AI Search (AI SEO or GEO) is transforming how people discover brands. PR agencies are perfectly placed to lead the charge.
In this webinar, Josh Blyskal, who leads AEO Strategy & Research at Profound, and SUSO’s Partner Growth Manager Cara Corbett share actionable tactics on how to optimize websites for AI search.
➡ What AI systems actually look for when evaluating content
➡ How to structure and optimize content for maximum AI discoverability
➡ The must-have elements of Technical SEO for AI Search
➡ Trends from Profound
Scroll down to watch the recording.
Click here to access the presentation slides.
Webinar Transcript
00:00:00.009 –> 00:00:01.189
Okay, cool.
00:00:01.229 –> 00:00:07.956
Well, we’ll get started because we do have limited time. But yeah, thanks so much everyone for joining today.
00:00:08.275 –> 00:00:11.599
I’m your host, Cara Corbett. I’m the Partner Growth Manager at SUSO Digital.
00:00:12.359 –> 00:00:22.468
We’re an SEO firm that specializes in agency partnerships, meaning that we work with agencies of all sorts and sizes, including PR agencies, as their technical SEO delivery partner.
00:00:22.507 –> 00:00:27.652
So welcome to our webinar, Optimizing for AI Search, the PR Agency’s Tactical Playbook.
00:00:28.172 –> 00:00:31.835
And this is the second webinar of three in our AI search series.
00:00:31.896 –> 00:00:38.442
So if you tuned into our first webinar, which was all about the AI search opportunity for PR agencies, it’s really good to see you again.
00:00:38.503 –> 00:00:42.606
And of course, if this is your first webinar with us, we’re really happy to have you joining us today.
00:00:43.508 –> 00:00:55.795
Today’s session is really all about getting super tactical. We’re going to really dig into all the ways that you can increase your client’s visibility in AI search engines beyond just doing traditional PR tactics.
00:00:55.835 –> 00:01:02.939
So especially as this relates to structuring your own content and also your technical optimization of your website.
00:01:03.420 –> 00:01:08.064
And to help me break this down, I’m joined by Josh Bliskell, who leads AEO Strategy.
00:01:08.084 –> 00:01:14.950
There’s another acronym for you. It stands for Answer Engine Optimization. He also digs heavily into the research at Profound.
00:01:15.031 –> 00:01:22.257
So Josh is a top voice in AI search optimization, and I know he’s going to provide a lot of color and data to today’s conversation.
00:01:22.739 –> 00:01:28.183
So, Josh, do you want to introduce yourself and tell everyone a little bit about Profound and what you do there on a day to day basis?
00:01:28.780 –> 00:01:29.560
Absolutely, yeah.
00:01:29.579 –> 00:01:36.884
Thanks for having me, Cara. Yeah, Profound is the first platform that enables marketers to understand, take action on, and tangibly improve their visibility in AI search.
00:01:36.944 –> 00:01:45.668
So what we do is we interrogate answer engines at the scale of millions of questions every single day to understand where brands appear, where products appear.
00:01:46.069 –> 00:01:56.694
We understand the influence layers, everything from what makes a URL, what makes a page actually get cited in answer engines and influence what the answer engine says across blog, affiliate, owned, social content.
00:01:57.093 –> 00:02:38.633
and my role specifically at Profound is head of AI strategy so that means i’m really leading on the front end of what these actual strategies look like how we can develop tangible real tactics and strategies to influence how answer engines AI search engines basically cite content brands and think about how we can build out visibility defensively in these new answer engines it’s awesome yeah lots of color you’re going to provide for sure so just so everyone knows we will have a little bit of time for questions at the end of today’s session so feel free to pop them in that comments chat in the top right corner there and then we’ll be sure to try and get to everybody’s at the end So here’s what we’re gonna cover today a little more deeply.
00:02:39.293 –> 00:02:43.197
First, we’re gonna discuss how large language models work at a very digestible level.
00:02:43.257 –> 00:02:49.382
Don’t worry, we’re not gonna make it too technical, which is fundamental to understanding how to optimize for them.
00:02:49.861 –> 00:02:56.647
We’ll look at where and how these tools get their answers, how each one differs, and then what that means for your client’s visibility strategy.
00:02:57.127 –> 00:03:07.858
And then we’ll move into our first tactical session, which is all about structuring content for discovery and how to package your own content in a way that AI systems can easily understand and extract for answers.
00:03:08.399 –> 00:03:21.731
And next, we’ll cover the technical SEO and crawler accessibility, which might sound a little bit geeky, but it’s absolutely crucial because if AI systems can’t actually access your content due to blockers, then it will never have the chance to be factored.
00:03:22.147 –> 00:03:32.019
And then finally, Josh is going to share real trends from Profound’s proprietary data around what they’re seeing is actually working to get their clients more brand mentions and citations in AI systems.
00:03:32.079 –> 00:03:40.248
So let’s get into it. And first let’s start by setting the stage. Say it’s, you’ve just sent out a press release on the wire for your client.
00:03:40.588 –> 00:03:45.093
And it’s getting picked up by relevant news outlets and then generating some interest from journalists.
00:03:45.294 –> 00:03:51.800
And if you optimize your press release with well-researched keywords, you might even have it show up in Google’s organic search results.
00:03:52.281 –> 00:03:56.245
And whether that’s one of the ten blue links on the page or in that news carousel.
00:03:56.564 –> 00:04:04.272
Now, from an organic perspective, this was considered a success as you’ve achieved visibility not only from news pickups, but also on search engines.
00:04:04.912 –> 00:04:13.959
But in 2025 and beyond, organic success looks a little bit different. It’s not just about getting stories written about your clients or taking up space on the Google results page.
00:04:14.459 –> 00:04:20.403
It’s also about whether that story helps your client become a recognized and recommended brand by these engines.
00:04:21.064 –> 00:04:24.187
Increasingly, people are using tools like ChatGPT instead of Google.
00:04:24.247 –> 00:04:32.593
So our opportunity for organic visibility is really now lying and getting that story cited, which is when the LLM uses your brand as a source and links to it.
00:04:32.634 –> 00:04:38.596
So we can see an example there, or if it uses their name in the answer.
00:04:38.636 –> 00:04:45.680
So we can see in that listicle style response, State Street Corporation is both mentioned and cited and their press is mentioned.
00:04:46.641 –> 00:04:54.146
So really their name is front and center at the exact moment of decision. And this is really the new gold standard for visibility in the AI age.
00:04:55.163 –> 00:05:11.194
And if you’re at our last webinar or maybe you’re just really tuned in to the AI search discussion, you already know that earned media and other off-page trust signals are some of the most important pieces of AI search visibility because these models are looking for authoritative and trustworthy sources for their answers.
00:05:11.653 –> 00:05:17.858
And that’s really created a rebirth in demand for PR because you guys have already been doing these things for years and years and years.
00:05:18.619 –> 00:05:39.382
if you also tuned into our last webinar you know that earned media and traditional pr tactics alone are not going to get your clients results in this domain there’s a whole other technical side to AI search where you can optimize your content and website in a way to help AI easily find interpret and include those answers And the reality is most PR agencies lack this skill set.
00:05:39.442 –> 00:05:49.555
So the message today is simple. PR already has that storytelling side nailed down. The next step is mastering that technical layer. And that’s exactly what we’re going to walk through today.
00:05:50.523 –> 00:05:58.646
But first let’s talk a little bit about how these AI engines work, because of course, if we want to optimize for them, we have to understand their inner workings a little bit.
00:05:59.326 –> 00:06:13.163
LLMs don’t work like Google. They don’t index the web and spit back a ranked list of blue links for users to search through based on the keywords that they, you know, they typed into the bar and then find their answers, do their own research, and really come to their own conclusions.
00:06:13.723 –> 00:06:21.305
Instead, they provide you the answer. So it’s no longer about where you’re ranking in the pile. It’s now whether or not you’re actually part of that answer at all.
00:06:22.165 –> 00:06:28.387
And so if AI doesn’t rank and it just gives answers, the obvious next question is, where does it actually get those answers from?
00:06:28.507 –> 00:06:35.870
And really, there’s two main sources that we’re concerned with here when it comes to optimization, which are training data and grounding data.
00:06:36.649 –> 00:06:45.101
So each of these models have their own training data sets that are sourced from things like the web, digitized books, articles, and public reports and documents.
00:06:45.422 –> 00:06:53.713
And unlike humans who continuously learn or Google algorithms which are constantly updated, LLMs have their knowledge effectively frozen at a specific point in time.
00:06:54.613 –> 00:07:01.637
you know, ChatGPT’s latest model, ChatGPT five, their knowledge cutoff is as of October first, twenty twenty four.
00:07:02.358 –> 00:07:06.560
So it actually doesn’t have any knowledge of anything beyond this point in its training data.
00:07:06.639 –> 00:07:14.605
So, for instance, you can test it out yourself. If you ask the latest model who is the president of the United States based on your knowledge, it will tell you Joe Biden.
00:07:15.646 –> 00:07:23.651
So with all this in mind, if your client’s brand was already being covered, mentioned consistently and part of the public conversation, chances are they were likely part of the training data.
00:07:24.271 –> 00:07:27.273
And if they weren’t, you cannot just go back and add them later.
00:07:27.973 –> 00:07:34.817
That’s why long-term PR presence really pays off. It really builds that awareness and background knowledge that AI models learn from in that training phase.
00:07:35.517 –> 00:07:41.761
And then that second main source is the grounding data, which is that live layer that some models use when they don’t have all of the information.
00:07:42.261 –> 00:07:59.850
So, for instance, if you ask ChatGPT that same question about who’s the president of the United States without that added context of based on your own knowledge, it’ll actually send out these automated robots or crawlers to query the live web and then gather up to date sources to fact-check and cite in order to ground that data.
00:07:59.930 –> 00:08:06.795
So this process is called retrieval-augmented generation (RAG). I’m not going to say that again.
00:08:07.235 –> 00:08:21.144
It’s a tongue twister, but this is really where we have the biggest influence when it comes to showing up in these engines, because if our client’s content and website is set up in a way to aid these AI crawlers, they’ll have an easier time seeing them and using them in their answers.
00:08:22.204 –> 00:08:28.750
So what exactly are these crawlers looking for when they decide what grounding sources to use for their answers?
00:08:28.790 –> 00:08:35.576
And we’ve really condensed this down into four main components, which are relevance, authority, context, and patterns.
00:08:35.635 –> 00:08:40.860
So relevance is about how well does your own content match the user’s actual prompts?
00:08:41.500 –> 00:08:45.322
Think about it like, does it contain the right keywords or entities?
00:08:45.361 –> 00:08:52.845
So entities are things like specific names, places, or technical terms that signal to the AI that it’s related to that topic.
00:08:53.284 –> 00:08:59.567
And now once it finds the stuff that’s relevant, it then looks at authority and trust, essentially asking, can I rely on this source?
00:08:59.988 –> 00:09:09.812
Just like a journalist wouldn’t rely on a random blog, AI also leans on trusted sources with factual information, proper authorship, and content that cites actual studies or official data.
00:09:10.768 –> 00:09:13.428
Now for context, think of AI like a giant mind map.
00:09:13.808 –> 00:09:19.451
It’s not just storing facts, but it’s also storing the connections between entities and topics.
00:09:19.850 –> 00:09:24.133
It wants to know, does the brand and its content clearly connect to an industry or niche?
00:09:24.513 –> 00:09:54.500
So for example, if your client makes medical devices, but their website and press only talks about awards, or funding but never about healthcare patients or hospitals to the AI the brand might actually look like a finance or startup rather than a healthcare related brand so that strong context makes sure that AI knows exactly where your client belongs in the bigger picture And finally, patterns are about consistency, meaning does your client’s name, product description or key messaging show up consistently across the web in many places?
00:09:54.580 –> 00:10:00.702
Because when this is done, AI starts to gain the confidence about the source and it’s likely to share that repeated narrative.
00:10:00.763 –> 00:10:10.009
So for PR agencies, this means your coverage and content. It’s not just telling a story to people. It’s also shaping the signals that AI systems use to fuel their answers.
00:10:10.956 –> 00:10:20.928
So Josh and I are gonna talk about some of the ways that the main AI model is different from how they work, visualize information and what sources they use and which platforms are effectively best for your client’s needs.
00:10:20.989 –> 00:10:32.217
So ChatGPT is much more of a paraphraser. It’s really good at pulling together clean conversational answer out of multiple different sources. But unfortunately, it doesn’t always show these sources.
00:10:32.278 –> 00:10:39.140
So that means your client’s content could be in the response, but without their name or their website link attached to it.
00:10:39.181 –> 00:10:48.105
For its grounding data, it uses a hybrid index using SERP API. So it’s often pulling from Google or Bing through that API.
00:10:49.044 –> 00:10:59.912
Now, Gemini powers both Google’s AI mode and AI overviews. So AI overviews are those quick snapshot summaries at the top of traditional Google search results. And AI mode is that more conversational style model.
00:11:00.831 –> 00:11:09.197
And the good thing about Gemini is it actually uses Google’s search index, which obviously they spent the better part of the last two decades perfecting with all of their algorithms.
00:11:09.817 –> 00:11:16.182
But it’s also heavily interlaced with everything in the Google suite, like your search history, your Gmail, your location, your preferences.
00:11:16.562 –> 00:12:04.302
so i like to think it’s the most personalized LLM it’s great for those localized searches if you’re trying to find something local and something interesting to note is that google’s AI mode uses this method called query fan out where it essentially breaks down the prompt or query that you put in into multiple related sub prompts or questions in the back end and then it pulls answers for all of those questions from multiple different sources and essentially synthesizes it into one comprehensive answer so that’s why when you ask it one question you get more than what you ask for because it’s already anticipating those next follow-up questions and then perplexity is by far the most transparent of all the three LLMs it always shows its sources clearly which i personally really like because i’m a You know, I like to fact check things myself and just to make sure it’s not hallucinating.
00:12:05.384 –> 00:12:08.205
They actually developed their own index, which is smaller than Google’s.
00:12:08.264 –> 00:12:15.788
But one of their people said that they’re focused on serving the most popular and high quality content from sources most likely to be relevant and trustworthy.
00:12:15.849 –> 00:12:23.192
So, Josh, how about you talk to us a little bit about some of the citation data that you have seen for these and where you think that that fits in?
00:12:23.904 –> 00:12:34.033
Yeah, nothing is more evident right now in the differences between the infrastructure and underlying fundamentals of each of the answer engines than the top sources that they cite and the way that they cite those sources.
00:12:34.472 –> 00:12:36.894
So a great example right now is in ChatGPT.
00:12:36.914 –> 00:12:42.659
Not many people know this, but Wikipedia is actually the number one cited domain in ChatGPT, period.
00:12:42.679 –> 00:12:47.023
And that sits at around ten percent of all citations. That’s massive.
00:12:47.222 –> 00:12:52.967
Wikipedia acts like a legitimacy layer within ChatGPT, legitimizing companies if they do have a Wikipedia page.
00:12:53.447 –> 00:12:57.212
And then not being able to find companies if they don’t. It’s a really interesting situation there.
00:12:57.712 –> 00:13:01.817
Reddit’s also the number two domain. Interesting thing about Reddit, Reddit’s seeing a bunch of shifts lately.
00:13:01.837 –> 00:13:03.178
We’ll talk about that in a little bit.
00:13:03.219 –> 00:13:09.947
But very interesting there because obviously ChatGPT is going to be the most used answer engine, the most used pure answer engine there.
00:13:10.467 –> 00:13:14.052
On the Google side, very interesting. There’s been a ton of different headlines about this.
00:13:14.601 –> 00:13:22.168
Google AI mode, Google AI overviews, both are very obsessed, some could say obsessed, with actually citing Google SERPs.
00:13:22.609 –> 00:13:27.073
So it’s actually citing Google search results pages rather than citing any other page.
00:13:27.514 –> 00:13:34.178
And that kind of continues the trend of Google Really trying to keep its own interface and its own infrastructure self-contained.
00:13:34.318 –> 00:13:37.500
The second most cited source in Google doesn’t help either. It’s YouTube.
00:13:37.821 –> 00:13:44.605
Google is actually able to read in YouTube transcripts and full YouTube data to power its AI search results, which is unique to it specifically.
00:13:45.065 –> 00:13:47.726
Lots of good YouTube citations there. And that’s very interesting.
00:13:47.767 –> 00:13:51.729
And it’s a very unique vector for brands thinking about their different social plays.
00:13:52.068 –> 00:14:35.941
Reddit’s also obviously there in um in the google models it’s the third most cited source and then for perplexity very interesting perplexity has a massive bend towards Reddit so when we’re looking at Reddit and perplexity it sits at around six percent of all citations uh when we look at uh the second most cited source in perplexity which is youtube that’s around two percent so it’s a three x bend towards Reddit um perplexity is very ugc driven perplexity has kind of always been about capturing the community idea so if we’re looking at running shoes for instance perplexity is going to show you across you know Reddit across um you know quora across any different answer site where people are talking about running shoes what the community is saying and that’s that’s a unique advantage that they have right now Yep.
00:14:35.961 –> 00:14:46.585
So yeah, the big takeaway is this. If you’re going to do GEO or AEO or just AI search optimization, whatever acronym you want to call it, you got to know where your audience is and then optimize for that specific platform.
00:14:47.485 –> 00:14:50.826
But Josh, you said there’s some interesting changes happening pretty recently.
00:14:51.549 –> 00:15:00.678
Yeah, Reddit has seen a pretty massive decline in its number of citations and something that’s not captured on this graph, but it’s still all the more relevant is that Wikipedia has also been crunched as well.
00:15:01.038 –> 00:15:11.830
And so this is all happening at a time when in the traditional SEO world, we’re actually seeing some pretty big changes from Google in terms of them deprecating some search parameters that usually allow people to search deep in the search results.
00:15:12.309 –> 00:15:18.552
So Google is changing its fundamental infrastructure. Literally, it did that on September 11th. You can see where that citation drop off occurs.
00:15:18.892 –> 00:15:23.193
We know ChatGPT launched GPT five and they’re also testing some new things with commerce.
00:15:23.214 –> 00:15:26.995
So they’re changing how their rag model works and how all those different things basically build out.
00:15:27.475 –> 00:15:31.096
So we’ve got plenty of plenty of volatility happening in this phase.
00:15:31.669 –> 00:15:41.633
The end result, which is hard to attribute to any one of those causes, is that Reddit’s gone from being right around twelve percent of all citations to being right around one percent of all citations.
00:15:42.092 –> 00:15:50.735
The interesting thing there is that despite that decrease, Reddit is still the number two ranked cited source within ChatGPT.
00:15:51.176 –> 00:16:01.258
So it’s not that Reddit’s disappeared from influence, it’s that the magnitude of that influence is just much more akin to its peers and closest competitors than it was previously.
00:16:01.278 –> 00:16:08.721
I mean, Reddit was just flying sky high before that, but very interesting trend to stay aware of. And it just shows us how much volatility exists in the space.
00:16:08.541 –> 00:16:16.563
Yep. SEO people, they can never get away from volatility. There’s always going to be algorithm changes and all that. So that’s true. All right, Josh. Yeah.
00:16:17.263 –> 00:16:31.490
And I think the interesting thing here right now is that if we think about the relationship, and this is some Profound data here, if we think about the relationship between earned, owned, and social and competitor content, we can really start to see some of these groups kind of coalescing.
00:16:31.870 –> 00:16:42.654
So the biggest kind of content, the biggest group of content, period, in your AI search world is going to be the earned content because this is where the models co-correlate and build consensus.
00:16:43.134 –> 00:16:48.038
Answer engines want to see the same narrative many times across the internet in different places.
00:16:48.057 –> 00:16:50.958
They want to see that a lot of different people say the same thing.
00:16:51.339 –> 00:16:56.701
But here’s the thing, your own content is your fortress. It is your fortress of solitude.
00:16:57.081 –> 00:17:03.044
It allows you to build out the accurate, clear, concise, detailed information about your brands, products, and services.
00:17:03.485 –> 00:17:34.419
in a way that the answer engines will use as the source of truth and as the source of record where the earned media is where you can go out and prospect it’s where you can go and pioneer find new narratives talk about new things product launches and understand how to build a cohesive narrative across both those things so we make the facts happen on our own site we can start building out the narrative and then we disseminate that out across the earned media space across social it’s really exciting to see this kind of makeup here it’s a lot of potential for brands All right.
00:17:34.960 –> 00:17:38.382
So let’s get into why we’re all here, which is to discuss the tactics.
00:17:38.402 –> 00:17:44.526
So this chart that we put together shows the main tactics that have been shown to influence visibility in AI search.
00:17:44.905 –> 00:17:47.768
And we’ve mapped them all by impact, timeframe, and effort.
00:17:47.867 –> 00:17:52.791
So we’re going to walk through these tactics in the next few slides, break down what they mean in practice.
00:17:52.832 –> 00:18:03.338
And then each tactics slide will also have a link to a useful resource or a statistic or research paper for further reading, which you can then view when we send out the PDF after the webinar.
00:18:04.358 –> 00:18:17.221
The bigger point is this, we know PR agencies have strong storytelling abilities, but most of the technical levers that you see here like structured data, site hygiene, and server log monitoring usually sit outside of the PR toolkit.
00:18:17.821 –> 00:18:29.084
There’s a little bit of a capability gap, but really our goal with this session is to make sure that you don’t lose that ground to show you how you compare your strengths with these technical foundations and then really own the AI search opportunity.
00:18:30.037 –> 00:18:35.259
So let’s start with content. And the golden rule of structuring content for AI visibility is really simple.
00:18:35.699 –> 00:18:39.099
It’s make your content easy for AI to read, trust and repeat.
00:18:39.180 –> 00:18:44.521
So if your content is messy, vague or buried in jargon, AI is just going to skip it.
00:18:44.961 –> 00:18:48.044
And if it doesn’t come from a credible source, AI won’t trust it.
00:18:48.203 –> 00:18:51.944
And if it’s not structured clearly, AI won’t know how to use it.
00:18:52.566 –> 00:18:57.453
So what AI wants is clean, digestible building blocks that it can lift straight into an answer.
00:18:57.554 –> 00:19:04.806
So let’s look at some practical tactics that you can start employing right away to make sure that your content is packaged in exactly that way.
00:19:05.753 –> 00:19:13.059
So when it comes to making your content easy for AI to pick up and reuse, here’s some three simple things you can start doing right away.
00:19:13.559 –> 00:19:15.801
So first, use question style headings.
00:19:16.762 –> 00:19:19.125
AI is literally trained to answer questions.
00:19:19.204 –> 00:19:24.388
So if your headings mirror how people actually ask them, they’re much more likely to be sourced in answers.
00:19:24.449 –> 00:19:34.596
So in PR, that could mean shifting release headline from brand launches a new fintech tool to how does brand’s new platform help small businesses manage payments?
00:19:35.057 –> 00:19:43.902
And on the client side, it might mean titling your blog posts or your explainer pages with natural question based headings like what are the benefits of switching to solar energy in 2025?
00:19:44.922 –> 00:19:48.463
The next is breaking explanations into bullets and clear summaries.
00:19:49.265 –> 00:19:51.645
AI doesn’t actually want to dig through dense paragraphs.
00:19:51.726 –> 00:19:56.167
It looks more for clear, factual points that it can use to directly answer the query.
00:19:56.288 –> 00:19:59.750
And that’s because AI doesn’t actually work like Google.
00:19:59.769 –> 00:20:02.612
They don’t read the full page when it’s referencing sources.
00:20:03.132 –> 00:20:08.019
Instead, it’s breaking content into passages or chunks that directly relate to the query.
00:20:08.099 –> 00:20:17.994
So if your content is formatted in a predictable and skimmable way, it helps the models access what they need more effectively, and you’re more likely to be cited.
00:20:18.576 –> 00:20:26.019
And then third is adding FAQ sections because AI is trained on natural human language and we know that it’s trying to answer questions.
00:20:26.078 –> 00:20:33.281
So Q&A snippets do a great job at mirroring this format and they do specifically well on product pages and service pages.
00:20:33.342 –> 00:20:36.623
But I know Josh has some other things to add on this as well.
00:20:37.065 –> 00:20:43.287
The interesting thing here is that answer engines, and this is a little tidbit here, they pull eighty to a hundred characters of content from any given website.
00:20:43.467 –> 00:21:05.234
And so by actually building out the question answer format and concisely aligning that content to the things that people are directly asking and using bullets and summaries, what you’re doing is you’re allowing the answer engine in that eighty to a hundred character snippet to extract the direct, self-contained, most meaningful part of that text that directly answers the user’s question and give that back to the user.
00:21:05.595 –> 00:21:08.958
The answer engine can see, all right, here’s the full page content.
00:21:09.258 –> 00:21:14.883
But what it needs to do is to be able to easily find the direct question that answers the user’s question and then extract it.
00:21:15.163 –> 00:21:21.068
If you have long, verbose paragraphs, if you have unstructured data, unstructured text, it’s going to be much harder to do that.
00:21:21.169 –> 00:21:23.671
The answer engine can’t just summarize the whole page all in one.
00:21:24.090 –> 00:21:32.679
So you want to provide a good landing pad for that answer engine to come onto that page and get what it needs to get, get in, get out, and then give the value to the user.
00:21:34.744 –> 00:21:41.508
And AI models also favor fresh content because they’re trained to prioritize sources that look current and actively maintained.
00:21:41.847 –> 00:21:43.828
And there’s actually some hard data to back this up.
00:21:45.089 –> 00:21:55.034
Right here linked on this is a recent Ahrefs study that shows that on average URLs cited by AI tools are about twenty five percent fresher than those in traditional search results.
00:21:55.114 –> 00:22:01.656
So ChatGPT is most likely to cite newer pages, while Google’s AI overviews, by contrast, are more comfortable citing older content.
00:22:01.717 –> 00:22:03.657
But freshness is still going to give you an edge.
00:22:04.218 –> 00:22:11.162
So the good news is, you know, PR teams already work like newsrooms, they’re fast, they’re reactive, they’re always tied to what’s happening now.
00:22:11.261 –> 00:22:13.643
And that rhythm is exactly what AI rewards.
00:22:13.722 –> 00:22:16.265
So continue to publish timely updates.
00:22:16.765 –> 00:22:23.568
And you know, this is second nature, press releases, campaign launches, newsjacking, you’re already putting clients into conversation as they happen.
00:22:24.048 –> 00:22:28.211
The extra step is just making sure that your clients mirror that cadence on their websites as well.
00:22:28.271 –> 00:22:34.255
So use their blog or their newsroom to post about product launches, industry awards or timely opinion pieces.
00:22:34.855 –> 00:22:38.718
And you can also refresh older posts instead of letting old content go stale.
00:22:39.397 –> 00:22:43.720
For PRs, this can mean repitching or updating older thought leadership with new data points.
00:22:44.121 –> 00:22:56.407
And on the client side, it’s about revisiting case studies and product pages or even evergreen blogs and then adding context like as of April, 2025, because these small updates keep content relevant in AI’s eyes.
00:22:57.230 –> 00:23:01.790
And then make sure to highlight recent stats because data carries more weight when it’s clearly dated.
00:23:01.871 –> 00:23:07.192
So you can strengthen, uh, strengthen coverage with phrasing, like according to March, 2025 survey.
00:23:07.211 –> 00:23:14.153
Uh, and you can update the client’s research pages, blogs, or resources to show physical, uh, publication dates.
00:23:14.252 –> 00:23:16.653
Cause again, AI notices those time signals.
00:23:16.673 –> 00:23:23.414
Uh, and we already talked about FAQs of course, because the models pull those answers verbatim from the sources that are easiest to parse, but.
00:23:23.835 –> 00:23:53.257
you should also maintain that um you should also maintain them and keep them up to date as well so that also means helping clients identify the questions people keep asking about their pricing their leadership sustainability whatever is topical and then ensuring the answers evolve with those questions on those service or product pages even the smallest edits will signal freshness to the AI models uh josh uh what are you seeing fresh coverage yeah i mean we’ve worked with hundreds of brands that have a great content catalog.
00:23:53.676 –> 00:24:03.343
And because of the way that Answer engine access these pages, the actual lift is relatively low to get that content updated for 2025, refactoring that content.
00:24:03.663 –> 00:24:11.749
We’ve seen probably three, four dozen brands in the last two to three months who come to us and they say, hey, we want to get optimized.
00:24:11.888 –> 00:24:13.930
You’re already ninety five percent of the way there.
00:24:14.309 –> 00:24:17.211
It really is about freshness for you because you have so much excellent content.
00:24:17.731 –> 00:24:23.253
If it’s not about updating the content for 2025, it’s just about saying, hey, this content’s already updated for 2025.
00:24:23.575 –> 00:24:24.535
Let’s get those signals.
00:24:24.555 –> 00:24:29.436
Let’s send those signals out to say, according to this 2025 study, putting 2025 in the title tag.
00:24:29.777 –> 00:24:38.040
Because another tidbit here on the technical side is that when that query fan out process that Kara was talking about happens and the answer engines go out to the Internet to look for results.
00:24:38.401 –> 00:24:47.867
One of the most interesting things that we’re seeing, and this is vertical dependent, but between ten and thirty percent of the time, answer engines like ChatGPT, We’ll add 2025 to the end of the query fan out.
00:24:47.907 –> 00:24:48.888
That seems benign, right?
00:24:49.469 –> 00:25:05.821
But the result here is that the actual pages that ChatGPT is looking for and pulls back, they’re structurally biased towards having 2025 in the title tag, 2025 in the meta description, sometimes 2025 in the URL, the the first eighty two hundred characters of the content.
00:25:06.362 –> 00:25:15.346
Basically, because of that 2025’s inclusion in the actual query fan out, You want to be aligned and have 2025 in your content directly.
00:25:15.686 –> 00:25:19.973
So date alignment is just a structural way to increase your number of citations.
00:25:22.256 –> 00:25:28.417
And when it comes to AI selecting their sources, surface level content is just not going to cut it as well.
00:25:28.978 –> 00:25:31.518
AI systems favor depth, credibility, and clarity.
00:25:31.597 –> 00:25:33.778
So here’s four ways to make that happen.
00:25:34.419 –> 00:25:37.980
And the first thing is having topical breadth across your client’s website.
00:25:38.019 –> 00:25:41.921
And this is also a really important thing in SEO, by the way, just traditional SEO.
00:25:41.961 –> 00:25:50.143
But you want to cover a good depth of content related to the industry because we know that AI maps relationships between entities and overall topics, right?
00:25:50.603 –> 00:25:55.405
And they’re using methods like query fan out, breaking those queries into multiple sub questions and intents.
00:25:55.546 –> 00:25:59.508
So in order to do this, we follow a hub and a spoke model for content.
00:26:00.189 –> 00:26:05.612
So you essentially have a hub page provides a high level overview that touches every core aspect of the topic.
00:26:06.011 –> 00:26:10.095
And then you have spoke pages which go into each of those core aspects in depth.
00:26:10.555 –> 00:26:18.799
And then, you know, so if we put that into context here, let’s say you have a fintech client that wants to put out a piece around the future of digital payments in 2025.
00:26:19.160 –> 00:26:20.340
This would be the hub page.
00:26:20.661 –> 00:26:26.204
And then some of the spoke pages could be about regulation and compliance, cross-border and global payments.
00:26:26.545 –> 00:26:29.945
security and fraud prevention, case studies, et cetera, et cetera, et cetera.
00:26:30.846 –> 00:26:33.048
And all of these would link back to the hub.
00:26:33.428 –> 00:26:37.529
And then if it made sense, some of the spokes could even link to one another.
00:26:38.309 –> 00:26:46.212
And all of this in-depth content will give the AI more signals and entry points for them to associate your brand with the topic and also answer more queries.
00:26:47.054 –> 00:26:52.175
You also want to use expert voices, of course, because AI looks at authority like we talked about.
00:26:53.336 –> 00:27:04.482
So the same way a journalist does, you know, a release that just says company launches new tool won’t carry as much weight as one that features a quote from a CEO or a chief scientist explaining why it matters.
00:27:04.583 –> 00:27:09.925
So for a green energy client, that might be an engineer explaining the carbon savings in this example.
00:27:10.925 –> 00:27:14.169
And next, keep messaging consistent across editorial.
00:27:14.209 –> 00:27:20.792
If your client’s product names and boilerplates or bios shift from one release to the next, it can actually confuse AI a little bit.
00:27:21.493 –> 00:27:30.961
But when the messaging is aligned across press releases and blogs and about pages, the system then gains confidence and it’s far more likely to surface the story in that consistent narrative.
00:27:32.098 –> 00:27:35.101
And finally, including definitions and examples in your content.
00:27:36.082 –> 00:27:38.243
AI is trying to explain things back to the user, right?
00:27:38.284 –> 00:27:42.386
So if you introduce a technical phrase, make sure to define it in plain English.
00:27:42.467 –> 00:27:49.333
So an example that I have in front of me here is a vague line like, regulation is a big challenge for fintech brands.
00:27:49.732 –> 00:27:51.292
That’s not teaching the model much, right?
00:27:52.134 –> 00:28:05.997
But if you change that and you say fintech regulation refers to the rules that govern digital services like payments and lending, for example, Europe’s PSD two law requires strong customer authentication for online payments.
00:28:06.477 –> 00:28:12.838
Now you’ve actually given AI two things that it can cite one a definition and then also to a real world illustration.
00:28:13.239 –> 00:28:17.880
So definitions make your content more retrievable for those like what is or what are questions.
00:28:18.299 –> 00:28:22.423
And examples give real examples for those give me an example questions.
00:28:22.482 –> 00:28:27.586
And together, those are dramatically increasing your chances of being cited in an AI generated answer.
00:28:27.685 –> 00:28:30.327
So the takeaway here is really, really simple.
00:28:30.749 –> 00:28:33.269
Comprehensive content doesn’t mean long winded.
00:28:33.309 –> 00:28:37.613
It means covering the context, showcasing authority and explaining it clearly.
00:28:37.853 –> 00:28:42.717
And then, of course, keeping the message consistent because that’s the kind of content that AI trusts and reuses.
00:28:43.557 –> 00:28:53.220
And now even as AI reshapes search, EEAT, which stands for expertise, experience, authoritativeness, and trustworthiness, this still matters.
00:28:53.960 –> 00:29:04.423
So for those of you that don’t know, EEAT was a key framework in the SEO industry, still is, that Google actually had human reviewers use to assess the quality and credibility of online content.
00:29:04.442 –> 00:29:09.384
So they would go through the website and they would manually look at pages and they would assess EEAT signals.
00:29:10.423 –> 00:29:13.945
And the difference now is that it’s not just humans reviewing these signals.
00:29:14.145 –> 00:29:19.247
AI is also considering who’s speaking and why they’re credible and whether their story is backed up.
00:29:19.307 –> 00:29:21.827
So with this in mind, you should do the following.
00:29:22.628 –> 00:29:25.049
The first thing is adding expert bylines and bios.
00:29:25.170 –> 00:29:27.549
And for PRs, of course, this is already in your wheelhouse.
00:29:27.611 –> 00:29:28.631
You’re placing your clients.
00:29:29.230 –> 00:30:00.634
uh in op-eds and interviews and thought leadership but you can take it further by making sure every piece has a clear byline and also a short bio that highlights their credibility so this also applies to your clients own blogs or resources on their site articles should have a real author not just an admin i’ve seen that quite a bit guilty and ideally the byline links to an actual bio page for that author And then on that bio page, include information that shows that they’re a credible and real source to be talking about this topic.
00:30:00.694 –> 00:30:11.210
So include a professional headshot, information about relevant credentials, publications and experience, and then link out to their social medias as well and any other credible content.
00:30:12.375 –> 00:30:20.358
And you should also ensure that your clients have strong about pages on their website so that you can help them tell their story in a way that’s consistent and transparent.
00:30:20.960 –> 00:30:31.525
You essentially wanna help the AI understand who’s behind the brand and what they do and what they stand for, because this makes it easier for LLMs to associate your brand with those topical clusters we spoke about earlier.
00:30:32.506 –> 00:30:40.569
And ideally this page highlights the company background, the history, the team, the mission, The location is important as well.
00:30:40.789 –> 00:30:44.752
And then any other trust signals like awards and accreditations and certifications.
00:30:45.173 –> 00:30:49.455
And we know PRs are already highly skilled at highlighting these credibility markers and pitches and coverage.
00:30:49.496 –> 00:30:53.739
So we just need to make it known on the website as well that the brand is recognized and respected.
00:30:54.444 –> 00:30:58.127
And then finally citing external resources and references.
00:30:58.209 –> 00:31:07.376
So PR people can strengthen coverage and press releases by weaving in credible third-party validation, citing industry benchmarks, referencing academic studies.
00:31:07.396 –> 00:31:11.080
It’s kind of like we’re going back to university and citing our sources.
00:31:11.820 –> 00:31:15.942
linking to independent reports, but we should also be doing the same on client websites.
00:31:16.342 –> 00:31:18.481
They publish a stat or make a bold claim.
00:31:18.823 –> 00:31:22.722
You should ideally link out to the original source because this transparency builds trust.
00:31:22.762 –> 00:31:26.003
And then both people and AI are picking up on these signals, right?
00:31:26.624 –> 00:31:30.944
So takeaway here, simple again, you’re already doing a lot of things to build authority.
00:31:31.825 –> 00:31:39.467
The opportunity is just making sure that those signals aren’t just in the press, but also baked into the client’s own website where AI can actually read them and reuse them.
00:31:40.765 –> 00:31:47.567
Now, one of the questions I’m asked all the time by our PR agency partners is how to format a press release to enhance AI visibility.
00:31:47.728 –> 00:31:55.490
And in addition to the other tactics that we’ve already covered and we’re going to continue covering today, there are three more that are especially important for this.
00:31:56.211 –> 00:32:04.214
And we’ve attached a great resource to this slide from Notified that gives a great template for you to use to make your press releases more visible to AI.
00:32:05.015 –> 00:32:10.880
First, front load your key details because AI systems grab the most important facts from the very start.
00:32:10.920 –> 00:32:17.785
So if the who, what, where, when, how, why are buried in that fourth paragraph, there’s a good chance that they’re going to be missed.
00:32:18.105 –> 00:32:21.887
Put all the essentials right at the top so that journalists and AI have the context.
00:32:21.969 –> 00:32:25.090
And, you know, the good news is this is a standard practice.
00:32:25.191 –> 00:32:28.133
I don’t think anyone is writing press releases in another way.
00:32:28.212 –> 00:32:30.875
But so nothing’s changed too much so far here.
00:32:31.736 –> 00:33:00.440
second uh using conversational phrasing so press releases they can’t just tell anymore they also need to answer and AI thrives on natural question driven content so instead of writing in statement form like we typically do with press releases build in the questions that people will actually ask in an faq format so you know when is it launching what problem does it solve and then answer those directly in the copy And then third, optimize multimedia with context.
00:33:00.519 –> 00:33:08.067
So a press release often comes with images, sometimes videos and infographics, but AI can actually only reuse them if it’s properly labeled.
00:33:08.146 –> 00:33:17.634
So adding things like alt text, captions, and clear file names that describe exactly what’s happening will give the model even more entry points to understanding the story.
00:33:19.997 –> 00:33:31.362
Now, the hard truth about AI search is this, and we’ve kind of alluded to this, but if the AI robots can’t reach your content in that grounding phase, you may as well not exist or have control over your narrative.
00:33:32.001 –> 00:33:36.023
It doesn’t matter how good the story or the content is or how much coverage you’ve secured.
00:33:36.423 –> 00:33:42.006
If the technical foundations aren’t in place on your own content, AI tools won’t see it and they won’t use it.
00:33:42.086 –> 00:33:44.946
So for PR agencies, this is often the hidden gap.
00:33:45.267 –> 00:33:50.068
You can land the headline in the perfect outlet or create impactful, thoughtful thought leadership pieces.
00:33:50.108 –> 00:33:57.290
But if it’s locked behind a paywall, if it’s published as a PDF or if it’s buried on a slow, clunky site, AI will just skip it.
00:33:57.330 –> 00:34:06.194
And so that’s why pairing storytelling with technical SEO is really important here because the story gets you noticed, but the technical setup ensures that AI can actually pick it up.
00:34:06.234 –> 00:34:12.318
So If AI systems can’t reach your content quickly and clearly, they won’t use it in their answer.
00:34:12.418 –> 00:34:16.340
And that’s why the basics of site accessibility matter more than ever.
00:34:16.659 –> 00:34:24.204
So the following tactics, they might sound pretty technical without any developer background, but don’t worry, I’m going to try and make them as digestible as possible.
00:34:25.224 –> 00:34:29.266
So the first thing is we want to ensure that our client sites have crawlable sitemaps.
00:34:29.706 –> 00:34:33.548
So sitemap is essentially a file hosted at the root domain of your website.
00:34:33.588 –> 00:34:37.451
So, you know, website.com slash you’d have sitemap.xml.
00:34:37.490 –> 00:34:44.815
And it’s basically a roadmap of all of the important pages on a website to help the machines find your fresh content quickly.
00:34:45.255 –> 00:34:52.260
So if your client regularly publishes releases or thought leadership, a sitemap ensures that those updates are visible and not left behind.
00:34:53.141 –> 00:35:10.800
we also want to ensure that the website has a clean site architecture so this means keeping your navigation and your url structures very logical and simple journalists users and AI crawlers should be able to find your newsroom your about page and your product pages in just a few clicks.
00:35:11.239 –> 00:35:13.601
And this also extends to how pages are labeled.
00:35:14.061 –> 00:35:24.827
So using clear URL slugs, making sure every page has a descriptive meta title and meta description, Josh is going to talk about this a little later as well, giving you some more definitive advice here.
00:35:24.887 –> 00:35:34.376
But And then another thing is your site speed is also critically important here because slow pages just straight up get ignored because AI will move on to faster alternatives.
00:35:34.416 –> 00:35:40.384
So there’s things you can do in the backend to help speed up your site, such as optimizing images, caching.
00:35:40.664 –> 00:35:42.106
Sometimes it’s a server problem.
00:35:43.487 –> 00:36:09.503
and then you know you can actually take a you can do this right now you can take a client’s website of yours and you can put it into a tool called page speed insights it’s literally pagespeedinsights.com i’m pretty sure um and then see how they’re scoring and identify some of the assets on their website that’s causing the website to be slow so both on mobile and on desktop And we touched on how AI is mapping relationships between entities and topics and that you need to cover a good topical depth on your site.
00:36:10.204 –> 00:36:13.865
And I also mentioned that you want to link those spoke content pieces to the hub pieces.
00:36:14.226 –> 00:36:17.027
And this is a process called internal linking.
00:36:17.047 –> 00:36:20.230
So it shows AI the relationship between those key pages.
00:36:20.610 –> 00:36:31.936
And so when you link your fintech digital payments page to a related page like fraud prevention or cross-border transactions, you’re teaching the AI that they belong to the same story under the same brand related to this main topic.
00:36:31.956 –> 00:36:36.440
All right, Josh, talk about visibility blockers.
00:36:37.240 –> 00:36:38.362
Yeah, there are a few.
00:36:38.681 –> 00:36:39.543
It’s quite interesting.
00:36:39.663 –> 00:36:42.364
So the first thing that I really want to talk about is JavaScript.
00:36:42.606 –> 00:36:46.789
JavaScript is like the gotcha for so many big websites right now.
00:36:47.269 –> 00:36:58.398
So even to this day, across Claude, across ChatGPT, across Gemini, across Perplexity, none of the answer engines that I just mentioned can see and render JavaScript.
00:36:58.958 –> 00:37:08.144
So if you’ve got a beautiful page with great content, but it’s rendered client side, it’s rendered through JavaScript, the answer engine sees a blank white page.
00:37:08.445 –> 00:37:09.286
It doesn’t see anything.
00:37:09.726 –> 00:37:11.927
It basically isn’t going to be able to get that context.
00:37:12.347 –> 00:37:17.293
So making sure that our content is server-side rendered, what that basically means is making sure it’s in the HTML.
00:37:17.733 –> 00:37:25.702
Making sure the content is in the HTML and it’s already there for the Answer Engine to see is the first big step towards getting our content pulled out.
00:37:26.063 –> 00:37:28.485
The second thing is PDF only publishing and reporting.
00:37:29.246 –> 00:37:31.907
So answer engines don’t typically download files.
00:37:32.007 –> 00:37:43.891
So if you’re going onto a web page and you have this incredible research report or this incredible white paper, but it’s in a PDF format and it has to be downloaded or you have to put your contact information in, that’s not going to be digestible.
00:37:44.231 –> 00:37:49.054
And even the PDF URL itself, really perplexity is the best at this, but it’s still kind of spotty.
00:37:49.534 –> 00:37:56.755
It’s not always something that’s going to be easily indexed and easily understood where and how it should be contextualized based on the way that the PDF is going to be read in.
00:37:56.775 –> 00:37:58.416
So PDFs are pretty dicey right now.
00:37:58.936 –> 00:37:59.697
Hard paywalls.
00:37:59.797 –> 00:38:07.302
Obviously, there’s a little bit of litigation going on there with OpenAI, Perplexity, New York Times, all those different companies are trying to figure out what paywalls mean in the first place.
00:38:07.643 –> 00:38:11.626
But basically, here to say that paywalls are imperfect.
00:38:11.706 –> 00:38:15.449
They’re not a good idea if you want your content to get picked up and understood through.
00:38:15.869 –> 00:38:21.134
Basically, there is a debate that answer engines can see through paywalls, but we don’t want to rely on that for visibility.
00:38:21.373 –> 00:38:26.518
We want to keep our ungated content that answer engines are supposed to see ungated, and our gated content should be gated.
00:38:26.858 –> 00:38:30.302
So right now it’s a soft gate, but paywalls are kind of a gate.
00:38:30.561 –> 00:38:33.525
And the biggest one of them all I think right now is robots.txt.
00:38:34.161 –> 00:38:39.402
Robots.txt is like the Great Wall of China for these answer engines.
00:38:39.443 –> 00:38:43.684
If you’re blocking the answer engines in robots.txt, they’re not going to be able to get in.
00:38:43.704 –> 00:38:50.505
And I’ve worked with companies where they’ve bought Profound, they’re ready to get started, and we go and we look at the data and they’re getting zero percent citations.
00:38:50.525 –> 00:38:55.568
Well, we ask the web dev team, we look at the robots.txt, oh, it looks like they’re literally blocking OpenAI.
00:38:55.927 –> 00:38:57.048
It’s a really easy fix.
00:38:57.088 –> 00:39:01.389
Hey, guys, we really want to have OpenAI cite our pages and look at our pages.
00:39:02.110 –> 00:39:05.731
So going in there, checking your robots.txt, making sure it’s not blocked, really important.
00:39:05.751 –> 00:39:14.795
And there’s even now CDN providers like Cloudflare who are actually in your own Cloudflare instance, you might see Cloudflare blocks bots sometimes by standards.
00:39:14.815 –> 00:39:16.215
So this is something to go in and check.
00:39:16.534 –> 00:39:18.775
Make sure you’re not doing that as well if you want those citations.
00:39:19.909 –> 00:39:20.891
Very valuable advice.
00:39:22.492 –> 00:39:26.476
And if you want AI to understand your content, you’re going to need to label it clearly.
00:39:26.516 –> 00:39:29.858
And that’s exactly what structured data or schema markup does.
00:39:29.998 –> 00:39:36.804
Essentially, a little piece of code that you put on the back end of the website on your content that tells the AI what the content is.
00:39:36.885 –> 00:39:38.445
So here’s the interesting bit.
00:39:39.246 –> 00:39:43.527
Even Google has come out and said that structured data plays a role in AI overviews.
00:39:43.909 –> 00:39:51.911
Their own guidance highlights that markup helps their systems interpret content, and they specifically advise that structured data should match what’s visible on the page.
00:39:51.931 –> 00:40:01.054
So the takeaway is, if Google is mentioning this in the context of AI overviews, you can bet that they’re at least looking at structured data as part of the signals for visibility.
00:40:01.134 –> 00:40:05.255
So let’s look at some of the most practical schema examples for this in PR.
00:40:06.016 –> 00:40:17.501
And by far the most important one that you guys can start implementing is news article schema, because when you mark up a press release or news coverage with this schema, you’re telling AI systems exactly what the coverage is.
00:40:17.800 –> 00:40:22.963
So whether that’s an announcement, you’re telling them who wrote it, when it was published and then who it’s about.
00:40:23.063 –> 00:40:28.005
So that makes it far more likely to surface when people ask questions about your client or the news.
00:40:29.106 –> 00:40:37.934
And then how to schema is really great for those step by step explainers and guides, you know, like how to integrate our technology with QuickBooks or something like that.
00:40:39.148 –> 00:40:45.271
Uh, FAQ schema helps up mark, uh, those, uh, common questions and FAQ content that we talked about earlier.
00:40:45.731 –> 00:40:49.753
And then video schema, make sure that branded videos, they’re not just sitting on your website.
00:40:49.974 –> 00:41:01.659
They can also show up in AI generated answers because we know that, uh, some of the LLMs are actually multimodal, meaning that they, they can answer, um, you know, con, uh, responses with video content they’re learning from.
00:41:02.112 –> 00:41:03.693
different types of content as well.
00:41:04.693 –> 00:41:16.438
So if you want to check if your clients have schema markup on their site or on their service pages or product pages or whatever, you go to schema.org and then you can see examples of all of the different kinds of schema markups that you can put on your content.
00:41:16.458 –> 00:41:17.637
There’s literally hundreds of them.
00:41:18.358 –> 00:41:21.398
And then you can even check or validate on this.
00:41:21.438 –> 00:41:25.800
Put your client’s website in and then see if they’ve got any types of these schema markups implemented.
00:41:27.543 –> 00:41:27.742
Okay.
00:41:28.003 –> 00:41:42.976
We started this session with this graph and I just want to bring it back to end on the full picture because by now you’ve seen how the different tactics break down, you know, the content side where PR teams already have quite a bit of strength and then the technical side where that capability gap typically is.
00:41:43.416 –> 00:41:49.862
And you can see that actually many of these technical items are done pretty quickly within the first two months of the campaign, which is great.
00:41:50.182 –> 00:41:56.429
Sometimes it’s just a small change like that robots.txt that we talked about that will really start to skyrocket your visibility.
00:41:56.489 –> 00:42:08.199
So if PR teams can bridge this gap, whether that’s by building this knowledge internally or collaborating with a technical partner like Profound or SUSO, you don’t just protect your relevance in the AI era, you’re leading it, right?
00:42:09.268 –> 00:42:18.097
So Josh is going to take us through some very specific trends from Profound’s data that he’s been seeing and talked about, I think, recently at Brighton SEO and all that.
00:42:18.197 –> 00:42:19.398
It’s been a pleasure there.
00:42:19.780 –> 00:42:26.806
Yeah, so we’re looking right now, this data set right now is showing us that listicles and comparative content are dominating in AI citations.
00:42:26.827 –> 00:42:31.550
This is a sample of over two point six billion with a B citations.
00:42:31.992 –> 00:42:35.775
So we took our entire citation catalog and we annotated it.
00:42:35.876 –> 00:42:44.525
We looked at every single citation and we figured out that comparative listicle content right now is the single most common cited type of content.
00:42:44.865 –> 00:42:45.967
And it makes sense why.
00:42:46.563 –> 00:43:00.413
When answer engines are going out to the open internet, they’re, again, relying on that structured information, direct comparisons, cohesive, factual, and authoritative takes on the top products, brands, or services in any given industry.
00:43:00.873 –> 00:43:14.427
And the formatting and style of comparative listicle content tends to synergize most easily and most effectively with the kind of content that answer engines are capable of bringing back to the user and the kind of questions that users are asking.
00:43:14.786 –> 00:43:20.393
So that’s a twenty five point three seven percent distribution towards comparative content, a twelve percent distribution towards blog.
00:43:20.673 –> 00:43:23.757
So that’s a two X bend right now towards that piece of content.
00:43:23.916 –> 00:43:26.039
I do think this is going to change in the long run.
00:43:26.559 –> 00:44:03.804
but i think right now because most people aren’t thinking about releasing content in this easily digestible aeo format there is this natural bias towards comparative content yeah the second piece right now this one’s a real this was really interesting for us to find semantic urls get eleven point four percent more citations on average so what that basically means is semantic url that basically means that urls with natural language queries like if i’m looking if i if i’m writing a page about the best crms a semantic url for a best crms page would be the dash best dash twenty twenty twenty 2025 dash CRMs dash compared.
00:44:04.344 –> 00:44:05.445
That’s a semantic URL.
00:44:05.505 –> 00:44:13.230
They are an eleven point four percent more citations than a normal URL like eight three GF two slash CRMs slash list.
00:44:14.090 –> 00:44:18.974
Reason for that being that answer engines are going to their list of results.
00:44:19.434 –> 00:44:22.597
We know that SERP API, for instance, is what ChatGPT is using.
00:44:23.036 –> 00:44:24.597
SERP API sends a list of results.
00:44:24.978 –> 00:44:32.483
ChatGPT looks through the list and is like, which of these pieces of content is most likely to give me the answer to what the best CRM is?
00:44:32.844 –> 00:44:34.784
What piece of content is going to tell me that information?
00:44:34.844 –> 00:44:37.065
And it sees the best CRM is 2025 compared.
00:44:37.085 –> 00:44:37.887
It’s like, that’s pretty good.
00:44:38.327 –> 00:44:39.387
That gives me some confidence.
00:44:39.427 –> 00:44:42.650
And then it sees eight GF two one oh four CRM list.
00:44:42.690 –> 00:44:43.510
It’s like, I don’t know about that.
00:44:44.210 –> 00:45:17.967
eleven point four percent more citations there very interesting and that’s just a free eleven point four percent citation game for anyone who wants to implement yeah social media and ugc are also very interesting i used to say that social media and ugc were like basically bar none, they’re everywhere but microsoft copilot had to go and stop indexing Reddit so i had to change it to situationally so seventeen point four percent of all chat gpt citations are for ugc so we’re talking about Reddit we’re talking about linkedin quora yelp TripAdvisor, Instagram is even in there a little bit.
00:45:18.128 –> 00:45:19.809
Perplexity is fifteen point eight percent.
00:45:20.168 –> 00:45:21.610
Google AI overviews is twelve percent.
00:45:21.929 –> 00:45:24.672
Microsoft Copilot doesn’t cite Reddit, doesn’t cite YouTube.
00:45:24.931 –> 00:45:26.233
So it’s at four point six percent.
00:45:26.353 –> 00:45:27.213
Very interesting there.
00:45:27.552 –> 00:45:31.514
But the thing to take away here is that this is no longer just an SEO thing.
00:45:31.534 –> 00:45:39.561
This is no longer, you know, one channel, you know, on website URLs, talking about some of the same units that we’re playing with.
00:45:39.960 –> 00:45:41.202
This is expanded in scope.
00:45:41.282 –> 00:45:51.829
Search teams and just marketing teams in general have to start talking to each other because social is a new landscape in product and brand discoverability, especially with these answer engines.
00:45:51.849 –> 00:45:53.929
So it’s very interesting to see how prevalent this is.
00:45:55.434 –> 00:45:55.795
Awesome.
00:45:55.815 –> 00:45:57.637
Thanks so much for all of your insights, Josh.
00:45:58.416 –> 00:46:05.322
I do want to just take a couple of minutes just before Q&A to share a great way to dip your toes into AI search and GEO without any investment.
00:46:05.362 –> 00:46:12.726
And that’s with our partner club, which is a completely free resource that gives agencies and consultants access to a few main things.
00:46:12.766 –> 00:46:17.451
So we give free private workshops to upskill our partners on SEO and AI search.
00:46:17.490 –> 00:46:19.311
So it’s very similar to today’s webinar.
00:46:19.672 –> 00:46:23.195
So if you like the content, you want to spread that love with the rest of your agency.
00:46:23.894 –> 00:46:51.210
um yeah you join the partner club and then you just coordinate a workshop with me and i can give that free training to your wider team we also have a twenty four seven help desk that connects you directly with our seo specialist so you can get answers to your questions about AI search and seo within twenty four hours without any investment and then We also do free SEO and AI visibility reports for our partners, white labeled in their branding, and they really give a good snapshot into how their clients are showing up on traditional and AI search engines.
00:46:51.489 –> 00:46:52.471
They’re not automated.
00:46:52.490 –> 00:47:14.739
We have actual people on our team that put about two hours of analysis on the website and they look at competitors and they identify things on the actual website that might be holding their performance back, whether that’s like technical things like structured data or the lack of on page, whether that’s the way the content is structured or even if, you know, maybe technical and on site issues, off site issues we mentioned.
00:47:14.778 –> 00:47:15.019
Right.
00:47:16.199 –> 00:47:19.362
So there is you can see some pretty cool logos here.
00:47:19.402 –> 00:47:22.925
These are just some of the over fifteen hundred agencies that are part of the partner club.
00:47:23.505 –> 00:47:25.766
They use these free resources to get upskilled.
00:47:26.367 –> 00:47:31.771
We’ve worked with some agencies on actual paid SEO projects as well as their back end technical delivery partner.
00:47:31.791 –> 00:47:34.552
So very easy way for you to fill that technical gap.
00:47:35.072 –> 00:47:37.114
Just sign up on the website using that link there.
00:47:37.173 –> 00:47:50.920
But we’ll also have a QR code at the end of the presentation for this as well Uh, and just the teaser for our final session in this series, which is going to tackle the big question that always comes after strategy, which is how do you measure success in AI search?
00:47:51.121 –> 00:47:55.105
Um, you know, AI search is becoming a core part of how people discover brands.
00:47:55.144 –> 00:48:00.688
So rankings alone, aren’t really enough and we need new ways to track and report on visibility.
00:48:00.750 –> 00:48:09.217
So in the next webinar, I’ll be joined by Chris Andrew from Scrunch AI, where we’ll dig into exactly that, what kinds of KPIs we should be looking at in this new era.
00:48:09.916 –> 00:48:18.103
how to actually monitor brand presence and generative answers, and even how to track traffic and conversions on your site from AI using Google Analytics and more.
00:48:18.164 –> 00:48:19.565
It’s actually a lot easier than you think.
00:48:20.286 –> 00:48:23.367
So join us on Thursday, October thirtieth for AI visibility.
00:48:23.387 –> 00:48:25.951
You can measure AI search reporting for marketers.
00:48:26.271 –> 00:48:29.914
The QR code to sign up for this will also be at the end of the presentation as well.
00:48:30.922 –> 00:48:35.427
So we’ll head into some Q&A and I’ll put all those QR codes on the screen.
00:48:35.766 –> 00:48:37.608
You can check out Profound there.
00:48:37.648 –> 00:48:41.532
You can join our partner club and you can also sign up for our webinar there.
00:48:41.932 –> 00:48:43.635
So I see one question so far.
00:48:43.655 –> 00:48:47.579
I’ve got my colleague Deep is managing it in a separate tab for me.
00:48:47.998 –> 00:48:54.806
But Nick Herf has asked, do hard paywalls also include gated owned content on website for the purpose of driving leads?
00:48:55.085 –> 00:48:57.407
How can you drive leads through content without gating it?
00:48:57.568 –> 00:48:58.309
Great question.
00:48:59.510 –> 00:49:05.833
So, yes, they’re still going to treat them the same way as they would with like gated press, right?
00:49:06.155 –> 00:49:10.257
So what you can do is you can create dedicated landing pages for that gated content.
00:49:10.838 –> 00:49:17.063
And you want to almost spoil a little bit of that content in like a key takeaway section on that landing page.
00:49:17.163 –> 00:49:19.844
Unfortunately, I know that you want to protect that.
00:49:20.746 –> 00:49:22.867
But that is unfortunately like the workaround.
00:49:23.268 –> 00:49:24.688
Josh, do you have anything to add there?
00:49:26.467 –> 00:49:27.628
I think that’s an excellent answer.
00:49:27.849 –> 00:49:39.777
I think the only thing that people are doing that’s kind of crazy right now, there’s a crazy strategy where people are basically serving Answer Engine user agents different versions of the page than your Googlebot or other agents.
00:49:40.137 –> 00:49:42.960
That’s called cloaking, and it’s kind of a gray area in SEO.
00:49:43.380 –> 00:49:48.682
But it’s an interesting test because people will see this and they’ll say, I really want to have this piece of content.
00:49:48.744 –> 00:49:58.027
If ChatGPT was to go to this landing page, Maybe I want a markdown version of the page to load or some sort of scrapped down version of the page that has a little more context into what’s going on on the page.
00:49:59.547 –> 00:50:00.306
It’s worth a test.
00:50:00.387 –> 00:50:02.327
It’s really interesting, the results so far.
00:50:03.768 –> 00:50:07.009
Not too earth shattering, which is good, I think, honestly.
00:50:07.048 –> 00:50:09.070
It means that we don’t have to change our websites too much.
00:50:10.559 –> 00:50:15.543
The interesting thing is that the answer engine on average was able to pull just a little bit more content through in the test that we were doing.
00:50:16.083 –> 00:50:24.588
I don’t think it’s gonna change the world, but if you’re absolutely dead set on getting a certain white paper or something like that digested by an answer engine, there are ways that you can do it.
00:50:24.668 –> 00:50:31.893
And Josh, like what’s your opinion on, cause it does very much feel like we’re in sort of that like black hat era of AEO.
00:50:32.472 –> 00:50:38.476
Do you think that like things like that, like cloaking are going to eventually sort of be deprioritized?
00:50:39.197 –> 00:50:41.559
I think cloaking is going to get legitimized, if anything.
00:50:42.221 –> 00:50:50.327
I think there’s going to be real frameworks for actually giving answer engines a slightly, I guess, wordier or contextual or uglier.
00:50:50.666 –> 00:50:51.708
You can define whatever it is.
00:50:51.748 –> 00:51:01.436
But answer engines, it should be a two-tiered system, I think, where on the bottom system, answer engines get like this text version of every web page, and they can just go through and get them MVP.
00:51:01.456 –> 00:51:03.438
They get all the details, all the text that you need.
00:51:03.797 –> 00:51:06.297
And the human version of the web page should be beautiful, gorgeous.
00:51:06.338 –> 00:51:08.679
It does not need to be optimized for machine readability.
00:51:09.059 –> 00:51:19.221
And right now, we’re in this situation where basically each of those different answer engines and each of the search engines are using the same system and the same kind of units to browse and understand what’s going on.
00:51:19.740 –> 00:51:24.461
Where we really want is the ability and the freedom to kind of unshackle both of those systems.
00:51:24.802 –> 00:51:26.461
So for now, we’re in this awkward state.
00:51:26.661 –> 00:51:32.862
I think we’re actually going to get to a more peaceful, harmonious state where the answer engines have the ability to access content that’s made just for them.
00:51:33.826 –> 00:51:35.748
Especially as commerce starts to get important.
00:51:35.769 –> 00:51:37.012
Especially with commerce.
00:51:38.311 –> 00:51:40.411
That more agent to agent experience.
00:51:40.552 –> 00:51:41.833
Yeah.
00:51:41.853 –> 00:51:42.054
Okay.
00:51:42.114 –> 00:51:45.056
Lorraine says, can you explain briefly what schema is?
00:51:45.155 –> 00:51:45.376
Yeah.
00:51:45.436 –> 00:51:47.817
So it’s essentially a piece of code.
00:51:47.856 –> 00:51:51.920
It’s a developer looking piece of code that you put on the back end of your website or content.
00:51:52.621 –> 00:51:54.561
I can show you an example of what it looks like.
00:51:56.003 –> 00:51:57.164
This is what it looks like.
00:51:57.224 –> 00:52:00.846
Essentially, it’s not, you know, seen to the naked eye.
00:52:00.865 –> 00:52:03.807
You’d have to actually go inspect the code and you’d be able to see it.
00:52:04.427 –> 00:52:05.849
But like I said, you can go to schema.org.
00:52:05.989 –> 00:52:08.010
I can pull it up right now quickly.
00:52:09.092 –> 00:52:12.054
And then you can see all of the different kinds of schemas here.
00:52:12.074 –> 00:52:12.974
You can even search.
00:52:13.195 –> 00:52:20.260
So for example, I talked about news, there’s news article, there’s news media organization, newspaper, all of these different schemas.
00:52:20.719 –> 00:52:24.503
And they’ll tell you essentially how are you gonna code that into it?
00:52:24.543 –> 00:52:27.344
And then you’d put that into the backend of the website.
00:52:27.706 –> 00:52:29.306
I’m not a developer, I’ve never done it.
00:52:29.467 –> 00:52:35.672
So ask your local web developer, or like I said, consult with some technical experts like our team or Josh’s team as well.
00:52:37.797 –> 00:52:42.322
Kimberly asks, how can you build out a media list that is ranked for AI search visibility?
00:52:42.481 –> 00:52:45.405
The great question. So I have some insight onto this.
00:52:45.485 –> 00:52:55.576
I don’t know if Josh also has some additional insight on this, but there are specific publications that have tied deals and made partnerships with some specific like tools.
00:52:56.396 –> 00:52:59.077
I’m happy to share that in a follow up email as well.
00:52:59.197 –> 00:53:09.500
Lily Ray actually put together a really comprehensive one that says exactly what perplexities deal with this publication is and how they’re providing information for their search answers.
00:53:09.840 –> 00:53:17.103
Actually, let me pull it up now and I’ll just I’ll drop it in the chat. So we all have it because I do think it’s a really valuable partner.
00:53:19.829 –> 00:53:22.070
closed my Slack here, I gotta pull it up.
00:53:22.210 –> 00:53:25.833
But in the meantime, Josh, do you wanna add some color while I look for that?
00:53:25.853 –> 00:53:33.838
Yeah, I think that one of the most interesting things is that you have to see what’s being cited to a certain extent to understand where the arbitrage opportunities are with some of these citations.
00:53:33.878 –> 00:53:39.181
Like what we’ve seen right now because utility-first content is such a mandate.
00:53:39.240 –> 00:54:19.244
When I say utility-first content, I literally mean like content that gives answer engines the direct answer um from any source because some of these traditional seo strategies matter a little less in AEO, what we’re seeing is that those sources are very vertical dependent i would talk to people in the pr in the pr industry uh in the ad industry who i show them here’s the you know for your client you know let’s just i’m gonna throw a company i was gonna say for your client burger king Um, these are the top cited sources in fast food and they’ll be like, okay, what’s weird is that, you know, fortune business insider Forbes, you know, all of the, you know, Bon Appetit, none of those are at the top list.
00:54:19.664 –> 00:54:26.925
You know, the ones at the top list are like eatthis.com and like fastfoodeats.com and like davesburgerblog.com.
00:54:27.266 –> 00:54:56.043
So it’s really interesting to see as you start to dive into hyper verticalization, the actual media list gives you innate opportunities to arbitrage and hit sites that have much less traditional visibility, but because their content is so specifically geared towards the industry and the answer engines don’t really read into their authority that much, or they do, but they don’t, not as much as traditional SEO, you have an opportunity to start playing in those specific fields and in specific places.
00:54:56.083 –> 00:54:56.905
It’s really interesting.
00:54:58.233 –> 00:54:58.353
Cool.
00:54:58.373 –> 00:55:03.916
Yeah, I just shared the link to the document that was put together by Lily Ray, who’s like an esteemed SEO voice.
00:55:05.117 –> 00:55:14.422
And yeah, so there’s a whole list that you can see here, you know, link to the publisher, who the AI company that they have that partnership with, and then key details of that partnership.
00:55:14.483 –> 00:55:23.188
So definitely as like a little strategy would probably recommend prioritizing some of these publications if you’re doing some pitches and things like that.
00:55:24.480 –> 00:55:26.702
I’ll go back to my presentation.
00:55:27.882 –> 00:55:38.331
And then Tom says, does paid for content like advertisements or advertorials drive citations as well as editorial if it’s on a trusted source?
00:55:38.471 –> 00:55:42.253
Or can LLMs see that it’s paid, so therefore lacks credibility?
00:55:43.135 –> 00:55:51.380
I can speak to a specific stat done by Muck Rack that pretty much showed that paid sources aren’t contributing to LLM visibility at all.
00:55:52.722 –> 00:55:54.862
Josh, what are you seeing on your end?
00:55:55.623 –> 00:55:57.324
Yeah, it’s the same thing.
00:55:57.364 –> 00:55:57.684
Yeah.
00:55:57.724 –> 00:56:00.445
So yeah, easy answer.
00:56:00.766 –> 00:56:04.067
Paid for content is not going to get you as much AI visibility.
00:56:04.527 –> 00:56:06.949
That question will come into, it might change, right?
00:56:06.969 –> 00:56:13.672
Because ChatGPT is inevitably going to start including ads in their thing. So stay tuned on advancements there.
00:56:15.112 –> 00:56:19.074
I believe that’s all the questions. We’ve got about three minutes left.
00:56:19.094 –> 00:56:25.416
So if there’s any final questions, drop them in the chat. And then otherwise, you know, our Helpdesk is open.
00:56:26.416 –> 00:56:29.757
Josh’s email is in the presentation. So we’ll share that over.
00:56:29.838 –> 00:56:36.780
Definitely follow Josh on LinkedIn. He’s always sharing like the latest news and research from Profound.
00:56:37.920 –> 00:56:39.702
And yeah, and join our Partner Club because it’s free.
00:56:39.742 –> 00:56:40.141
Why not?
00:56:41.192 –> 00:56:46.898
I’ll give everyone just a couple more minutes and then I’ll give everyone their day back.
00:56:46.918 –> 00:56:51.568
Last questions? Nope. All right.
00:56:51.608 –> 00:56:54.672
Well, thanks so much for joining once again, Josh.
00:56:54.831 –> 00:56:55.893
It was great to see you again.
00:56:55.932 –> 00:57:01.536
Josh and I did a talk in New York at Tech Week together. It’s always good riffing off each other.
00:57:02.717 –> 00:57:07.661
But yeah, and stay tuned for our next webinar. Definitely sign up for that and hope to see you all there.
00:57:08.021 –> 00:57:09.362
Enjoy the rest of your week.
00:57:09.503 –> 00:57:12.184
And yes, you’ll get the slides. We’ll send that over in an automated email.
00:57:12.465 –> 00:57:14.606
Thanks, everyone. Thank you all. Take care.
