of the blog is “How to optimize for AI search: an answer engine optimization (AEO) primer” by the author “Mel Gutierrez”, and so on. For LLMs to have an easier time describing your product, your product pages should all have [Product schema](https://schema.org/Product). Your larger website should have [Organization schema](https://schema.org/Organization). And like any good SEO will tell you, your website architecture (how your website is structured) should make sense. If you have a product parent page /product/, all child page products should be nested under it rather than what we call “orphaned,” which are free-floating pages with no referring parent page. The more isolated a page and the harder it is to find, the less likely it’s going to be crawled by a search engine or an LLM. Machines are smarter than they used to be, but that doesn’t mean it hurts to make finding your products easier on them. ## The schema types built for direct-answer extraction ### FAQPage schema and HowTo schema Two of the most commonly used types of schema that are often lifted directly into AI Overviews is [FAQPage](https://schema.org/FAQPage) schema and [HowTo](https://schema.org/HowTo) schema. FAQPage schema is one of the most powerful tools you have as an ecommerce brand if you apply it not only to your FAQ page but to individual products. For example, digging into your CX data, if you notice one specific product always gets the same kind of questions about it, you can add a small FAQ section to that product listing. If someone asks the question in Google or with an LLM, both AIs can pull from your neatly packaged FAQPage schema to answer that question. Similarly, HowTo schema is a great way to get cited by LLMs if your brand is an expert in, it’s in the name, how to do something. If you’re a shoe brand, you probably have a blog on how to choose the right shoe for your activity. That blog is where you’d implement the HowTo schema. If a potential customer asks Google “How do I choose the right kind of running shoe”, and the AI Overview has the option of choosing between your blog and your competitor’s blog, the blog with the HowTo schema is more likely to get surfaced. If you want your customers to be able to ask your brand about product questions directly, you’d want to implement a kind of chatbot designed to answer those questions. But what if that chatbot also had the context of who that customer is if they came back to your site? One AI chat capable of pre-sale and post-sale responses? To learn more about that, check out how Gladly can put [conversational commerce](/conversational-ai-for-sales/) to work on your website. ### What is an llms.txt file? You’ve heard of a robots.txt file. It’s a bunch of code that tells crawlers what parts of the site they’re are allowed to see and which they shouldn’t visit. For example, you want search engines to access all of your available products, but you don’t want them surfacing your hidden seasonal special pages out of season. A llms.txt is similar, but built for LLMS. It points AI tools directly to your highest-value pages and gives it the context-rich content you want it to see. While there’s no evidence llms.txt files have any impact on search visibility, it doesn’t hurt. Providing a machine-readable summary of what your website is, who your brand is, and what you offer allows you to have more control over your branding than what a standard crawler can see through the button and navigation coding on a standard page. When making your LLMS.txt file, make a list of the highest-value pages you’d want an LLM to see, such as your homepage, key category pages, and your best authoritative content (such as an About Us page). ## Content clarity: answering questions directly vs. keyword stuffing Another way AEO doesn’t differ from SEO is avoiding keyword stuffing. Keyword stuffing is an old SEO technique (so old it was used by Aristotle old) that involved putting a specific keyword in the title and on a web page as many times as possible, even if it read terribly. Google’s Panda update in 2011 fixed this issue by lowering the rank of low-quality, thin content or content that used this technique to game the system. AEO is similar. Keyword stuffing won’t help your case. If you’re selling high heel shoes, you can put “high heel shoes” all over a product page, and it won’t be more likely to get cited by an LLM. However, a product page with information like available sizing, materials, and color? That’s where the sweet spot is. People asking LLMs for product information are looking for, usually, something specific. “My prom dress is pink, and I need to find a comfortable pair of flats. Here’s a picture. Can you find me something within a $50 budget?” or “I’m going on a fishing trip in Southern California. What kind of brown jacket can I buy that works for warm weather during the day and cool weather in the evening?” SEO and AEO both require that you provide relevant and clear information so that people searching for your product can find it easily, even if they’re relying on an LLM to do that search. ### How to find the real questions Now that I’ve lectured you about keyword stuffing, here’s where you need to look to find what information people are looking for: - Check out the People Also Ask section in Google. This section on Google takes your query and suggests questions related to it that people have asked Google before. - GSC query mining: Use your GSC data to find out what queries people are searching for that are surfacing your pages. There may only be one or two impressions for these, but if one person has that question, you can bet that more might. ## Product feed and catalog hygiene as an AEO input Making sure your potential customers find you on Google or via LLM involves site hygiene. This process tends to be more time intensive with larger ecommerce sites, but it’s vital to ensure a buyer isn’t getting outdated information that results in a poor user experience. An updated product feed is nothing new, but clean and updated structured data on your products is what LLMs pull to answer its users' questions. Clean and consistent titles and descriptions on the back end make it easier for LLMs to match with what a searcher is looking for, and crawlability of the page itself is part of catalog hygiene. Is your schema JS-rendered only? It shouldn’t be. Will your robots.txt file block any new AI user-agents? These are the kinds of questions your SEO should be surfacing, answering, and fixing with your dev if there’s anything amiss. And speaking of files bots read before they read anything else: this is exactly where your llms.txt file (the one we talked about a few sections back) is supposed to be doing its job. If your product feed is spotless but you never told the LLMs where your good stuff lives, you've done the hygiene work and skipped the invitation. ## Measuring visibility in AI-generated answers (what’s possible today) One of the hardest things to measure is AI-visibility, since we, as users, don’t have access to the data Anthropic or OpenAI do. We don’t know how many people deliberated on framing a question to Google search versus just asking their instance of Claude or ChatGPT. What we can do is use Google Search Console to estimate. Google Search Console is a tool through Google that lets you, for the most part, see what queries people are searching Google for, how many people see your site in search results for that keyword, where on Google it appears, and how many people clicked on that result. For example, let’s say you’re a clothing brand that just put out a blog last month on how to select the best Fall sweater. Using GSC, you should be able to input that URL, adjust the timeline to the last 28 days, and find out how many people saw your blog for the query “fall sweaters.” Two years ago, before AI Overview, a good CTR might be 0.7%, where for every 100 impressions, 7 people clicked. Now, that number might look more like 0.3%. The impressions are still 100, but more people are seeing the answer for “what are fall sweaters” in AI Overview and aren’t clicking. High impressions and low clicks used to indicate that your meta title and description might need some work. However, if you have high impressions, lower clicks, and your average position is relatively good (position 1-10), AI Overviews is probably stealing your lunch. Another way to track LLM success is to start tracking LLM-referral traffic in Google Analytics. Referral traffic means someone on the Internet referred your site to someone, and that someone visited your site. AI referral traffic means one of the many LLMs cited your website, and its user visited your website. The higher that number goes, the more likely it is that your site is getting cited by more LLMs, especially if the referral sources are diverse. ### AEO tech Another way to track your AI visibility is to use a dedicated tool. You can’t swing a cat without hitting an AI visibility tool, so you have your pick and most tend to be relatively reasonably priced. Gladly currently uses [Scrunch](https://scrunch.com/) to track AI visibility. It requires that you know your industry well enough to create the hundreds of prompts possible and track whether any of the most prominent LLMs on the market are surfacing your content, citing your website, or surfacing and citing your competitors. ## Building an AEO workflow alongside existing SEO practices Here’s where AEO starts to differentiate itself. SEO on the website alone isn’t enough to create more traffic for your website. AI Overviews is eating your lunch. So are LLMs. What you do have control over is how you present your website to those sources. Solid technical SEO is the foundation for strong website presence, but in addition to that, you need to start, if you haven’t already, having a strong social media presence. LLMs now site Reddit and YouTube if a website doesn’t have enough information, or if a user is looking specifically for what people *think* about a product. The average user is extremely skeptical these days, especially since the economy, while not in a recession, is K-shaped, meaning the average consumer is more likely to save then spend as the stock market is performing well but the job market is not. The average consumer isn’t going to believe your on-site reviews. You control your website, which means you control how many 5-star reviews you show and how many 1-star reviews you hide. Listicles are often sponsored, so many “best x” searches have to be taken with a grain of salt. People are using social media as a direct communication tool with people who have actually used your product and want to know how they felt about it. If you’re actively on social media, responding to comments and questions about your product, you’re providing potential customers and LLMs with more branded information that you control. Be consistent across all channels, whether it comes from tone, branding, or post timing. In addition, try to get your happy customers to post positive reviews on third-party review websites. Start a referral program. SEO involves backlink growth opportunities, so pair it with offsite do follow links from authoritative affiliates you trust. Lastly, make sure you constantly audit your website for opportunities to improve your technical SEO. Add product schema. Be descriptive. Have a good keyword on the page, and *in addition* to that give your future customers (and LLMs) as much information as you can about your product so that LLMs have the context they need to surface it to their users. Here's the bigger picture hiding underneath all of this tactical stuff: support and content used to live in different departments with different KPIs. Now they're feeding the same machine. Every review you earn, every FAQ you write, every support conversation you resolve well is a data point some LLM might use to decide whether to recommend you. We get into that argument in more depth in [our other post](/blog/seo-to-aao-ai-powered-cx-discovery-channel/) on why purpose-built CX is becoming the new discovery channel. Worth the read if you want the strategic "why" behind everything I just told you to go do. --- *This content is provided by Gladly. Visit [gladly.com](https://gladly.com) for more information.*