View AI search as an opportunity to plug the gaps in your digital footprint

Crystal says: “Use AI search optimization to deliver value across your business and your whole digital footprint.”

What does the whole business look like these days?

“With the emergence of AI search optimization and LLMs as a channel that people are using, the importance of multi-channel marketing and having an omnichannel presence is very acute.

LLMs like ChatGPT, Perplexity, Claude, and Gemini are surfacing content from across the web. It's not just what's on your website; it's what's happening on Reddit, it's what's happening on YouTube, and it's what's happening on Instagram. That means that you need to think about the whole of your digital presence in order to move forward.

Some people say that it's the same as SEO, which is potentially true. Ideally, they should complement each other in the same way that YouTube optimization should complement your core SEO activities, but they also can help you identify gaps in the information about your business across the web.”

Do you have to research your competitors to see what platforms are most successful, and try to mirror that?

“I think it's important to look at the sources for these queries, and they will vary for the types of queries.

For instance, if somebody is looking up ‘best microwave’, a lot of LLMs will pull out review sites and take big lists of reviews from their core training data and citation sources. Therefore, if you haven't invested in getting your product out there and getting it reviewed by some of these key players, then you're probably not going to be visible there.

If you were trying to be top-of-mind for informational queries about data-heavy or heavy consideration topics, then you might need to shore up some of your data kudos and make sure that you've got good core information on your website. Additionally, if you have a larger brand, you might want to think about your entity and how that is represented on sites that have entity management, like Wikipedia, because that is a core training data location. That's really important.

When you're thinking about how to position yourself in all of these LLMs, it's not just about copying what other people are doing. It's also about making sure that you are present where users are going to be looking for you, and where the LLMs are going to be looking for information about you. You do want to be competitive, but it's important to be involved.

A classic example is Reddit. People have suddenly become very interested in Reddit now that LLMs are big players in search and information discovery. If you want to be involved in that, you have to be genuinely committed to that process. Even if you're just posting a few times, you do have to post, respond, and monitor that conversation, and you will get more action if you are more entrenched in those channels. You can't go about it half-heartedly; you have to get involved.

You could say that you’re trying to be there because your competitors are there, but you're also there because your users are there as well. Users are there asking questions, and ChatGPT is there looking for answers because users are there asking questions.

You should be there to satisfy the users who are there anyway, to satisfy search (because search is also looking at Reddit), and to satisfy the LLMs.”

What step do you need to take first?

“The first step is to have a look at how urgent this is for you.

LLMs and AI search agents are emerging as a source of information and discovery. This means that some people are on it and some people are not. On the AI Search Lab by Wix Studio, we have a great article from Rand Fishkin that talks about how to measure whether or not your audience is even using ChatGPT.

The last time I checked, ChatGPT and the other AI search agents had around 6 billion visits in August. That's a big number, but Google's pulling in 80-84 billion every month. Not everybody’s on it, so you should check that the market is actually with you.

If you look up ‘programming’ on SparkToro, for instance, you'll see that lots of people who are interested in programming are using ChatGPT as a regular tool. If you look up ‘potatoes’, then you won't see ChatGPT in the top seven results. It's not something that the potato people are particularly interested in.

Have a look at how fast you need to move. Have a look at where your audience is, so that you can see the total addressable market in that regard. Also, have a look at your data. We have an AI Visibility Overview tool at Wix, which is built into the CMS, and it allows you to see whether or not bots are coming to your site to look for answers from users. It allows you to see whether or not you're being surfaced in queries on ChatGPT, Claude, Perplexity, Gemini, etc.

With these tools, you can see whether you're visible in the first place, you're already doing pretty well, you need to get yourself together but not panic, or you're out of the running and there are already other people much further ahead of you. Have a look at the addressable market. Go into your GA4 and have a look at how many people are coming to your site from those tools.

Then, try querying the bots and doing the due diligence to fill in some of those gaps. If you have a pie shop and you sell blueberry pies, you can go on to ChatGPT and ask, ‘Does Crystal's Pie Shop sell blueberry pies?’ If it says, ‘No’, why does it think that? What is the issue that’s blocking it from understanding that information? Is it not listed on your website properly? Does the page say it’s out of stock? What's going on there?

It's important to understand why the chatbots are saying what they're saying, and you can do this in a couple of ways. One really good way is to go to Perplexity and use something called Perplexity Pages. We have an article on the AI Search Lab from Maddy Osman, which talks about ranking on Perplexity. One of the things you can do is create a page on Perplexity. It will spin up a page, and they'll pull up lots of resources.

This isn't the same as a query. With a query, you ask a question, and they will answer however they want. For Pages, if you were to put in something like ‘Crystal’s blueberry pies are fantastic,’ it will create a narrative around that, and you can see how they're justifying that claim. If you put in, ‘Crystal’s blueberry pies are terrible,’ then you would see how they're justifying that claim too.

It's a way to see gaps in the information that's available about you on the web, and that the tools are using to discover information about you. This is really important because, for some of these things, it can be tricky to change.

A lot of the tools use various types of information retrieval methods. There's training data retrieval where, if you were to ask for a recipe for a blueberry pie, they're not going to go on the web to look that up. From the training data that the LLM has, it can put together a blueberry pie recipe. It knows what blueberries are, and it knows what a pie is.

However, if you want to know what Crystal Carter's blueberry pie recipe is, it would go to a website to look that up. That's a RAG retrieval. If you were to ask a more nuanced question (like, ‘Is Crystal Carter's blueberry pie healthy for somebody who is recovering from a marathon and is lactose intolerant?’), that has lots of different caveats.

When you ask queries like that, it triggers something called a Mixture of Experts (MoE) response, which will basically pull in lots of different subject networks and portion your question into a bunch of different sections. It will look for some of it in their training data, some of it will be from the web, and then it will reason it all together to give you a response.

When you see that happen with your content, it gives you a lot of insights into how they're presenting your brand and how your brand is visible online, and it also helps you to identify where you can improve. It helps guide the kind of content that you should be creating.

For instance, if your goal is to be cited more often in LLMs, maybe you should have more of the kind of content that needs to be retrieved rather than the kind of content that they would have in their training data. If you have a generic blueberry pie recipe, that's not going to get you into an LLM. However, if you have a report or an opinion piece on why blueberry pies are the best pies, that might be something that LLMs would cite.

Depending on your goals, you need to look at your content to make sure that it aligns with that, and also look at your web presence to make sure that aligns as well.”

Do you also need to optimize for search behaviour that is likely in six months' time, to make sure that you're not left behind?

“You need to stay ready so you don't have to get ready.

If you're in the potato area, then you could potentially have a first-mover advantage as people become more attuned to these tools. We're starting to see more and more corporate teams encouraging their employees to do prompt training, for instance. As Google Workspace rolls out, more people are going to have Gemini in their workspace every day, and Copilot is becoming more integrated now as well.

People are going to be more used to being AI native, both at work and on their phones. The Google Pixel 10, for instance, has Gemini baked into it. They've added Gemini to smartwatches as well.

If people aren’t taking up AI in your sector just yet, then you have time to learn. You don't need to panic, but you do need to prepare. You should be reading, getting involved, testing, iterating, and tracking the traffic.

You also want to make sure that you are being efficient about it. Some people are saying that it's all or nothing. I don't think it's that at all. When you have those training data responses, you can specifically ask for a website or a product name, and it won't necessarily give you a link, so you have to go to Google to get the link. There's definitely a process where they overlap.

We're in the very early days, but if you are active and you're involved, then you're going to be able to move quicker when you need to.”

Crystal, what's the key takeaway from the tip you shared today?

“Get involved and assess your brand.

Don't look at the gaps that you see in AI as a problem, necessarily. Look at them as an opportunity to improve the way that your brand is communicated online, improve the way that your brand is visible online, build more connections with users, and build more bridges to your domain and your business.”

Crystal Carter is Head of AI Search and SEO Communications at Wix. Find out more over at Wix.com/studio/ai-search-lab.