BrightEdge introduces AI Hyper Cube to help brands understand AI search. Here’s how it works

Every brand wants to know how it shows up in AI-based search results and how it can influence those results, but figuring it out is no straightforward task when the large language models (LLMs) are basically black boxes. Enterprise search engine SEO provider BrightEdge says its new AI Hyper Cube is the answer.

BrightEdge was created in 2007 as a Search Engine Optimization and Search Intelligence company. It built the Data Cube to help companies understand how they appear in search results and what actions they can take to improve their digital experience. But today, there are multiple types of search engines and AI engines, in addition to traditional search, used in the customer journey, and companies need to understand their impact as well.

This led BrightEdge to create a whole new layer of technology that helps companies optimize for how their brands are discovered in search. It’s called the AI Hyper Cube.

When the company started testing AI search, it found that AI handles many different conversations and questions. So, when you think about the customer journey, CEO Jim Yu explains, there is an entire universe of prompts that a marketer needs to consider and target. AI also doesn’t invent information (in a search, at least); it finds and cites it. The question is, where does it find that information?

According to Yu, AI would have an opinion (something traditional search doesn’t), and that means sentiment is a very important part of understanding what AI is saying about a brand.

These are the questions the AI Hyper Cube helps answer. Yu argues that if you know these three things (the prompts, where the information comes from, and the sentiment), a brand can learn which sources affect the specific areas of AI conversations it cares about, and what to change.

How exactly does the AI Hyper Cube work? Because the Large Language Models (LLMs) don’t share prompt information. They are essentially black boxes.

How the AI Hyper Cube works

BrightEdge works with companies across all industries and has built a corpus of information on topics and journeys by assembling data on what people are searching for and what drives demand. The company combines this data with hundreds of millions of prompts happening in AI engines to get a comprehensive view of the entire customer journey.

From there, it indexes and understands what the AI returns when a certain prompt is given. By decomposing what the AI is saying, it gets the entities, brands, and sentiment. All of this is then indexed.

A brand can ask the AI Hyper Cube questions about itself, such as where it is mentioned, what competitors are mentioned, and where positive (and negative) things are said about the brand or its competitors. The Hyper Cube will provide the different sources.

It’s not just about finding where a brand is mentioned, and what’s being said about it, though, Yu said. The Hyper Cube also helps brands understand what to do next. For example, it will share the actual video, creator, or source of information the AI is using, so the brand can take action.

But how does BrightEdge know the prompts? Great question. It doesn’t know, with 100% certainty, the exact prompts entered into AI search engines. But it is able to use all the information it collects to make a very educated guess.

To understand what the prompts are, the Hyper Cube learns which web pages and content AI agents visit, and it stitches this information together with SEO data and other sources (which Yu did not share) to understand how everything fits together. Yu pitches that they can even see the differences between what each AI-powered search engine is doing (e.g., Gemini versus ChatGPT).

BrightEdge has also introduced new Agent Insights technology that examines what AI agents are doing on websites. For example, ChatGPT sends out a user agent in real-time on behalf of a user entering a query (prompt) to search websites.

A lot of Marketers underestimate the amount of traffic from agents, Yu said. He explained that agents are a major signal for understanding AI activity (and there are many different types of agents), sharing that agent traffic is about one-third the size of organic search traffic. But you don’t see Agent analytics in traditional web analytics because agents don’t render JavaScript. Instead, it appears in the web server's log files. Understanding what the AI agents are looking at on your website is key to understanding the prompts.

Taking a holistic view of SEO and AEO

Is Answer Engine Optimization (AEO) taking over from SEO? No. Both are equally important, and brands need to take a more holistic view of the two. Yu says:

You definitely want to combine and look at it holistically, because the way the AI gets its information is by searching. If you think about the AI, what it's doing is it has base information that it's trained on, and when you ask it a question, it knows what information it already has in its model. And then it goes out and searches for things. So it does this thing called a query fan out, where it comes up with a plan of all the things it's going to search for, and it goes and literally searches for that information and then summarizes the information from the search in constructing that response to the user.

It’s very important that marketers continue to optimize for the questions users are searching for, using AI. Yu said good AI optimization starts with good search optimization.

Yu walked me through his company’s process for onboarding a new client. He said they start with looking at the landscape in the Hyper Cube to show the client where they already have authority in AI and search. This helps them see easy opportunities to build on.

Because BrightEdge looks at both traditional search and AI search, it generates an optimization plan that supports both. As it generates recommendations, the Hyper Cube will show what a brand needs to improve about its web pages, website, and new content to create. Plus, it looks at other sources the brand needs to influence. Yu explains: 

If you think about all the other parts of marketing, like your social outreach, like your video programs, like your PR programs, how do you integrate all that? Because those are where we look at both. From a search perspective, we look at it from a link perspective; from an AEO perspective, we look at it from an influence perspective; as a citation source for the AI. So what are the components of what people do as they start and then optimize with BrightEdge? And from there, they can see the before-and-after in the reporting. So here's our baseline as we did these actions and optimizations. Here are the results of the optimization. They can use our platform to execute that whole program.

Most of the work a brand does to support SEO and AEO is the same, like schema tags, but there are some differences to note, he adds: 

I’ll give you a very concrete example. Search engines don't need a separate information source, right? So they crawl your web page. They're very good at it. They can interpret all the code. They can strip out the things that they don’t need. They can just ingest HTML and JavaScript. But the AI engines, and it's not just ChatGPT, but all AI engines, they don't have all that web infrastructure. So for them, new formats that are more for agents do make some sense.

Yu says the AI Hyper Cube is not a black box. He said it’s not about the methodology but about what it reveals, showing brands an entire universe they don’t know about. He shares that, in their research, they found that Google’s AI is negative 44% of the time, more than ChatGPT. It's also the case that Google’s AI is used earlier in the funnel, and ChatGPT is more for research and consideration journey stages. Hyper Cube shows the differences across multiple AI search engines, so a brand can see the entire customer journey, says Yu:

I was just talking to a brand the other day. They had no clue that Google's AI pulled up essentially bad publicity from a couple of years ago for some terms around their brand. So when people are on ChatGPT, they didn't realize some old product reviews are showing up in prompts when people are shopping and comparing. And this is where ChatGPT often shows more negative sentiment. And so it [the AI Hyper Cube) surfaces these very important things that are in that AI journey that they just wouldn't even know about.

How fast can a brand fix an issue the Hyper Cube reveals? AI engines like fresh content, and they assign higher authority to a brand’s own information. However, some things are harder to influence, and bad publicity is among them. Before AI search, Yu said you could control bad publicity through your channel, but now the Hyper Cube shows that pages traditionally buried in search results are being pulled to the front and cited.

Another revelation from BrightEdge's Hyper Cube: more than ever, the entire customer journey, including service and support, is moving toward AI search. It has become essential that a brand think carefully about the information it provides to training agents to teach them about the brand.

How does a brand ensure the most recent, relevant, and up-to-date information is available for customers? Yu said they need more than a traditional content management system infrastructure. He said schema makes a difference, as does using markdown, and BrightEdge’s technology helps brands figure it all out. 

My take

Answer Engine Optimization, much like search engine optimization, feels very much like an art than a science. BrightEdge does seem to bring more science to the equation, but there’s still so much that’s hidden from view and hard to figure out.

It took me a bit to understand how the AI Hyper Cube works, but what I finally came to understand is that when you have a lot of data from Search Engine Optimization, mapped together to analysis of agent activity on the website, and other data (which is likely more proprietary to BrightEdge’s underlying models), you can make guesses backed by data.

Once you start adopting recommendations, implementing changes, and measuring outcomes, you will see if the Hyper Cube works. Am I completely sold on this approach? It makes sense how it works, but without understanding the full datasets and how they are analyzed, I remain optimistically skeptical.