Sean says: “Well, I believe that us SEOs should be exploring more around large language models and the results it gives us. Start learning from what it's showing, the kind of content it's displaying, and what we can learn from that.

If it's showing our content, why is it showing it? If it's not, why isn't it showing it? What sources is it showing instead of us?”

So, why would an LLM be showing our content, as it were?

Is it traditional SEO that lets us do that, or are there certain optimization methods that we need to actually undertake in order for LLMs to pick up our content?

“I think, because it is still new, we are still learning about the patterns that large language models are using to find the right sources. It's not 100% concrete right now but what is clear, during my research and during other people's analysis, is that traditional SEO techniques are still very important in terms of getting yourself seen. Building that authority, the freshness of content, the way it's laid out, structuring your content, navigation, headers – all those traditional SEO techniques that are very important still apply.

However, obviously, certain things do help more than others. I have seen that, in terms of content, you've got to be well structured, getting to the point quicker in terms of answering the question directly, not beating around the bush, and not putting jargon and fluff in there.

AI is here to stay, and it's matching pretty much how users are changing their behaviours. They want answers quickly. They want to get straight to the point. They don't want any of that fluff anymore. We've seen that with the rise of social media and AI overviews in Google. That's being reflected in these large language models.

A lot of the time, what I’ve found when our content has been displayed on large language models like ChatGPT and Perplexity, it's the ones where we very much cut out all that fluff, all that jargon, and we got straight to the point. It's really well structured. The headers are clear. Obviously, there are things like structured data you can implement but, generally, it's about that content.

I also found that it's about building authority in the field. Obviously, that's hard to do but, if you can get backlinks from trusted websites – whether it's BBC, Huffington Post, or medical, depending on what sector you're in. These do, as usual, benefit you in the long run.

These things like content freshness also have an impact because these large language models are working on live content now, in terms of search. So, they do still rely on the freshness of content. That is another way of getting seen.”

Content freshness – I'd like to dig into that and how you go about doing that to actually make it more appealing to LLMs.

But you mentioned ChatGPT and Perplexity there as well. Are those the two top sources of LLM traffic at the moment?

“From what I've seen, that is the case, but it could be different for everyone. It really depends.

From what I've seen, that is the case, especially for where I work at Macmillan, those are the two that we're getting the most traffic from. That might be because we're in that sector that's heavily searched. People go to those LMMs to search for that kind of information. It depends.”

What kind of volume are we looking for in comparison to the traffic available from Google at the moment?

“That's the thing. People are going to say, ‘Well, look. Google is still the dominant player.’ – and they are pretty much the dominant player – but there's no doubt about it, large language models are growing, and it's chipping into that market share, bit by bit.

We've seen quite a good increase, month-on-month, of traffic coming from those larger language models.

Apart from that, I get the point that Google is still dominant but AI really does, as I said before, represent how user behaviours are changing. I think it's a great insight into really seeing how they display your content and what kind of content they are displaying, to really give you insights into developing your own content.”

What about establishing the value of brand exposure on LLMs? Because you're not necessarily just going to get direct traffic. I would say a lot of people are going to get direct answers as well, and perhaps have some sort of exposure to your brand as a result of doing that.

How many searches that a brand appears on are you likely to get direct traffic from? Then, for the second part of that, do you attempt to measure the value of that additional brand exposure?

“I guess that's the tricky thing right now, measuring that kind of visibility on large language models because, right now, we're not getting that kind of data easily – if at all. Obviously, traffic is slightly easier to track. It’s not 100% certain either because through Google Analytics, you can track this kind of stuff, but Google is being a bit shy in terms of telling us about AI overviews, for example. Obviously, we don't really get that kind of data from the large language models.

You can do your own manual searches around the top queries that people are searching on your website and see what's showing up if you are being shown. If you're not, why are you not? Who's being shown instead? What are they doing that you're not doing?

Something I did want to mention was around information gain because the large values models are looking for things that stand out that they can't answer directly themselves. What can they add to the query to really add that extra information? It could be stuff like statistics, facts, or quotes from experts. This is the kind of stuff you should be throwing into your content because that's going to make it stand out from all the other similar answers across the Internet.

Just like Google's been saying the same thing around EEAT, it's the same principle with large language models.”

You touched upon content freshness as well. How do you go about demonstrating that to LLMs and perhaps search engines as well?

Is it a matter of simply updating the ‘date posted’ on a blog post? Or do you have to add an additional significant volume of new content to that page?

“It depends, doesn't it? You should always try to be creating fresh content with unique insights (depending on what industry you're in, that's easier in some than others), but you can use large mortgage models to find opportunities.

If they are featuring a bit of information backed up with statistics or figures, and it's from a couple of years ago, and you could get that information which is more up to date, you should go around trying to update that because you might be the ones being featured in the next result because you have more fresh content compared to your competitors.

I do think that's where you could find opportunities and, as usual, track your website traffic. If there are pages that are focused around statistics and people are landing on it and engaging with it, but it's out of date, then maybe look to update that because, like Google, large language models will pick up on the fact that this is popular, but they also pick up that it's out of date, and they might not use it. If you update it, you have got a bigger chance of being picked up for those results.”

You also touched upon the importance of including expert content within your page.

Is there any optimal way of structuring that? Do you use schema? Do you use links to social media profiles to demonstrate who that expert is?

“Yeah, 100%. It's exactly the same as with Google and Bing. These large language models will pick up on those signals.

It doesn't matter what industry you're in. If you're a builder or if you work in the medical industry, showcasing who you are, your credentials, and where you've done other work is always going to be good for being picked up for that sort of expertise.

I do believe that it's similar to what we've been told before but I think that, with this whole thing with large language models, it's really just highlighting and underlining all these other things that people have been talking about in the industry in the last couple of years, but they're really saying, ‘This is now. It's the future. This is what you have to do.’ They're just kind of underlining it.

Yeah, all of those things. Schema does come into play. People have been arguing whether it has an effect or not. I do believe it's important, but using it right in the right places. Social media, 100%, they are being picked up in these large language model results as well. It's all important, those factors, for sure.”

If you're analysing why a competitor appears in an LLM instead of your brand, what are some common reasons as to why that competitor is appearing instead of you?

“Again, it depends on industry. I can only talk from my industry, where I've done some research. They do favour those .org, education sort of domains – those trusted domains. That's one thing that is kind of out of your control, but what is in your control is obviously the structure of the content.

As I said before, make sure there are clear headers, clear subheaders, and clear structured content where people are really getting to the answer – and it's actually a unique answer backed up with relevant data, statistics, and quotes, if you can, to make it a bit more unique than what others are offering. And also making sure, as I say, to keep it fresh. All those sorts of principles. The structure of the content is really, really important.

I was going to mention something as well, around topical authority and topical relevance. It's important that, when you're covering a subject on a page, you don't veer away too much from other areas because you want to be known for a particular topic or subject area, then they're going to be evaluating this content.

If you start talking about random things on the page, they're going to think, ‘There are other people talking about this subject that are really focussed on it, and covering it in a better way.’ It’s just like Google did, but it's a really polarised version of this.

Also, every word counts. They're getting better and better at understanding the connection between words, so you really have to focus on making sure that it's all as relevant information as possible when you're covering a certain topic.”

To stay on this topic for a second, how do you go about deciding on what your topic should be? Because, obviously, you don't want to go too broad or or too niche either.

“It all depends on what your subject area is. It really does depend on your area of expertise.

If you're a plumber and you're talking about how to fix things around plumbing, you don't want to then start talking about gardening, because it is definitely going out of your area of expertise. Just stick with that. Obviously, you can broaden it out. There are ways to go around it. It might be which tools to use or the best places to store things and stuff like that, but you don't want to go too far away from your particular subject.

We've seen it, in recent history, where certain websites have tried to do that, and they've been badly hit. LLMs won't pick you up because they just don't think you're the right expert source to go to for that particular subject.”

So, essentially, know your audience, know what your expertise is, and then perhaps have a bit of a persona and a few paragraphs defined as to what type of content is likely to appeal to that audience.

Then, if you come up with a new topic, there are some more keywords that you uncover, or there are some People Also Ask-type questions that you're considering writing content for, then obviously run this past what you've defined as your audience and what your expertise is, in order to decide whether or not to actually write content about it.

“What I think is really fascinating as well, that I found during my research, is that, if you're doing a service, for example, they're generally not directing people to these service pages. What they are liking is blogs.

Funnily enough, people are saying that blogs are dead and there’s no room for blogging anymore but, actually, the large language models are pointing people to a lot of our old blogs that weren't ever being picked up before by Google.

It's a really great opportunity to expand your content on your website with blogs still – but, obviously, they have got to be relevant and still follow all the principles we've discussed before. So, if you are a service, don't just stuff your service pages with loads of random information like answering random questions. Put it in a blog and cover it well, show your expertise, and you've got more chance of being picked up, I feel.”

We've been talking about how SEOs need to use LLMs. How are you finding other departments' involvement with LLMs as well? I'm thinking maybe customer service, social media, user experience, etc.

Should other departments be open about what LLMs can do for them? How should SEOs go about engaging with them about the use of AI in content?

“It's hard to tell, in terms of how they can use it. I think everyone should be using it because it can make everyone's jobs easier. I think everyone should be using it to a degree, depending on what they do.

In terms of how they can help SEOs and influence being picked up on by large language models, they have a huge influence, as they do with Google, because they can help you build that authority. Obviously, PR can go out and get those pieces picked up by large publications if you have that capability. UX is helping you build a more user-friendly website, better navigation, and all those things that they do. That's very important. Social is getting your name out there, reposting, and showing your expertise in the area. These things get picked up.

Generally, customer service is one of the most underrated sources – at any time, but definitely now – because they can pick up on trends and what people are asking in your industry/field quicker than any search engine can because they're getting it straight from the horse's mouth, so to speak. If you can get that intel from them, start writing about it, and answering those queries.

It's just like Google but LLMs are here to stay. They're only going to grow, and they're producing this content pretty much on the fly. If you can answer a query before anyone else can, you'll be the one to go to for that particular issue, quicker than ever before.”

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?

“I believe they should focus less on those vanity metrics – the kind of keyword rankings, and even traffic to a degree.

I think people should be focussing on things like share of voice, visibility, and, obviously, conversions because that is what really matters, especially if you're revenue-focussed. That's what I think.”

Sean Barber is the SEO Manager at Macmillan Cancer Support, and you can find him over at SearchwithSean.com.