How to get recommended by ChatGPT and cited in Google AI Overviews: the full video transcript

To get recommended by ChatGPT and cited in Google AI Overviews, your business has to pass four gates: AI crawlers must be able to access your pages, retrieve them, extract a clear answer, and trust that answer enough to cite it. This AEO video transcript covers each gate, how to test your own visibility, and what to fix first.

A note on terms: where the video says "AEO and GEO", we treat GEO as part of AEO, not a separate discipline, and AEO leads to SEO. The transcript is lightly cleaned for reading: fillers, false starts and repeats are removed, spelling is British, and a few product names are corrected. Nothing else in Simon's words has changed.

By Simon Young, Founder of Question.Marketing. Published 28 September 2026.

Prefer to read the method in full? Read the definitive guide to AEO. It sets out the four gates, the testing method and the sources behind every figure.

Chapters

• 0:00 Why is now the moment to be first into a gap?
• 2:09 How are AI assistants changing search?
• 5:39 What is AEO and where did it come from?
• 7:47 What are the four gates to getting cited?
• 10:25 Why won't AI slop get you cited?
• 14:57 Does the AI decide who to recommend before it looks for sources?
• 17:41 What is a citation, and whose sources count?
• 22:23 Why does being talked about beat being linked to?
• 28:21 What is the difference between mentions, citations and recommendations?
• 32:02 How do you test whether AI recommends your business?
• 37:55 Which prompts should you test?
• 43:34 What should you record when you test?
• 48:26 Are you blocking the AI crawlers?
• 50:54 Why does a local business show in the map pack but not in ChatGPT?
• 53:38 Why must every fact about your business agree?
• 1:00:25 What is brand entity and why does it matter?
• 1:03:42 What is answer-shaped content?
• 1:07:52 What do schema and corroboration each do?
• 1:12:19 Which tactics should you stop paying for?
• 1:13:50 How do you measure AI visibility honestly?
• 1:18:51 What is overhyped?
• 1:21:09 What should you do first?
• 1:26:24 How do you spot a real AEO supplier?
• 1:30:01 How long does it take to get cited?
• 1:32:11 Where can you read more?

Why is now the moment to be first into a gap?

Video: 0:00

How to get cited and how to get recommended by ChatGPT and Google AI Overviews. This is a definitive guide on how you can get your business listed in not just Google, but Perplexity, Claude, ChatGPT, and all of the other LLMs. I'm going to take you through a few steps, some recommended actions, how you design your own playbook, how you can work with your knowledge base, and how we then take all of that, publish it, and make sure that your site is the one that's showing up in front of others.

And I just want to stress at this point that it is now that the opportunity exists. There is a huge opportunity for you to be first into a gap. And when I talk about gaps, and I will do as we go through this presentation, you'll see that if the AIs do not have the definitive answer to a question or a, or an answer to something that they consider that answer could be improved on, you have the opportunity to pretty much jump straight into number one in a matter of hours.

I've just proved this with one of my clients. We published a video and took the steps that I'm going to outline in this video over a period of a couple of days, published that and a few other items which we'll go through, and that overview, that video overview, as well as the web page and the blog that goes with it, is now cited as number one on Google and number one in AI recommendations for Google as well. So it's straight in. These AIs and LLMs are looking for new knowledge, and they will eat it for breakfast very quickly. They really, really do want to be able to get hold of new knowledge, and that's the point I'm stressing here.

How are AI assistants changing search?

Video: 2:09

So we are finding that impressions are flat. In the web, and I've talked about this a lot already, the web is not owned by Google. Google thinks it owns the web. The internet is a knowledge base, a repository of information that's been deposited there for well over 30 years now. I was lucky enough to see the dawn of the internet, and it feels now that we are at a point in, let's not call it search anymore, assistance. So we're moving into a world where AI agentic assistants will very, very quickly take over from search.

I hear a lot of people talking about SEO. SEO's dead. We are now moving to a world where AI agents will serve you the information that you want, and that's the whole... That was what Google's point was, and Google never achieved this. Google would serve you a page full of links and multiple pages full of links and answers. It would serve you the people also ask answers. But what it never was able to do was tailor that to the person who was searching or give them the answer as quickly as it should have done. So, in an example where I'm searching for, let's say, a set of golf clubs it should know my age, my ability, my budget the types of brands that I already like, et cetera.

Now, as it stands, if I do that search on Google and I go through links and forums and I start reading reviews, et cetera, I'm not getting advised. I haven't got somebody holding my hand and bringing me the information that would be as helpful as possible. Yes, I probably will get to the answer, and it will take a little bit of time. AI and agentic models are moving into a world where it will do that for you. It will interview you. It will fan out a load of questions and get you to that answer.

So searches are still happening obviously, and those of us that are using AI more are using Google and the internet and search engines a lot less. And I believe that we're very, very close with the launch of Muse in the US last couple of weeks. Not in Europe yet, unfortunately, but a free agentic assistant, so it's more like a personal assistant, that product is flying. But who knows where we go with other agents, the pricing, the models for business, et cetera. The fundamental principle though is people are getting what they need a lot more quickly which is meaning less clicks through to websites, and it's meaning that answer comes from AI.

Now, the important thing with that is because the answer is coming to you and talking to you, that then means that you are far more likely to convert when you do end up onto somebody's website.

What is AEO and where did it come from?

Video: 5:39

So what is AEO, and where did it come from? What is the acronym? So answer engine optimisation. I've been talking about AEO since 2019 when originally I was talking about the interactions people were having with Amazon Alexa and other voice agents. So probably the biggest drop or drop of the ball seems to be Apple with Siri, and if they'd pushed forward and were still really marketing that could well have been really the way forward. People didn't really use the speaker system very much. I believe we are going to end back in that world and it... and Alexa was probably just early.

We've just seen Mark Zuckerberg announce that they are going to have an assistant that goes along with Muse, and they're going to launch that before Christmas 2026. So we are changing very quickly how we are interacting with data, and I can't honestly highlight enough how much knowledge is worth now. Knowledge... People say knowledge is power. Knowledge is now a currency, and as that knowledge comes out of people's brains onto, not onto the page, it could be onto the page, it could be into video, it could be into all sorts of different ways that people are able to access knowledge.

For example, podcasts. So this episode does end up going out on my podcast. It will be cited by AIs, it will be read and interacted, and if people trust my opinion, then it will then go further and further. But what we need to look at, how this video's going to cover some of the basics, and there's an awful lot more that you can do. There is a ton more that you can do.

What are the four gates to getting cited?

Video: 7:47

So I'm going to talk about four gates. You've got to first allow the AI agents to access your information, and as those machines get to access information, we will then look at how citation actually works. First off, you have to allow them through the gate. So things like Cloudflare sometimes block the other AI agents from getting through onto sites or wherever you've published your information. You've got to allow them to be able to retrieve the information and extract it. So structuring the data correctly means that it sits in a format that is very easy for whichever agent is picking that information up to understand, and they've all got different preferences.

I've talked for over 20 years about Google Schema being, Google's language, so to speak, and structuring data properly will give you an advantage and no more so than now. So as we move forward and we're all publishing more data, think about making sure that you put the right tags, the right schema, and the right metadata into pieces of content that you're publishing so that it is therefore consistent. That's one thing. But we need to get the answer to be trusted as well and picked up by all of the platforms.

And there is no science at the moment that sits behind saying, "Right, this is how you are then cited across all of the different models." But what does citation mean? Citation means that your piece of knowledge or whatever you've published is confirmed then by a third party. So we could all go and publish a website today and get it to go and scrape, the entirety of the internet, so to speak, and publish a site, and then we could put some of our own data into it.

It doesn't mean you're right. So we're moving from an area where in the past you could almost publish anything, and you could, in theory, rank anything using particular keywords and end up with content on Google. Or at or near the top of Google.

Why won't AI slop get you cited?

Video: 10:25

The way you need to work now is you need to firstly work out what you're going to publish. Now, there's a lot of talk around what we would call AI slop at the moment. I cannot advise you strongly enough to just stop for a minute. Don't use ChatGPT or Perplexity or Claude or whatever to make your content and then publish it as is. Because if you do that, there is no information gain, and one of the things we will talk a lot about is information gain.

We mustn't just plagiarise the information that already sits out there on the internet or on any of the other engines. And when I say the internet, Google would probably say the internet. Things like Meta might say that they have a database that is harvested from WhatsApp and ChatGPT might reference lots of information that's published and spidered by Google. And let's remember that not everything is indexed. So there's a great big chunk of... I think people presume that if you publish something on the internet and you've got a website, particularly if it's new, if that site has content on it will be indexed, and that is absolutely not the case.

A lot of the agents, the amount of time that these bots spend spidering and indexing, there is not enough compute power to just go out everywhere. So they... Well, there probably is. But what they want to do is make sure that they use their resources efficiently, and they are spidering the best content. Things like Perplexity has a hunger for brand-new knowledge. So, and when I say brand new, we are seeing content that is published 24 hours ago, going to the top of Google, going to the top of some of these search engines, or I don't know that we want to call them search engines anymore.

I think the EU has just announced something around they are classifying ChatGPT as a search engine. I would say that is very naive of them and that they're not, that's... It is not a search engine. It's an agentic assistant or an AI search data aggregator where it's pulling data from a corpus. Now, where that data comes from, they're, most of these search engines know more about us than we probably ever know. It, and I think it'd be, we would all be very, very scared about what they actually do know about us.

So one day, this will all come out, and they will be able to see the footprint that, for example, I wish I could see my footprint that I've left on the internet forever. Maybe it exists, I don't know, and how much of that is catalogued and able to be attributed to my personality, I don't know. As we move forward, I think the internet may well not have that possibility as much as the AI agents will. So as you search and interact with agentic AI, I think you'll find that there's an awful lot more information being garnered about you as you, ask it questions and navigate your way through that knowledge base to get to an answer.

I think that there is a possibility that now a lot of that can be a lot more personalised, and there'll be a push, pushback against that. We will see, I think we'll see steps taken by some agents to become user agnostic and not harvest data and openly say that's not happening. But will they be as good as the big boys? Your Googles, ChatGPT, that new model GPT-6 Astra that's out now. I was listening to Sam Altman talking about that, and that is very scary, the level of knowledge that it's actually able to even process.

Does the AI decide who to recommend before it looks for sources?

Video: 14:57

So we, the model will decide who to recommend, and a lot of people talk about, are those searches the flip of a coin? Are they? At the moment, I think we are in a flux. We're in a point where it is new. It is so new, and all of the agents are changing. They're all experimenting with different ways of serving data and what to trust and what to cite and how it serves certain people. So there's a lot more going into this than there ever was with just a plain old search engine of just let's serve a bunch of links.

I spotted this a few years ago when I was able to rank YouTube videos at the top of YouTube in 20 minutes. That's still a thing. But now we're seeing... And the reason, let me just explain that. So the reason that you could rank videos at the top of YouTube in 20 minutes was because YouTube, being owned by Google, YouTube wanted to be able to show new content and fresh content all the time. But what it would do is it would immediately put it to the top of YouTube and show it to what it perceived to be its target audience.

And if that target audience then interacted with that content i.e. they found it useful, then that content would probably stick. If the first few people, let's say maybe the first 50 or 100 people that saw a piece of content, found it boring or just turned them off, it wasn't correct or whatever, that content would disappear into the huge ocean that is YouTube. And it then disappeared. But I've proven that this week, that we can take a piece of content, publish a video, put it on YouTube, and you will see not only that video going to the top of YouTube. Again, it needs to be useful, factual. It needs to be able to be confirmed and accurate. One of the worst things you can do is not have accurate information, but also have conflicting information on one piece of content and another piece of content on your site, because then the AI agent is going to go, "Well, which is right?" So it won't then cite either of them.

So one thing you will see across a lot of what we do is you need to make sure that across all your sites and properties, be that social media, your website, anything else you publish, your videos, or even interactions that are happening on other social medias, that you're not contradicting yourself.

What is a citation, and whose sources count?

Video: 17:41

So citation needs to be something that you're going to pay attention to. The model appears to choose which brands largely to recommend from what it's learned from training, and then it goes for looking, it will then go out and look for sources to support the choice that it's made. Now, that means in the past, on good old traditional search, you could put up a piece of content and then send Let's say you wrote a blog about a particular topic, you could send that blog out to tons and tons of other websites depending on their domain authority.

That's an old school word now. But if, there were websites that were particularly relevant or had high domain authority and your article was cited on them, so that's a citation, a third-party confirmation that your piece of content is real, then Google or another search engine could look at that and go, "Okay, that's confirmed. We've measured that. We can see that, yes, Simon's published something, and on a third-party site, that's also confirmed." So that's a... It's just giving you a thumbs up in layman's terms. But now, as these AI agents go around, they're looking for more and more confirmation on, is a piece of content true?

So we can't see necessarily what those models' workings are. They're not going to make them public. But the leading hypothesis is that we need to have a piece of content that is cited and cited by a third party. So you'll see, we'll go into it, but you'll see, for example, content that is posted on LinkedIn, not in the, not in normal posts, but in articles, counts as a citation. Content that is published on YouTube can count as a citation. You will see, for example, that obviously Google owns YouTube, so therefore YouTube videos being published therefore lead to citation in Google AI Overviews.

If you look at ChatGPT, you will see that it very, very rarely, and I mean, we tested, I think we looked at 25,000 plus citations, and within that on ChatGPT, there was only one YouTube video. So it looks like ChatGPT or OpenAI are actively not citing content that's on YouTube, so they're not allowing Google across the fence into the OpenAI database, so to speak. And they are trying to... All of these models do feel like they are trying to protect their data sources. They're circling the wagons, so to speak, to have their own data corpus that they are very protective of. This actually goes against Google's original raison d'être. Their reason for living was that they were going to give everybody fair access to all the information on Earth, and that's why they became the search engine or as a lot of people, if you asked everybody what the internet is, they would say, "Well, it's Google." We're very, very quickly moving into a world where it is not Google.

Somebody called me a crackpot recently on LinkedIn, but for saying that I think maybe in a year's time, we don't even have search engines. They will exist. They'll be there if you want to access them, but it'll be a bit like going to the library. Rather than searching something on a computer, it'll be the repository of information in the background. We may get to a point in the future, well, I'm sure we'll get to a point in the future where you don't even need a website.

You're, you will feed your own AI its information and let it learn about you and what you do and what products you sell and how you help and why you're unique or maybe, how you help people in your local area because that's going to be something huge. Local AI listing and search is coming very quickly. So, if you are a local business, you should be working, or you've got a particular territory that you work in, you should very much be working to publish information around your expertise in your area.

Why does being talked about beat being linked to?

Video: 22:23

Being talked about beats being linked to 100%. You'll see, I still see, because I'm in the industry, I get offered link building packages one after the other because people still think SEO works. SEO is the basics, the very basics of what we're doing here, and you can literally train your AI agents, be that Claude or Grok or GPT-6 Astra. You can build all of your SEO into your site pretty easily now. You need to just... You can just tell it what you want to achieve, what you're trying to do. Just even say to it, "Here's my website. I would like to SEO it. I would like to improve it. Can you give me some suggestions?" And talk to it, and it will do all of that.

It will do miles more work than a lot of external SEO agencies will do. A lot of them do on-page work. Well, that's done. That's covered now by any of the agents. They will then say, "Maybe we'll do some link, link building," right? Link building I would probably not bother with. You want to be talked about. Be that on a podcast or be that on a third-party piece of PR. Reach out to people. Get your brand cited. And you won't do that with plain content. I've worked with lots of PR agencies, and if I sent out a PR, a press release.

If I sent a press release out to an agency and it had nothing new in it, and it had no real news, think of it as that. Would a newspaper publish a story that's already been out there? If it's already known knowledge, they're not going to publish something today that's been already talked about in their publication. So that's how you should think about how AI agents want to engage with you and your brand. You need to be publishing something new and some statistics or some data, that works really well listicles, how-tos, how to use a product, all these sorts of things.

If an answer exists already, so go and search, and there are tools to do this. I'm building one myself. It will be called Known Gap, but i.e. what is the known gap in the information corpus? So, if you go out there and you search around your product, I would say to you, go and search the top 20 or 30 questions that your customers might ask about your product or how it helps or what problems it solves or if that's a service business, what your... How are you different?

I asked somebody this week on a meeting, "How are you any different to the tens of other people that I've talked about?" You know, somebody trying to sell me a lead generation system, and he couldn't really answer it. He just went, "Well, we're better." And it doesn't prove to me that you are. And the AI Overviews want you to prove it. They need somebody else to talk about you. So that is the overarching citation recipe is get a third party. It's all right standing there. If you're your own website or your own channel and you're shouting all the news, everything you can say, nobody's listening.

You need third parties to pick it up. So, we need branded web mentions, and that is what is appearing in AI Overviews, not backlinks. Backlink correlation is disappearing, and this is why you're able to then rank content very, very quickly above sites that may have spent years and years of time and money ranking themselves number one on Google. Those clicks are disappearing now because people are seeing the AI Overview first, getting the answer they need, and then going off to the, to, or maybe asking a few more questions and then either not clicking through or ending up on somewhere where they're already pre-qualified.

And we know for a fact, 100%, that the traffic that comes from AI Overview converts five or 10 times better, depending on which industry you're in already. And these conversions will eventually happen inside the LLMs. So I believe we're already seeing ChatGPT launching sales agents. So in the golf club example, I could well be talking about trying to find out what set of clubs is the latest and greatest and may suit me, and then an agent may step in and start talking to me about products. It could eventually then perhaps have my bank account attached and then move to taking a sale without me actually going even out of that LLM. So obviously this is a threat to things like Amazon, because as soon as the agent starts selling the product and it's getting delivered and I can take care of it and it keeps me up to date.

Imagine a scenario where, I've... In ChatGPT, I did some research and I've decided I'm going to buy a product, and it says to me, "Oh," tomorrow morning it says, "Simon, that product has shipped. It'll be with you tomorrow. Here's the tracking number. Oh, by the way, you don't need to go and do anything with a tracking number 'cause we'll tell you when it's getting near your house," et cetera. So all of this sort of stuff is happening.

What is the difference between mentions, citations and recommendations?

Video: 28:21

So you need the mentions, you need the citations, and you need recommendation. So trust signals like from Trustpilot, from Trustmary, from any other G2, for example, in the software space is probably the biggest signal that is showing trust because of the reviews in it. They are not all the same. So there are mentions and citations and recommendations, and it's, you're going to get, you know... To get mentioned, this is where we're talking about the difference in how that recommendation comes up. You could have pieces of content from your website or from a channel or an article all mentioned.

Or you could get mentioned in social media. I am experimenting with lots and lots of social media content now because we have the ability to push out, or you could send out hundreds of posts a day onto social. I've seen Gary Vaynerchuk posting about 300 pieces in a day to test whether it goes to different audiences. If you take me for an example, I've got 30,000 LinkedIn connections, and when I post, I tend to get I'll tend to get 300, 400 initial interactions with that piece of content.

Now, what's to say if I post more and more content that doesn't go to a different 300 and another 300 and another 300? And I'm not exactly annoying people at this point because some people presume that if you post one piece of content, it goes to all your network. Well, it doesn't. It clearly doesn't. And what we see is situations where if you post and you post, it's not wanting a piece of content to go viral that matters. You would assume, and I did assume this until we tested it a couple of weeks ago, I assumed that if I make an amazing piece of content and it goes viral And therefore it gets lots of interaction and lots of mentions, and maybe it gets some shares.

It's the shares that I would want. The saves and the shares of my content, not just the likes or maybe the odd comment, because those can also be engineered. So things like the LLMs will probably ignore most of that. They want new knowledge. They want them to talk about you. That will lead to a citation. So training of the... Training these LLMs is something that you need to be able to learn how to do and then make sure that it is able to be retrieved, but you'd need... You definitely need both.

And as we go on, we'll... The models are changing. They are changing quickly, and they're all going to have their different styles, and that's what fascinates me. We have to learn how to train and how to make sure that they retrieve our information over and above that of people that we're going to try and come up in front of. But remember, if there are gaps in the knowledge base corpus, if there is a question that gets answered, and there will be answers to all these questions.

But if you look at the answer to a question on Google, for example, and it's something that your customers are asking, and you think, "Well, I could go... Do a better job at answering that. I can give a more rounded, more concise, or more trusted answer," that's what you need to do, and you need to look at those gaps now.

How do you test whether AI recommends your business?

Video: 32:02

So as we move forwards I'd love you to go and test it. You don't need much to go and test all of these yourself. What you need to be able to do is go and get a bunch of questions. So we started talking about questions. What would your customers ask? Ask in their words, not yours. Just be careful that sometimes, we as run, we who run businesses, we are often very, very close or too close to our products or our business. We probably sometimes ask the wrong questions.

So a really easy prompt would be if you went into Claude, "What questions do you think, Claude, my customers would be asking about my business? And would which questions would then therefore lead to them wanting to deal with me as a business?" Now, you do need to train your knowledge base, and I can't stress enough how important that is. Teach, start teaching all of the AIs, all the LLMs. Start projects on all of them and start training them about your business. Start teaching them everything you can about your business.

This is an arms race. Not an arms race, but it's a knowledge race. It is build as much as you can. The barriers are gone. In the past, if you wanted to build a 200-page information-based website that told the world about every single aspect of what your business does that'd be pretty hard. Now you can do it in a matter of hours. With some of the tools that are out there, like obviously Claude and ChatGPT, but you can use those to go out and find information.

As long as you can add to it and put it in your own voice, that's really great. It's not necessarily always about complete information gain, but leading the way in a particular field. If you have the opportunity and you can put that information into that knowledge base, you will win. I've just taken recently a client's phone records, and they had thousands and thousands of phone calls. I then said, "Okay, let's remove all the phone calls that are less than five minutes," because they could just be interactions between staff, catch-ups, et cetera.

I wanted the stuff that had the real detail in it. So what I did then was I narrowed it into anything over above five minutes, over 10 minutes, and over 20 minutes, downloaded all of those phone calls, and then uploaded them into a knowledge base and got AI to go through it and read it all. It was even telling me things like particular salespeople that were on the phone. One particular person was just spending a lot of time with a chitchat, which is fine 'cause they were probably just getting to know the customers more.

That was their style. But it highlighted the fact maybe there was a training opportunity. Maybe that member of staff, if we look at their conversions, maybe that is the way to go. So it's not, it wasn't a bad signal. It just told me that person had a different style. But the knowledge that came from all of those recorded phone calls basically grabbed all the questions that their customers always ask, and they ask them again and again in different ways. So we want to be able to answer in all of those ways.

But if you take, go and take 10 questions, in your customer's words, and then go into fresh, a fresh ChatGPT or Claude or whichever one of the LLMs you use, ask it the questions and see if you are cited. See if your business is cited. See what is cited. Is the answer that is provided good enough? Does it satisfy... Put yourself in the customer's shoes. If they were to read that, imagine... And the answers won't be the same every single time. And this is the way the LLMs are experimenting.

So don't expect to put that question into one and get an answer, and then come back tomorrow and do the same thing and get exactly the same answer because they are experimenting with how they serve data to the people. So it's not like there's 10 links on a page of Google. This is thousands of opportunities for every business to come up in searches at the top. I don't believe we're going to get to a page or a, to a result where always a particular company is cited for every query because that would, one, be unfair, and two, wouldn't allow the models to learn more as they went forward.

So I think that's key. Take all the screenshots record where you were named. And was your site linked? And who appeared instead? Was it one of your competitors or was it a weak result. And then just do that again, run it again maybe a day or a week later and compare it. And this is just a rough ready reckoner. It's not a scientific experiment. But it will give you a good idea of whether you are cited at all or not. There are a lot of tools, a lot of paid tools that will give you a better idea, and you can run a lot more queries.

If you're really going to move into answer engine optimisation and you're going to work in generative engines and you're going to work on the questions that your customers need to find you for, you probably want one of those paid tools. But start off the way I've said.

Which prompts should you test?

Video: 37:55

But what you've also got to work out is which prompts actually matter. So if we think in the past, SEO companies would sell you keywords. Please be very specific. I know the temptation is if let's for let's say, for example, you're a recruitment business and you want to come up for recruitment Manchester. That's fine. That's not what you're, you want to come up for as a prompt because if you do and you don't satisfy the need of all of those people and you, some people, "Oh, yeah, of course, I do. I'm a recruiter in Manchester. I do." But you've got far more chance of being cited and recommended if you are specific.

Now, in the good old days of Google, if we talked about keywords, we would also talk about long-tail keywords. So an example in the recruitment space in Manchester might be I'm looking for a specialist recruitment partner in Manchester who finds software engineers. Now, recruiter Manchester or recruiting or recruitment agency Manchester might get searched thousands of times a month. The phrase that I've just given you might only get searched five times a month. Now, everyone's going to sit there and go, "Well, I want to come up for the one that is searched thousands of times, don't I?" No, you do not.

You want to come up for the very specific answer and you want to build your content around it because it lays the foundation for you becoming the cited source. You cannot... Well, I might be wrong, unless you are the number one recruiter in Manchester and have been for 30 years and you've got thousands and thousands of publications out there, then that might happen. But what you, what the AIs want is to direct somebody to the very specific answer that they are looking for to satisfy their need straight away.

Because if they don't, one of the others will and people will all migrate eventually to whichever of the LLMs gives the best quality answer and serves up a quality answer straight away. Saving us time and meaning that I don't have to go off to other websites or LLMs to go and find other information. So think of it that way. You would much rather come up for the long tail than you would for something very, very broad. And there will be people that disagree with that, but my mind leads me into the fact that we can then tailor the content.

So your prompt set and your content plan are the same document. So once you've decided what your prompts are, you've got to then go and build a content plan around it and most people build these separately and then they wonder why neither of them ends up working properly. So people who don't know yet that they need you, why does X keep doing Y? Is it normal that? Or how much should this cost? Where the volume is where the competition isn't. So where buyer trust or where the trust gets built months before anyone buys.

So it's a case of if you're solving people's problems, we need to see you becoming the answer, but you've got to lay down those foundations. You've got to, you need to have that content out there and get there quickly now. Do not waste any time because there will be people building content and brand building and as we move forward answering questions, hence my business is Question.Marketing. I want you to build the questions. I want you to answer the questions. I want you to be the question.

But you have to build the content around higher intent. Where there is higher content, higher intent in a search, that is where the competition will sit. So if I am in market, let's go back to my golf clubs example again, but if I'm in market for a set of golf clubs I might have searched that on Google and I might have bought them online, let's say, but I'm more likely to go and buy if I've, who's the best golf professional in the Doncaster area. I don't know that question is answered.

Who could I ask some advice of in Doncaster around golf equipment. And then it'll recommend someone, or it might cite someone, or an article, or a video. So think of it think of what your sales team is getting asked for. What do your recorded calls or your emails... Don't do it from a keyword tool. That's where we're moving to.

What should you record when you test?

Video: 43:34

So if you are named, was I named. If you're looking to be above your competitor you need to be, you need to have everything corroborated. And that means we have to be named. So we need to be accurate. Old addresses former staff, discontinued services, wrong prices, all of this turns up all over the web. We've all been maybe lazy about updating our web properties and what's out there digitally about our business. You need to make sure that look at who's appearing above you, for your competitive set of phrases or questions.

And if it is frequently not you may well have inaccurate information. So if an AI comes across and finds outdated information, it'll discount you. It won't even spider you. Are you getting recommended elsewhere? And what is the tone of that? What is the sentiment? I don't hear enough people talking about sentiment in business or brand sentiment. We are moving to a brand-led world where it's going to be who is the recommended brand and why, and who is saying so. Not just a bunch of links or somebody... I know a, an SEO business that owns 6,000 websites.

The way they did their SEO, they would take your content, they would publish it across all the blogs on all these 6,000 websites, and that they thought was job done. And in the last couple of years ago, pretty much that would work because they would get thousands of citations. But that is gone now. SEO is disappearing. The way this is all going to work is completely different. Make note of who appears instead and see why. You can ask, ask, ask. One of the things that people don't do.

Why are you not just doing a search in ChatGPT for a question that you should come up for and then looking at who has appeared, then ask it, "Why have you not named me or my business? Why are you not talking about our business? Why are you talking about our competitor instead?" You don't need a tool to do that. You can do that. You can set an automation to go and do that on a weekly basis and send you a report. Which sources have been cited for that competitor?

So why did they appear? What sources and recommendations? Where were they indexed? What sent this particular LLM to give the response it did? Ask it the question. So you then have a list. You have a list of sites that you could get cited on. Now, a lot of people would then talk about, "Okay, my competitor comes above me. They are mentioned on this blog this podcast, this LinkedIn article, this influencer," et cetera. And then they would go out and try and replicate that by going and, getting the recommendations from those sources.

One, that's probably quite hard to do because if a citation comes from a blog post of a particular influencer on a website and you're saying to them, "Well me too, please talk about us," well, give them a reason to talk about you. Don't just go, "Can we have the same article?" Or something similar to what you've already talked about. You'll see a lot of articles and a lot of blog posts on influential sites where they're talking about a list of products, 10, 15 of the best, SaaS products of in a technology niche.

Here's the list. Well, a lot of those are citations. Well, nobody's going to go back and edit that original blog post when they can go and publish something new. Give them something new to talk about. They don't want to repeat themselves. And in the example I gave earlier about publishing in a newspaper, if you were going to publish into a industry publication, I used to do a lot of work in the motor industry, they, those publications will not even consider a piece of news unless it is news.

I know that's a bit of a... That sounds a bit odd, but what I mean is you need to be new news to give news. I've made that up on the spot. It sounds all right. But hopefully you can get a feel for what I'm talking about.

Are you blocking the AI crawlers?

Video: 48:26

So are you getting yourself blocked? These, you know and none of this is to do with robots.txt or your llms.txt. There are still people talking about llms.txt files and selling people that. Yes, you might want one. It's not necessarily it's not necessary. Cloudflare is probably your biggest danger. So there is a setting in there that you need to switch on to make sure that you're not blocking the LLM robots from coming in. That might mean that you've got your firewall bot in it's configured incorrectly. You need your robots.txt, and if you don't know what a robots.txt is, maybe, you should research that, but most websites should have one.

Editor's note: Cloudflare changed its AI crawler settings on 15 September 2026. Its new Disallow AI Training setting writes a no-training preference into robots.txt, while Googlebot, Bingbot and Applebot keep crawling for search. Choosing Block now stops those crawlers entirely, search included. Existing settings carried over. Source: Cloudflare blog, 15 September 2026.

If not, again, say to your AI assistant, "Please create it and make sure that none of the LLM AI bots are blocking or are blocked from accessing our site." Make sure that your content is structured properly. So a lot of the time that means putting the content at the top, making sure it's in the right position. You don't want things like no index, and there are some plugins that can cause problems. Again, take this slide. Take a screenshot of this and then put it into your LLM.

And say to it, "Have... Do I suffer from any of these problems? Is this an issue for me?" And then, does it mean that your AI assistant will then go through your site, make sure you're not suffering from any of these basic problems. Some of the AI crawlers, well, they don't run JavaScript, so it is a reason to rebuild a site if you've built a lot of it based on script, it is not going to go through a piece of JavaScript and it's not going to read it properly.

Structured HTML, you must make sure that it's done properly.

Why does a local business show in the map pack but not in ChatGPT?

Video: 50:54

So you're in the map pack. A lot of local businesses will list themselves on Google Business Profile and local SEO is huge right at the moment. It's not necessarily transferring that data over into ChatGPT. I would also say to everyone who's listening to this, do you publish to Google Business Profile? Even if you... Like I would say, I don't know the statistics on this, but I would imagine 99.9% of businesses do not publish to Google Business Profile on a regular basis. Why not? You should be.

You should publish social media content, images, and blogs into Google Business Profile. Why would you not. Because you're just laying down more foundations and more opportunities to get found. So there is there's a massive opportunity. If you look at the locations that are recommended currently by ChatGPT, it's very, very low. This is because it's really a Google property. This is where we can start seeing the split. So we can see that OpenAI is not recommending local business places. Now, I've just told you should be publishing to it.

Of course you should because Google is picking that up. Google 100% will be picking that up. Out of 350,000 locations across 2,700 multi-locations brands, and this is US weighted. Only one 1.2% of that were recommended by ChatGPT, and 35% of that, or nearly 36% of that, of the same locations appeared in the Google local map pack. We have to enter a world where local AI LLM optimisation for search happens. This is where we are now. This will change. You will move into a world where, well, SEO companies will probably transmute into becoming AEO or GEO advocates if they're not already talking a lot about that.

You've seen some of the biggest SEO companies out there already moving towards, moving their offering from SEO to AEO and GEO. We are seeing that.

Why must every fact about your business agree?

Video: 53:38

So this is four things, and it's pretty unglamorous. Your site on your Google Business Profile or Apple Business Connect, Bing Places, trade directories, review platforms, Companies House, professional registers, anywhere where your site is mentioned or your business is mentioned, make sure it is consistent and the facts are identical everywhere. I had a scenario this week where I was dealing with a another recruitment business, and they had listed on their website, it was an, it was because part of their website was really old, and it had one of their old company numbers on it, and it had on the front page a different slightly different company name because they'd gone through a rebrand, they'd changed, they'd... I think they'd had a management buyout.

Now, that meant that when a crawler went on their site, it found ambiguity. So if any of these crawlers are finding ambiguity, did not the easiest word. And I hope that's been transcribed correctly into my subtitles, but I think you know what I mean. But if you're starting to have any sort of ambiguity across any of these platforms, it's a big problem So we need to see all of your properties consistent, across everything. Author data, business name, phone numbers, links, website addresses, social media profiles, names of those profiles, et cetera.

Where... How you're named in different business directories. Just don't be lackadaisical. Make sure that, for example, your social media profile references the correct company name and always talks about your brand the same way. I've got another business that I'm dealing with, and they are a Meta business partner. Now, they had been describing themselves as a part of the Meta business platform. Two different things, confusing. The computer says no. The AI says, "Well, this isn't the same thing, so I can't cite it." It doesn't want to get things wrong, and if it can't agree, it's one of the simplest things that you can do is make sure that, you've got your reviews describing your job.

So when somebody gives you a review, I've got a client who's a roadside recovery business. Now, when he does a collection off the roadside of a customer, they do quite often give him a review because he's very helpful. But what you don't want is a review that says, "Thanks, mate. That was a great job." You actually want a review that says, "Well Rick did a great roadside recovery. He picked my car up late at night, delivered me back to Doncaster and thanks very much." So it actually talks about the service that was undertaken.

So if encourage people, if they're giving reviews, talk about the product, talk about the service in the customer's own words. On more than one platform. All of this ties together. And it's just basics. Make sure you've got real location pages. Make sure that you have actual local questions. So don't say that you are a roadside recovery business in the example before. Say you're a roadside recovery business in Yorkshire or a roadside recovery business in Rotherham. So that when people search and they look for, they ask a question, who is the whose the best or the cheapest, because that's what people will say, or who can get to me the quickest?

I need a, I need my car recovered from the edge of the M1 in near Rotherham to get back to Sheffield. Then that person will be cited because they've answered the question, and it actually talks about their locations. So if you can get into things like local directories, business associations, community sites, all of these that are, that have been around for years and are already trusted, please look at that as opportunities. So not the sexiest piece of advice, but it works. And what you're trying to do is train the models to agree.

And to cite you on that basis because almost every business out there will have a contradiction. Make sure that's not you. It's why the machine will ignore you if you've got some of these pieces wrong, and it's really basic stuff. So as we know, the web says a couple of things about you. The models will refresh on their own schedules. There's no... Anyone who turns around and says, "Oh, yeah, we'll get you, listed in Perplexity in a week," blah, well, it's just they can't guarantee that because they don't, they're not able to do that.

Make sure you haven't got old profiles. Or a previous trading name. I think I just touched on that a minute ago. Or an address that's wrong, you know? Or a service that you stopped offering four or five years ago, or a phone number on a directory that nobody's updated. It's not going to take you long to audit. I think, a day to audit, I could get a Grok bot to go and do this for me in a day. And go and find out all of the information that's possibly incorrect.

Again, a Grok bot or Astra can do the outreach for you to get it corrected or do that yourself. I would encourage you to do it in the quickest way possible because every place that your business is described online, the sources and the engines will be probably using that to cite you. And if it, if they get that diagnosis wrong, then you're not going to get listed. So make them agree. Correct what you control and request the corrections from the places that are getting it wrong.

What is brand entity and why does it matter?

Video: 1:00:25

So we are also working in a world where brand identity, it matters more all the time. We talk about trust. Where most businesses struggle and they fail is they haven't built their brand identity correctly. Over time. It's not a, it's not an immediate fix. So we've already talked about a set of facts like your name, address, the date you were founded, the people. People. Talk about your people, what they do, what their job titles are. All of this can be mapped into your knowledge base and then served to the LLMs.

Let them pick all of this up. It might be stuff that you don't do at the moment. Does that give you an advantage over your competition? I'm working with a few people in manufacturing. That industry is traditionally, and they won't mind me saying it is quite old school, so they don't necessarily have the web properties, and they haven't done this piece of work. So get... What is the expertise of somebody's job title? What do they do? Describe that in some author and metadata, put it into pages.

Tell people about why such person is an expert in what they do and what they know. Get that knowledge from that person, publish it on your site. Make sure it's identical across different properties on the web. The structured data of an organisation will tie all this together. So you can have the markup and the links, and the machine will do the job of joining the dots. Providing it's not having to guess or it's seeing contradiction. We don't want it to do that. And then you can serve things into, for example, Wikipedia, Wikidata.

Quite hard to get in there, can be, but as long as you work with them, it can be well worth it because it is one of the most obviously trusted and cited sites out there and independent. And again, if you can get into the press and trade bodies and directories, podcasts. Podcasts, huge. One of my Grok bots, it goes out and looks for podcasts on AEO, answer engine optimisation, GEO, looks for SEO podcasts, looks for people in marketing, and then it submits me automatically asking, can I be on particular shows?

The reason, because then you get a third-party citation, you get a chance to talk and then you get links, you get other people watching you, other people talking about you. If you can get all of that done, I'll be very surprised if you get that done quickly. But if you work towards it, and this is what we're talking about, can you work to all of these points and make sure that you do it before your competitor does? Because then your position will be winnable.

What is answer-shaped content?

Video: 1:03:42

Answer-shaped content, extremely important. So I talk about Q&A. I've been talking about answer engines since 2019. That's not a brag, it's just the fact that I wanted to work in answering people's questions from a marketing perspective, because we all know if we answer a question the best, you become the authority and therefore someone should trust you, and especially if it is corroborated by a third party. And at that point, you're a good way along the line of becoming that person's supplier or preferred source of knowledge.

So make sure that the first 60 words quote you. And are able to be... You know how it is. If we read content, we all do it the same. We read the title, we might get a little bit into the first paragraph. We've already made our mind up. So if you just say first impressions work. First impressions matter hugely in search and AI search, particularly. The retrieval is looking for the question and the answer to be in that first 60 words. It's not to say it has to all the... You know, clearly, we can't always put the complete answer into a first paragraph.

But you want to open and code the answer with a question and answer schema. So in good old SEO land, we would have done title plus H1, i.e. heading one. If that then becomes an a question and an answer, even better. So if we put those questions down correctly, ChatGPT citations come from 44%, or this is a quoted figure, 44% of ChatGPT citations come from the first 30% of the page. I would probably say probably let's just round it to the top third of a page.

It's not bothered to go down a page and go further into detail, because what will happen is most of the web or the information that's out there tends to contradict itself, or not necessarily contradict itself, it weakens its position or dilutes the content as it increases its volume. So the other point here is when you are writing, write on topic, stay on topic, don't become broad, because when a bot comes through that page, if it gets confused at all or can't understand what the page is about properly, it's going to ignore it.

So heading, question. First sentence, answer. Give the evidence. Numbers, sources, specific examples, and the reason it's true. Then you want some sort of nuance. You need to answer the, how it changes, what the exceptions are, maybe what it depends on, or way, how you might be incorrect or whether this is a hypothesis or a guess. And then answer what is the next step. What, where do you expect this query to fan out from? And you'll hear a lot of people talking about fan-out queries. So if somebody asked, and we'll go back to the golf club example, I'm sorry if you're not a golfer, but if I, if my first question was, "I'm looking for a new set of golf clubs what have you got?" And it might come out with a few recommendations.

The next question might be, "Okay, I'm looking for something that's between, I don't know, up to £1,000. I don't want to spend too much." Then another question comes, then another set of data comes. Oh, well then, what's the latest models that have come out? Now we're already three questions deep. My fan-out query could end up 100 questions for all we know, and that's where I think the fascination will come with marketing is how those queries get built out, what the next steps are and where this ends up.

What do schema and corroboration each do?

Video: 1:07:52

Okay. So, schema and corroboration. My pronunciation is not the best today. One tells the machine what you are. The schema, and the other tells you whether to believe you. So it's all right publishing and then is this true is the other important thing. Can it... Can this information that you've provided be corroborated. The schema is going to describe the content. It can't be a substitute for it. So you, the schema sits behind what your publication is, and it lets the machine read the facts without confusion and who you are where you are, what works for you, what you sell, what you do, what all of that is written and retrieved.

And structured data is still list... If it's listing staff that left three years ago, that is better than none, but why would you not correct it? You must have it correct. Think all the time about how important the information that you're providing the front window to your business, why would you not have that correct and make sure it's correct all the time? It will not make an unknown business known. It's not... It's inexpensive to put in place. But heavily marked up pages with nothing worth quoting on them will... You've got to get unique information out into the public domain.

And code it correctly. So then we move on to the corroboration. We've talked a little bit about this already, so I'm not going to spend loads of time on it, but look for your trade and professional bodies. Can you be in those directories? But make sure you're listed correct, collect correctly. A lot of businesses I see, they listed their business five, 10 years ago in a trade directory. Go back and make sure it's right or add to it. Add information. Add something new. Can you get in regional and trade press?

Even if that's, something to do with a charity event or some other piece of news, can you be listed and for what reason? And can that then talk about your business? Can you get into podcasts? Yes. Try. Please do. If you are the expert, don't be afraid of it because you're going to need to do it and pull those transcripts and make videos and host them on your site. Not just on the platform where they're left. Please bring those into your own site and your own social medias, your own YouTube, onto your own LinkedIn embed it into your own site.

There's a lot of ways you can do that. Make sure you have reviews. If you've got reviews, answer them. If you're getting reviews, get people to talk about your service. We've covered that. It's really important. Not just add five stars or 10, whatever it is, a 10 out of 10 rating. Get them to talk about your service. Partner up with clients. Get them to talk about you. So get case studies. Get them to publish how well you've helped them, et cetera. And the other thing is work on industry reports.

So make sure that there are studies that cite you. Work with the authorities and the influence in your, influencers in your area. Make sure that you're not necessarily... Don't go into arrangements where you're paying influencers necessarily because that may well be seen as paid promotion and get ignored as well. So the test is, would a sceptical human believe that this was written independently. If it reads like you paid to place it, assume a machine can tell, and it will be able to tell. So don't assume that... Don't just write it.

You have to have it corroborated and talked about by a third party in a favourable manner.

Which tactics should you stop paying for?

Video: 1:12:19

What to stop paying for. Keyword dense rewrites, no. Link buying, no. Mass-produced AI pages, no. One-off schema setups, no. Single screenshot reporting, no. And ranking reports as the headline, no. Pretty much everything that SEO agencies are still doing, we should probably ignore. The llms.txt. This is not really a thing. There are people out there selling this as a service. Just go on your own web property or on in your Claude or your ChatGPT or whatever and tell it, "Can you make me a, an LLMs file?" It doesn't hurt. It's not a service though, and if someone's charging you for one and calling it AEO.

That should tell you that their offer is weak, and they don't really know what they're doing. I think we've seen the share of ChatGPT mentions falling from 60% to 10% in the last six weeks of 2025. Just for people, it's nothing to do with LLMs now. It's more about the content, where it's placed, is it true? And hopefully I'm managing to get that across to you in this video.

Editor's note: the 60% to 10% figure the guide cites is the share of ChatGPT responses citing Reddit, which fell from about 60% in early August 2025 to about 10% by mid-September 2025 (Semrush). Source in the guide.

How do you measure AI visibility honestly?

Video: 1:13:50

So let's talk about measurement and the numbers that nobody actually wants to publish. And there is a, if there's no margin of error, there is, it's not... If somebody's saying there's no margin of error, or they can give you an extremely accurate answer to how often you are cited, it's not true. It is going to be a range. Because depending on the number of times that you flip your coin, and by that I mean if you take one of your desired prompts that your customers should be finding you for and you run it into one of the LLMs and you ask it that question, you'll probably get a different response like seven, 10, 20, 30, 40 times.

We don't know those answers at the moment. But the more times you run the query, the more true answer you will get out of it, but there will be a margin for error. So if you, if for example, you run a query 1,500 times per source, per cycle, i.e. 30 prompts, you should be able to get down to a one or two percent plus or minus. You get, you'll get to a true honest answer-ish at that point. If you're checking once or twice, it could be five, 10, 15% different.

And it's not going to be true. If you're running this yourself, I've got tools to do this, but you'll get an idea. You won't get the truth. I think that's what we're saying. So people talk about share of voice. And how many times their business was named across the prompt set. That's a good measure. How many prompts have you won? So out of how many times you were searched, how many times were you, did you win? Were you above all of your competitors? Were you cited as the source?

And how often are your pages linked as the source? That's interesting. And then the accuracy. So is what the engines say about you actually correct? Because if it isn't, make sure you get that sorted. And more important than anything, when somebody does land on your site or makes an enquiry, we are very quickly moving into a world of can you attribute a piece of business or enquiry to AI. We already know that those AI visitors are far more likely to convert. Now it is more important than ever to add in attribution into your mix and find out exactly where the enquiries are coming from, because then you'll know when to where to place your effort, okay?

So Google's AI Overview report you'll see many people never end up coming through at all. We are entering a no-click universe, so people are getting the answers before they ever end up on a site. We're seeing that things are wildly different at the moment. Like it is the Wild West. But published sources sell services in an area. Quote the range, not the headline figure. Now, more more organic clicks for brands cited in AI Overviews, 35% at the moment. This cannot be... This can't be a fluke.

Conversion from ChatGPT referrals against 1.7% organic, 15, nearly 16%, so 10 times better in this example. And then 1.81% against 1.39% across 94 e-commerce brands. So it seems like that's lagging behind to some extent because we're not really seeing, for example, Google Shopping in Google AI Overviews at the moment. And if people are in shopping mode, let's say, we all know that people end off in, Amazon or they end in, they go through shopping. That's the part that's not really suffering yet, and I say yet.

So the key thing is, remember the query fan-out. People are already, when they arrive from an AI answer, they have already asked normally several clarifying questions. So they arrive further along. Clearly, they arrive much further along the buying journey.

What is overhyped?

Video: 1:18:51

So the honest counterweight to all of this and the 90-day cycle. The parts that are overhyped. So there are three things worth being straight about after all of this. 1% or thereabouts, or less than 1% of total web traffic is currently AI referral, roughly growing at 1% a month. This could flip, spin, grow very quickly. It will, in however long we don't know, become the norm. It's only a small channel at the moment. Everybody's panicking to be ready for it. There's probably 10 times more marketing dollars going into AI, AEO, GEO search and positioning than in anything else at the moment.

It is the buzz. It is the biggest thing. It's going to be the biggest thing in 2026, 2027 unquestionably. Google is half right. Its guidance says that you don't need to do anything special. All the SEO guys, they're all running around saying that the fundamentals are still SEO. You still do need to do the fundamentals, but a lot of the SEO businesses don't even do the fundamentals, and a lot of what I've talked about in this video hopefully is helpful, and you can see what some of the new fundamentals are.

But nobody can guarantee citation, not even me. I would love to be able to do that. It's the engines that choose who gets cited. And what sources they notice, and each of the models is different, and they're all clearly pulling from different databases, like Muse. No one really talks about Muse is pulling a ton of its data from WhatsApp. From Facebook Messenger, et cetera. Mr. Zuckerberg has his own data, like Mr. Bezos has his own data. Google has its own data. All of these platforms, they're all competing now for our now, our attention and who is the best.

What should you do first?

Video: 1:21:09

So, what can you expect? What should you expect in terms of speed to actually achieve anything. In a couple of weeks, you should be able to get all of your basics done. With your prompt set and see where you are. You can put an enquiry form on your website, or you can start implementing where did you hear about us, how did you come to us, and start looking at tracking in your Google Search Console where some of that traffic's coming from. Is it coming from AI already?

Make sure at that point in the next couple of weeks, does every description agree? You might want to do some of this in parallel and speed things up. So fix all of your organisation and person schema. Build the real author pages and correct anything that's wrong. That's a really good and fundamental step, 'cause if you've got things broken at the front end, whatever you do with all the quality content that I'm going to ask you to produce, you could already have, put a nail in your coffin.

You need to start building answer-first content as key pages within your site. Publish immediately the high value questions and transcripts and videos. I would love to see you one of the best pieces of advices I can give you is put all of this into video. But don't be tempted to go too broad, like we talked about earlier. Don't make a video where you start out talking about let's say the, recruitment agency in Manchester who's looking to work with software, is a specialist working for software recruitment, and then go into more generalised stuff in the later part of the video.

Make it specific, okay? If you can build those videos and then get the transcripts, publish that on multiple web properties and get it shared, jackpot. I've got an example of a client. We published a video last week, put it on all his web properties, put all the schema in it, shared it, even ran a few ads to it. But in 48 hours, top of Google, number one. Top of AI citation, number one. Multiple citations across many different queries. So where this differs to the web fundamentally is Google would say, "Well, if you want to optimise for a keyword or key phrase, you can do that." You might come up number one if you do all the work. But if you build a piece of content that is unique and can be cited, answers questions around a topic and is specific, you can start coming up for multiple long-tail fan-out queries for that one piece of content, and I've proven that it works. 100% it works. You've just got to do the work.

Then go out and get it mentioned. So go out and talk podcasts. Go and go to blogs and industry publications and get them to talk about this new piece of information or some statistics. Go and do a survey. One of my clients we, is working in recruitment sorry, in estate agency, and we've, we're running a study where we're going to go out and enquire, so send enquiries to over 100 in estate agents for properties out of hours, and we're going to then measure how quickly those respond.

We'll then turn that into a white paper, and all of that data then becomes public knowledge, so we can show the estate agency industry how quickly other estate agencies respond to enquiries and how they can do it better. Because the client in question is Stitch AI. They run a WhatsApp tool, and one of their key benefits to estate agents is that you can automatically respond to enquiries within five minutes. Well, if we see in the data, and I don't know what that'll turn out like yet, but if we see in the data that typically half those estate agencies don't reply to their messaging until the following morning. There's a huge opportunity to show a piece of data which is unique as a study, and then other estate agencies will see it. It hopefully gets cited, and we can send that out to news and PR, and then we're getting loads of links, loads of citations, and it all comes back to us talking about WhatsApp. So it's not rocket science, it's just hard work. And then once you've done all the work, go out, measure again, measure those measure the prompts that you've tested. Or you wanted to come up for. Make sure you pick the right ones. Don't go broad again. Go with what is going to work.

So look at the rest of how we work with this.

How do you spot a real AEO supplier?

Video: 1:26:24

So AEO and GEO some people you can't tell the difference. What were you doing before this was a category. This is new. Or it's reasonably new. AI is shifting the goalpost so quickly. I probably two or three months ago, I was looking at the SEO industry trying to just say, "Well, AEO and GEO are SEO." And now we're seeing some of the biggest names in the industry moving their offerings over to AEO and GEO because they can see that, for example, Google is not sending people to search anymore.

A good chunk of those people are just getting pushed straight into an AI Overview or straight into the AI chat. So it's moving very quickly. You can't necessarily show proper multi-LLM engine tracking at the moment against a defined prompt set. I can do it in a few, but I don't know that there's anything that does it properly across all of them. The problem being that if you want to track, say, 30 prompts across six or seven different LLMs, and you want to go and test it, multiple times a month across 30 or 40 different runs, like we showed 1,500 to get a one to 2% tolerance, you're talking hundreds of dollars in data credits to do that.

So to do it properly, there's very few that do it. GEOForge, one that I'm working with, does a very nice job on ChatGPT and Google AI. Some people might they might come back and they just say, "Look, you need Schema. You need a one-off setup." I don't think you should be working with those type of people. They might be telling you, they might be telling you that a third-party cooperation doesn't matter. It absolutely does. Hopefully, I've got that across in this video that the truth is where we're heading. Is the truth confirmed? Are you the source of the knowledge, and can you be first to do that.

Most of the people won't... Like all the SEO industry, I won't say all, 99% of the SEO industry will say link building, link building, link building. They don't normally do third-party cooperation or PR, and the PR industry is where this is at. I think it's a huge opportunity for public relations firms to go out and grab a chunk of brand marketing, which it should've been all along. And if you're looking at an SEO agency or somebody says they're doing AEO and GEO, but they're going to say that they're giving you a monthly rankings report rather than share of voice and citation and actual attribution of click be very careful because if people are still talking monthly ranking reports, that's frankly not what you want.

How long does it take to get cited?

Video: 1:30:01

How long does it take? It can be quite quick. I've managed to see citations in 30 to 40 days. I've seen them... Once we've been working on a client, I've seen them in 48 hours, but that takes a lot of content sharing, hard work, outreach. So realistically 30 to 60 days because you're not seeing the models refresh much quicker than that. If you want a meaningful and, well, I would say within four to six months, you can really, I would think, get to the point where you can dominate a market, depending on which market you're in. Obviously, some are far more competitive than others. So depending on how much competition and how much work's getting done in your industry, it could be a couple of three months.

It might be six months. Nobody... If anybody promises you anything, be careful and mindful that nobody is promising anything in this field at the moment. So anyone fast, promising faster than that they're either very lucky or they could well be lying and selling you hopium, okay? So I'd encourage you, find the gate, that you're failing before you spend a penny on it. Access, retrieval, extraction, and trust. Measure it. Go and do a sample. Do you come up? Where are you? Is it important? What should you come up for?

Where is the rest of the industry that you're in? But publish something that's true that nobody else knows. That's the biggest piece of advice you're going to get out of this video. And that has not changed since part one of the internet. And if we can be the knowledge and be helpful, you will win. Hopefully that... I've managed to get that across.

Where can you read more?

Video: 1:32:11

So everything that I've talked about, where I've talked about that I've published since 2019, I have got that in my receipts pages within Question.Marketing. Please go and read. There is a full guide. I have got a full guide to how you can work with answer engine optimisation, get cited. There's a lot more to it. Like, I couldn't get everything into probably an hour plus video. I would have loved to. Follow me and look, if you want an audit, I am looking for clients to work with at the moment, and I can give you a proper roadmap and we can have a... Look, just have a chat.

I'd love to talk to you. So hopefully that video's been useful. This has been part two of my AEO masterclass. Follow along. Things will change quickly and this video may well be completely outdated in a couple of months. Who knows? It's a very interesting space to be in, and that's what fascinates me. Somebody asked me what makes me tick. I'm a nerd. I love this stuff. So hopefully speak to you soon.

Read the full guide

This transcript follows the video. The written guide goes further on each gate, with the testing sheet and the published sources: How to get recommended by ChatGPT and cited in Google's AI Overviews: the definitive guide.

Simon's dated articles on answer engine optimisation, published since 4 October 2019, are on his page at Question.Marketing. For an AEO audit, start at Question.Marketing.

Simon Young, Founder of Question.Marketing. 28 September 2026.