Keyword & Prompt Research for AI SEO (AEO) | 2.2. AEO Course by Ahrefs

Hey, it's Ammo and welcome to the second lesson which is on keyword and prompt research for AEO. Now, in the last lesson, I walked you through how to run a brand gap analysis so you should have a clear picture of where your brand is showing up and where it isn't. Now, it's time to find the keywords and prompts to actually go after. And in this lesson, I'm going to walk you through the full keyword research process for both SEO and AEO. Why? Because there's a ton of overlap between the two. And half the battle is really identifying which is which and knowing how to approach them differently. So the first step is to build your keyword list. And this part honestly hasn't changed that much. You still need two things. Seed keywords, which are broad terms related to your niche, and modifiers, which are add-ons like best or howto that turn those seeds into real searches. Now, a quick way to come up with these is to just ask your AI assistant of choice. Something like, "I'm doing keyword research for my type of site, which makes money through my revenue model. My target audience is this group. Give me 10 seed keywords that are 1 to two words max, and five plus modifiers that will help me surface appropriate content formats I can use in my keyword research. The seeds and modifiers should not share the same words. And just like that, you've got a solid list of seeds and modifiers to work with. But these are just the starting point. Take your seeds and paste them into Keywords Explorer. Then go to the matching terms report and add your modifiers using the include filter. And just like that, you should have hundreds, maybe even thousands of real keyword ideas your audience is actually typing into Google. Now, the second step is to vet those keywords because some of them are a trap. Some of these keywords are going to look incredibly enticing. High volume, high traffic potential, low difficulty scores, all the right metrics, but not all of them are worth targeting. Before you commit to any keyword, it needs to pass three tests. I call this the bid formula. B is for business potential. Ask yourself, if I rank number one for this keyword, does it actually help my business? A keyword like what is espresso has solid volume and low difficulty, but someone searching that isn't looking to buy anything, maybe ever. Compare that to best espresso machine under $500 where the searcher is showing intent and has a budget. Always choose keywords that move the needle. I is for intent. Google the keyword and look at what's actually ranking. If every top result is an e-commerce page and you're trying to rank a blog post, it's not going to happen. The SER tells you what searchers want. Match the intent or move on. And D is for difficulty. You need to choose keywords you actually have a chance at ranking for. Check the referring domains and domain rating of the top ranking pages. In general, the more links and the higher the DR, the tougher the competition. If you see a few low DR sites in the top 10, that's usually a good sign. If a keyword passes all three tests, you should consider targeting it. Unless, and this is where step three comes in, the AI filter. Before you commit to any keyword, there's one more question you need to ask. Can AI fully satisfy the user for this query? Because even if a keyword passes bid, if the AI overview is so good, there's no reason for anyone to click through, that keyword might be a trap. And there's hard data behind this. AI overviews appear on about 21% of all keywords, but for informational queries, it's much higher. Nearly 58% for question queries, 46% for queries with seven or more words, and 99.9% of keywords that trigger AI overviews are informational in intent. So before you commit to a keyword, Google it. Look at what shows up. Put yourself in the searcher shoe and ask, "Am I satisfied with this answer or do I need to click somewhere to learn more?" If the AI overview nails it, that keyword might not be worth targeting the traditional way. But here's the good news. Even though AI is eating clicks for a ton offormational keywords, there's a whole category of searches AI hasn't touched. Free tools, search backlink checker, no AI overview, mortgage calculator, nothing. Word counter, nothing. Why? Because when someone searches for a tool, they need to actually use something. AI can't replace that yet. To find these opportunities, go back to keywords explorer with your broad seeds. Head to the matching terms report and add modifiers like calculator, checker, generator, tool, template, finder, planner, and maker. These are all actionoriented queries where someone needs to do something. AI can't satisfy that yet. So the organic click is still there for the taking. You can also filter for transactional intent directly. Just choose transactional in the intent filter and you'll get queries where people are looking to buy, sign up or take some kind of action. Now, the fourth step is to find your AI mention opportunities. A minute ago, I said if the AI overview nails the answer, that keyword might not be worth targeting the traditional way, but that doesn't mean you ignore it. These keywords just need to be targeted differently. And that's where AEO comes in. Instead of trying to rank and earn a click, your goal is to get your brand mentioned in the AI response itself. And to do that, you need to know which queries to focus on and which pages AI is pulling from. Now, it helps to know what AI is actually citing. We studied this across AI overviews and chat GPT and found that 43.8% of all cited pages are listicles. And in my opinion, these pages show up so often because they help AI build consensus. So, if your brand is mentioned across multiple lists, that's multiple sources recommending you and AI picks up on that. So to find the queries that matter for your brand, go to brand radar and enter your website. Find your brand, hover over the AI platform you want to research and click others only. This shows you the mention gaps where competitors are showing up but you're not. Then filter for queries containing best, top, versus, review, or alternative. And you'll see the queries where AI is mentioning your competitors but not you. and the ones AI is most likely pulling from when looking for brands in your space. This is your short list. Now, one thing to keep in mind, over 45% of citations change when AI overviews refresh, and that happens on average every 2 days. So, this isn't a one-time exercise. It's worth revisiting this report regularly to catch new queries as they come in. And that brings us to the fifth step, prompt research. When someone opens Chat GBT, Google AI mode or Perplexity, they're not typing keywords. They're having conversations. They're saying things like, "I'm a small agency owner looking for a marketing platform. Which one should I choose?" Or, "What's the best way to track rankings if I'm just getting started?" They use natural language with full context. And so, every person phrases it differently. So, you can't approach this the way you would keyword research, where you find the exact terms and go after them one by one. With AI, the same question asked 10 different ways can get 10 different answers with 10 different brands mentioned. And on top of that, AI fans each prompt out into many sub queries behind the scenes, most of which have zero search volume and will never repeat. If you're invisible for a topic, it's not about optimizing for one specific prompt. It's about building visibility across that entire topic, which is exactly what we'll cover in module three. So, with your topic list and your prompt gaps mapped out, let's go and close those gaps. I'll see you in the next lesson.