Gain visibility in generative search by managing your brand entity

Crystal says: “Manage your brand entity for visibility in generative search engines and tools.”

Why does generative search visibility mean that you should rethink your brand entity?

“One of the things that’s really important to consider is the way that LLMs work and the way that people are using them on a regular basis.

Over the last year, search has become very dissipated and decentralised. Rather than just using Google or Bing, people are searching across multiple search engines – including LLMs and tools like ChatGPT, Perplexity, Claude, and Gemini – to find their answers.

Even within the search engines, if you use Bing, alongside the main search results you will see an option to use Copilot to dig deeper into the topic. Similarly, Google has finally rolled out AI overviews to the EU and wider markets, so you also see that influence. We might see Gemini being rolled into the search experience there.

When we look at some of the ways that people are accessing and discovering content in analytics, a lot of traffic is starting to come from ChatGPT and Perplexity. A few people are starting to track that traffic and trying to see how they can guide it to get more visibility.”

Is the average person searching with LLMs, or is it mainly technically focused people?

“ChatGPT continues to grow its user base. A lot of people within the marketing space are using it really regularly, but it’s becoming a go-to tool for a lot of folks.

Even beyond that, you’re starting to see people using tools like TikTok and YouTube for search as well. Search is not as centralised to the main players anymore. If you start managing this trend now, when it does become more mainstream, you will already be well-placed to perform in the long run.

These things also tend to be connected, like Gemini with Google and Copilot’s links with ChatGPT and Bing. If you’re able to think about them holistically, you’ll be well placed for the future, which is going to include even more generative search than we have now.”

How many people are clicking through to the sources when they’re doing a search on a tool like Perplexity?

“That research is yet to be seen. However, we are starting to see that a significant amount of traffic is coming through from some of these tools.

From the analytics that I have seen, Perplexity tends to provide the second most click-through, but it is unique in that it has links for almost every query. Even if you ask it to help you write a LinkedIn bio for David Bain’s work with Majestic, it will pull up a link for Majestic underneath that.

If it sees any entities within there, it will pull out some of the top links related to that entity. That is one of the reasons why I am talking about visibility rather than traffic because, within certain generative search tools, you don’t necessarily get links. However, if someone is investigating a particular brand or a topic that’s related to a particular brand, and you have good relationships with your entities, then you’re more likely to show within the content.

I will give you an example. When the Barbie movie came out, I put the movie description into a tool called TextRazor, which is an NLP tool that pulls out entities, relationships, words, things, etc. The description doesn’t actually mention dolls or Mattel but, when I put it into TextRazor, one of the entities it pulls out was Mattel and one was ‘fashion doll’. Then, it pulled out ‘Barbie’ and ‘Ken’.

Even though the text didn’t say that Barbie is a fashion doll that’s manufactured by Mattel, it understands that. I’ve run similar tests in ChatGPT and Perplexity, and asked which entities they can identify from that piece of text. Sometimes it pulls out Mattel, Barbie, toys, etc. straight away. Sometimes you have to ask it which latent entities it can see before it will pull them out, but it understands that those things are related.

Additionally, if you do the reverse and ask one of these tools to name 5 fashion dolls, it will name Barbie. It understands that the entity of Barbie is a fashion doll, and it understands that, if you want to find out about fashion dolls, you should know about Barbie.

While it’s very difficult for anyone to replicate the IP of Barbie, it is important to think about which entities are related to your brand and which searches you should see within those. What we see from generative searches, AI overviews, and Copilot is that they’re trying to speed up the way that these things reply.

While you don’t necessarily get the exact same reply every time, there’s a corpus of content and information that they’re pulling from. Sometimes, they don’t generate in real-time; they already have an answer, and they just spit it out. With Copilot, you’ll put in a question, and it will pull out the entities, shorten your question, and then answer it that way.

It’s important to think about which entities are most related to your brand and how you articulate that within your website, your brand’s corpus of content, and across the wider knowledge graph as well.”

What do you need to be doing as a brand to optimize your entity presence further for generative search?

“Invest in things that help make you more clear and more disambiguated within the knowledge graph. If, when people search for you, they don’t get you – you don’t have a knowledge panel and you don’t have a knowledge graph – then chances are you need to do some work there.

If you’re a large brand, think about what your Wikipedia is saying and how it aligns with your Wikidata information. Make sure that you have schema markup that is accurate and reflective of your brand, and a LinkedIn that’s pulling through and is accurate and related to your brand. Make sure you’ve got brand consistency all the way through.

If you’re a smaller brand, it’s worth thinking about how you can build the relationships in your link profile, to connect with some of the main entities that help to identify you.

For example, if I say, ‘Steve McQueen is a man known for his films’, that statement could describe one of two people: Steve McQueen the American actor from the 60s and Steve McQueen the British director. If I said, ‘Steve McQueen, Academy Award winner and Turner Prize-winning director’, that includes his name and two named entities that are associated with him, so you can verify who he is. That description completely disambiguates him from the other Steve McQueen.

I have seen businesses say that they’re not in ChatGPT because they tried searching for something like ‘10 marketing agencies in New York’ and they weren’t being shown. There are two questions there. The first is timing. A lot of people think that ChatGPT is live data, but it’s pre-trained. It’s essentially like the phone book.

For anybody who is not old enough to remember, once upon a time, people would deliver a big book that had everybody’s phone number and addresses in it. The way that you optimized for that phone book was you had your name start with an A, so taxi drivers were always Aardvark Taxis, AAA Taxis, or something to that effect. That way, they were first on the list. Anyway, the phone book didn’t come out every week or every month – it came out once a year.

What’s important to remember is that ChatGPT updates their models, and the gaps between those updates will probably decrease. It could be that we have a model that is updating more frequently so it’s not such a big gap. However, at the moment, they’re only up to 2023, the last time I checked. Therefore, if you launched something in 2024, you’re not going to be in ChatGPT yet because they haven’t printed the next round of the phone book.

You have to set your expectations for that. While you’re waiting, it’s worth looking at some of the ones that are live-scraping. If you use Copilot or Perplexity, they have a fixed dataset of information that they use, but they also augment it with live data. Copilot is using Bing search results to augment its results so that they can provide real-time information.

For instance, at some point, there were fake pictures going around that Zendaya was at the Oscars – but she wasn’t at the Oscars, she was hanging out with her boyfriend in London. I asked Copilot, ‘What did Zendaya wear to the Oscars?’ and it said that she didn’t go because they had the live data for that particular query.

If your business is showing up in Copilot or Perplexity, that’s good. When the next round of updates comes around, check and see where you are within the ChatGPT ecosystem and how that’s working.

The other thing that’s important in terms of why your business isn’t showing up on ChatGPT is that you need to think about how distinct your business is. How solid is your entity? How clearly is your content attributed to you, and how connected is it to things that are definitely within the model’s training set?

Wikipedia is definitely a basis for that, and things like Common Crawl are in that as well. You can see how often a website’s been crawled by Common Crawl, if you’re a complete nerd and you go into the Internet Archive’s Wayback Machine. You can also look within your own crawl bot logs to see how often your website is getting crawled by some of these tools.

For example, Perplexity has its own crawl tool, and you can see whether you’re being regularly crawled by those. If you are, that’s a good sign that you’re more likely to be shown in some of these tools.

It’s also worth thinking about the kinds of queries that people are using within these tools and how you can support them. People tend to put more long-winded queries into LLMs because they can. When you use Copilot, they have a microphone first, not the keyboard. You just talk into it and say, ‘Hey, I’ve got two cucumbers, half a block of cheese, and a potato. What can I make for dinner?’ These are the kinds of things that users are asking.

People who are used to almost exclusively making informational content are going to struggle in this space. However, if you are creating more unique content, that is deeper and requires users to sit with the content, you’re more likely to be referenced within these things, be unique within these things, and be the kind of things that people actually click into.

For complex topics, I might go to a tool like Perplexity, Copilot, or Claude to get the top-level. If I wanted to get into sewing, I might ask what kind of things I need to get started. However, if I were to ask how to make a particular thing and I kept seeing the same link to the same website coming up, the same expert being quoted, or the same technique being mentioned, then I’m more likely to go through to a website that has all of that in one place. It’s worth thinking about that too.”

If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?

“It pains me to say this but, if you are spending a lot of time optimizing videos on your website, you can chill with that a bit because Google has changed that traffic.

I was a big advocate for this in 2023, however, they changed the traffic that they’re sending from the SERP to websites in terms of video visibility. If you haven’t got the memo, the video tab is almost exclusively YouTube now – it’s not 100%, but it’s almost always YouTube.

If you are stretched for time, embed it if you want to, but don’t expect loads and loads of traffic from that. Add it in for user value rather than as a traffic-driver.”

Crystal Carter is Head of SEO Communications at Wix, and you can find her over at Wix.com/seo/learn/webinars.