Beyond traditional SEO: Staying visible in an AI landscape

It feels like everything has gone from one to a hundred with artificial intelligence (AI) since ChatGPT was introduced in 2022. It genuinely seems like a new AI product or service is launching every week these days, making it really difficult to keep up with the latest trends.

AI has transformed how people interact with our content and brands. Gone are the days of relying on traditional metrics to understand how our business is performing online, since AI doesn’t conform to the rules we’ve been accustomed to.

Let’s be honest, that can be quite overwhelming, especially for businesses that aren’t geeky like us. So, to make your life a bit easier, we’ve pulled together this handy guide on AI. It should give you an understanding of how these AI models work and what you should focus on to maximise your visibility. We’ve also covered off some common myths and challenges that are worth remembering for the future.

This article at a glance

• Where does AI get its information from?
• When citing answers, what are models looking for?
• How do you plan content with AI in mind?
• Things to keep in mind with AI

Where does AI get its information from?

If you’ve ever used an AI platform like Google Gemini or ChatGPT, then you may have seen how they will regularly cite different websites and brands in their responses.

To understand how you get AI to mention your brand and products, we first need to understand what they’re looking for.

AI models are remarkably complex, but organically, there are only two mechanisms for getting cited or mentioned in their answers.

1. Training data

This typically appears in the form of a mention (no links). This means your brand is included in the original training data for AI, which usually happens through popularity (i.e. you’re a recognised brand within your industry and appeared in several sources during the training collection phase).

2. Grounding

Since AI models are simply complex predictive text models, they use formulas to determine the probability of the next words/phrases used in their responses. 

Platforms like ChatGPT will always do their best to answer your question using the training data they’ve already collected, but if your prompt is time-sensitive or requires in-depth fact-finding, then these models will usually launch their own web search to confirm their answers.

To do this, AI models have APIs linked directly to search engines like Google and Bing (think of it like a bridge between the AI model and traditional search engines). This is where you’ll typically see third-party websites referenced in the response.

When citing answers, what are AI models looking for?

You may be surprised to learn that AI models look for similar things to traditional search engines when they look for information to verify their responses with:

1. Recency bias

Several case studies, including this one from Seer, have shown that AI looks for fresh content to refer to. The theory is simple, really: if the content is recent, then it has a higher chance of being the latest (and most accurate) information. Therefore, having date stamps on relevant pages (i.e. blogs) will have a higher chance of being cited against those that don’t.

2. Third-party validation

AI wants to ensure it’s using the most credible sources when looking for information to use in its responses, and a sure-fire way of achieving this is to look for websites with a strong domain authority. Backlinks and brand mentions are your friend here, especially if they’re formed as part of a positive sentiment (i.e. they have something nice to say).

3. Semantic understanding

One of the challenges with tracking AI is how users interact with the platform compared to search engines. 

For the most part, users take a more conversational approach with AI by adding context about their scenario and interacting with the platform as if they were talking to another person. The issue this creates is that you could have 10 people looking for the same answer, but presenting their question from 10 different perspectives.

It’s really worth keeping this in mind since growing website visibility becomes less about specific keyword targeting and more about producing quality content that aligns with your audience’s intentions.

To try and put that more simply, AI models will look for content on pages that align with the intention of their users’ questions. They’ll use semantic understanding to help them identify the information they need, which can include things like:

• Short and succinct URL paths
• Correctly approached heading structures
• Data markup (schema)
• Formatted content that’s easy to understand (e.g. bullet points)

4. Original content, research and findings

Although this already applies to traditional search engines, we should always aim to thoroughly answer the questions your target audience has around specific topics with original content. 

We can’t emphasise the word ‘original’ enough in this instance. If you think about how much generic content can be created by AI these days, any brand that uses it for content creation is actually pulling information from the same pool as everyone else. In essence, you have nothing original to say, so why would AI bother giving your content a mention if it already has the answer?

To overcome this, think about how you can produce content with original findings or how you can actively demonstrate first-hand experience. Google actually covers this type of thing with their E-E-A-T content guidelines, which are well worth a read since the same principles apply to AI.

How do you plan content with AI in mind?

If you want to keep AI at the forefront of your mind when planning or creating content, here are a few things to remember.

AI prompts are popping up across the buying journey

There used to be a time when AI models were mainly centred around informational prompts and queries, meaning they were more focused on answering general questions as opposed to getting directly involved with the shopping experience.

Nowadays, they’re getting more involved with the commercial side of things. In fact, this research from Peec AI found that AI Overview (the AI-generated summaries you see at the top of Google) appears for commercially driven prompts over 90% of the time.

So, if you run an e-commerce business, it’s essential that your website and its product pages are well-structured, follow best practices, and are fully accessible to both crawlers and human beings. 

Listicle content currently carries significant weight

AI models are quite lazy when searching for information to use in their cited responses, which kind of makes sense since fetching content isn’t cheap for AI companies! They value accessible information that’s well-formatted and easy to process.

Listicle formats have proved very popular among cited responses. In fact, one study by Search Engine Land found that just under 22% of cited responses from AI models were in a listicle format.

This doesn’t mean you should look to spam the listicle format, but the key takeaway here is that structure and clear formatting = AI preference.

Digital PR is essential for AI visibility

Brand is a strong signal for AI models (and traditional search engines), which look to websites with strong authority within their industry for training and citations.

This is why we’d recommend leveraging any first-party data you have to launch Digital PR (DPR) campaigns, with the aim of building backlinks (a ranking signal for search engines), which, in turn, will harvest brand mentions (a signal for AI models).

What do we mean by leveraging first-party data?

Within things like a CRM system, you should have extensive customer data, including their purchases and demographics. With this information, you’ll be able to create unique insights that are only available to you. 

From these insights, you can create DPR articles that allow you to target backlinks and brand mentions from reputable publications. It could be something as simple as creating an article that discusses trends/buying habits by region, which can then be pitched to regional and national publications.

By using first-party data to create articles or studies, you will also have the added benefit of feeding into one of the other things valued by search engines and AI models: original content.

Things to keep in mind with AI

We get asked a lot of questions about AI and how to make brands visible in responses. And while there are things we can do to support your business with that, there are also a few things worth keeping in mind.

AI is not a search engine

This is a common misconception a lot of people have, but AI models aren't search engines; they’re language models.

Search engines use a dynamic index that’s consistently updated and return resources based on algorithms that are designed to surface results that are most relevant to a user’s query. In essence, their job is to find resources.

By contrast, an AI model’s job is to create. It uses its own knowledge to form an answer based on the prompt it’s provided with. It’s essentially a predictive text model that provides answers using its existing knowledge (often with help from its friends, Google and Bing). Because its answers are probabilistic (fancy way of saying inconsistent), you can get different responses using the same prompt.

The main thing to take away from this? Although AI models get treated like search engines, they’re actually different in how they operate, which means you shouldn’t use traditional search engine metrics to report on their contribution to your site’s performance.

Top ranking pages in dynamic indexes ≠ guaranteed citation in AI

You’d have thought that you’d need your site to rank highly for a topic in order for it to be cited within an AI model, but that’s not the case.

In fact, AI is open to citing sources from further down page 1 (and beyond) if it means providing a reliable answer, as supported by findings from SEO Site Checkup, which found that 76% of citations stem from beyond page one. 

This is important to keep in mind since it can alter how much emphasis is placed on single keyword performance whilst also rewarding sites with more comprehensive content on a subject.

Current SEO tools struggle to keep up

Industry-leading SEO tools, such as Ahrefs and Semrush, don’t provide a fully accurate view of AI data and how brands are being cited in responses.

This is because these tools are configured to report on static URLs and dynamic indexes. They enter a query and gain access to information on where pages rank for that query, which is then added to their database.

The challenge with AI is that its responses are rarely consistent, meaning a brand could be cited one day and then ignored the next. This is simply the nature of how language models work. 

Not only that, but prompts are difficult to track because they tend to be personalised and conversational. The basic answer someone’s looking for may remain the same, but several questions/prompts could lead someone there.

This doesn’t mean you should ignore AI when reviewing performance, but it does mean you should be mindful of the existing metrics available in the tools you use for reporting. 

Google and Bing offer their own tools to help you learn more about how your website is cited within their AI responses, which are worth checking out if you’re interested, but there isn’t anything like that for ChatGPT yet. 

GEO = SEO

You may have seen several articles across social media that herald the arrival of GEO (generative engine optimisation) or AIO (artificial intelligence optimisation) and how SEO is dead.

The reality is that the basic fundamental principles that underpin how AI models obtain the information for their training data and how they choose who to cite in their responses all stem from SEO.

You name it: content structure that search engines and humans can understand; thoroughly covered topics with original research or first-hand experience; addressing customer pain points in a human way; and building a brand worth trusting online. This is SEO in a nutshell.

Brand is more important than ever

We touched on this above with DPR, but the arrival and development of AI means your brand and its reputation are more important than ever.

AI models look to your domain’s authority (among other things) to determine if you should make the cut. Trust is an integral part of the customer experience, and AI models, alongside traditional search engines, recognise that.

So, whilst it’s important to focus on optimising the site and making it easy to navigate and understand for users, search engine crawlers, and AI bots, you must be investing time into online reputation management (ORM), since it determines if you get cited and what the sentiment of that information is.

Adapting in an ever-changing landscape

As we mentioned at the start of this guide, AI has come on leaps and bounds since its introduction, and it doesn’t look like that’s going to be easing any time soon.

The guidance and advice above should help you understand how AI works and what you can do to maximise the performance of your brand. If you think you’d benefit from some support in this space, then we offer a range of marketing services that will help you maximise online visibility, so get in touch to learn how we can help.