Most enterprise sites are not ready to be found by AI search, and the fix starts before you ever migrate. In Episode 46 of Beyond the CMS, Chris Bryce talks with Joel Goodman, VP of Growth Strategy at Squiz, about content intelligence, query fanout, and what makes a composable DXP AI-ready.
Joel spent 13 years running an agency in higher education before joining Squiz, a 26-year-old Digital Experience Platform built in Australia with a growing North American presence. They get into what makes Squiz different from a traditional headless CMS, why accessibility is now the same problem as AI discoverability, and how Squiz Content Intelligence audits your content the way an AI search engine reads it.
The sharpest idea in this episode is query fanout. Ask one question in ChatGPT and it breaks into 40, 50, sometimes 60 sub-queries your content has to answer, queries you never see. Joel explains how Squiz generates thousands of questions per topic to find the gaps before AI search does.
Key topics
- Why AI search reads your content differently than Google ever did
- What a composable DXP adds on top of a traditional CMS
- Why accessibility and AI discoverability are the same problem
- How Squiz Content Intelligence scores content across five metrics
- Query fanout, and why one question becomes sixty
- Edge rendering on Cloudflare, personalization, and the built-in CDP
- Getting content AI-ready before a migration, not after
Chapters
- 0:00 Welcome to Episode 46
- 1:43 What is Squiz?
- 3:21 CMS Plus Plus: the composable DXP
- 6:01 The white space between monolith and headless
- 7:20 Why migrations are heavy, and how AI eases them
- 10:23 Using AI to speed up a migration safely
- 13:20 Iceberg projects: AEO, GEO, discoverability
- 14:16 Squiz Content Intelligence
- 14:40 Accessibility and AI discoverability are the same problem
- 15:15 How LLMs read your content
- 17:31 Prioritized fixes
- 19:10 Rebuilding the platform to be AI-first
- 20:26 SEO, schema.org, structured content
- 22:05 Edge rendering, AWS, Cloudflare
- 23:04 Personalization, AB testing, CDP
- 27:02 The 3,400 questions test
- 28:17 Query fanout explained
- 29:17 Free AI visibility report from Squiz
- 30:21 What's next on the roadmap
- 31:12 Wrap-up
Frequently asked questions
What is query fanout in AI search? When someone asks an AI search engine one question, it breaks that into 40 to 60 sub-queries your content has to answer, none of which you see. Your content has to answer all of them to show up in the response.
Is Squiz a CMS or a DXP? Squiz is a Digital Experience Platform, a CMS plus plus. It pairs content management with site search, conversational search, personalization, AB testing, a customer data platform, and asset management in one product.
What is a composable DXP? A platform that gives you an integrated experience out of the box while letting you compose in the tools you already use, so you get the breadth of a suite without the heavy integration lift of stitching point solutions together.
How is Squiz different from a headless CMS? A pure headless CMS hands you content APIs and leaves search, personalization, and delivery to you. Squiz brings those together, and renders at the edge on Cloudflare, so you get headless speed with built-in experience tooling.
What is Squiz Content Intelligence? A platform-agnostic tool that audits the content across your site, scores it across five metrics for how clearly large language models read it, finds conflicting or duplicate answers, and returns prioritized fixes.
Why does accessibility matter for AI search? AI search crawlers run on the same technology as screen readers. If your site is not accessible, you block those crawlers from understanding your content, so accessibility now drives discoverability.
What is answer engine optimization (AEO)? Writing and structuring content so AI answer engines can understand it, trust it, and cite it. It overlaps with good SEO and accessibility rather than replacing them.
Why do AI search engines skip some websites? If a model cannot understand your content, or it finds conflicting answers across your pages and cannot tell which to trust, it moves on to a competitor or a source like Reddit that feels more trustworthy.
How do you get content AI-ready before a migration? Do the content strategy and structural work up front. Audit how language models read your pages, consolidate conflicting answers, and shape content to the structures of the platform you are moving to.
What is a CDP, and why does Squiz build one in? A customer data platform builds a profile for each visitor, known or anonymous, so you can personalize and run AB tests by segment. Squiz includes one and pairs it with edge rendering so personalization does not cost you speed.
Related from Dotfusion
If you are weighing a move off a monolith, see our headless CMS migration services. A dedicated Squiz implementation page is on the way.
Links: Squiz https://www.squiz.net | Dotfusion https://www.dotfusion.com
Full transcript
Edited lightly for readability.
Welcome
We're live, and it's episode 46 already. It's been three years of Beyond the CMS, and hopefully CMS stays a relevant term for the foreseeable future. Thank you to everyone making time on their Thursday afternoon in the Eastern time zone. I'm coming at you from Toronto, and we've got some of the World Cup games going on in the city, which is exciting. The whole idea of the show is to talk content operations and content management with some of the foremost thought leaders in the space. Today's guest is Joel Goodman from Squiz.
Chris, I had such a great time last time, so I've been looking forward to this. Happy to be back, and it's always good to dig into some nerdy tech again. You're joining us as VP of Growth Strategy for Squiz, so let's start at the top.
What is Squiz?
What is Squiz, and what should everyone know about it? The people who listen, on YouTube and the podcast, are typically VPs of Marketing and CMOs tasked with the big decision of managing digital publishing, so we'd love them to know more about you and Squiz.
Squiz is, as of this recording, about 26 years old. We were founded in Australia and headquartered in Sydney, with a huge presence across Australia and New Zealand and a strong presence in the UK. We're newish to North America. I came on board about nine or ten months ago, after owning my own agency for 13 years, working primarily in higher education. Squiz has a long track record in sectors that are content dense. In higher education you might have hundreds of thousands of pages. Government is a strong area for us, especially across Australia and the UK, and it's starting to pick up in the US. We've had good traction in professional services recently too. The thing that binds most of our customers together is that they rely on the content they produce and put on the web, and that tends to be very dense. So we offer a digital experience platform.
CMS Plus Plus: the composable DXP
Your listeners know what a DXP means, but as we come into some of these sectors, people ask: is that a CMS, is it different from a CMS, how does it play into the content operations we have to do? The easiest way to talk about it is a CMS plus plus. We have content management, the table-stakes page builder and WYSIWYG experience, and composable content on a page. On top of that we integrate site search. We have a strong site search product called Funnelback, and conversational search on top of it to bring that AI-powered experience you might have with Google AI mode, ChatGPT, or Claude onto your own website, with your own content. We have personalization and AB testing built into the platform, and a lightweight digital asset management tool.
The thing that really sets us apart is composability on the technical side. If you've already sunk time and effort into your own products, say a digital asset manager with thousands of images, files, and videos, you don't want to move those off. We give you the tools to integrate the products you already use into the main content management interface, so you can put together a full end-to-end data flow, from your front end into search, into content management, and into all the peripheral data points.
It's refreshing to hear DXP and composable together, that complete plus-plus flexibility. We're a partner of yours and really believe in the vision you have for the product.
The white space between monolith and headless
We serve customers with a slight marketing bias, helping them make decisions on their publishing or content platform. Squiz has a unique white space: you get a lot in one package. When you look at the monolith players, the Sitecores and the Adobes, they have their place, but it's a heavy integration lift. You need Java engineers, and that's fine for the use case, but imagine getting that full experience functionality in a package that's actually easy to get to market quickly. Right now marketers know they need to make changes to their publishing and content infrastructure, but there's a pause. Squiz is a strong way to get migrated quickly and in good stead, because you can compose your old platform alongside the new one.
Why migrations are heavy, and how AI eases them
Migration projects are always huge. As someone who ran an agency that did big website redesigns and migrations for colleges and universities, I felt the pain on the back end of managing all of it. The unique thing about Squiz's DXP is that over the last several years the company invested in overhauling the architecture so it can be AI controlled, an AI-first platform from a DXP and content management perspective. That opens up opportunities for automating migrations and easing friction points.
Migrations will always be a heavy lift, and easy is relative to your last one. A lot of the thought goes into cutting down the time to move content over, and into the technical side of mapping parts of your current site, your content and visual components, into our component system and design system, in ways that are easy and ideally invisible to the marketers who don't want to worry about it. You want the assurance it will be done to a high standard, with the reality that you still spot check and run a QA process. We have a great professional services team to handle a lot of that, and we're investing in easing those migration paths, because professional services teams don't enjoy the tedious work either. We've had larger clients go live this year where we moved them off other DXPs and CMSs and migrated hundreds of thousands of pages. Every time you do one, you learn something and improve the system.
Using AI to speed up a migration safely
The show is unscripted, but I'll ask on behalf of my clients and prospects: how can they use AI to enable the velocity of their migration into a product like Squiz?
We handle a lot of the heavy lifting. A lot of it comes down to content strategy, content design, and operational strategy within the tech you choose. Do as much prep work as you can ahead of time. Everyone talks about AI search visibility. Asking how ready your content is before you make a move benefits your current marketing during that liminal space as you work toward a migration. And if you're doing it with a new platform in mind, use AI to understand the structures your content will move into and the differences in mapping between your current platform and the new one. That forethought to reshape your content ahead of time makes it easier. Put a good planning structure in place with the delivery team or your agency.
The AI question is difficult, because there's a lot of groundwork to do to get really good results out of the models as they stand now. So it helps to have a trusted partner, Dotfusion for instance, that knows how to get content into the proper formats and structures for a new platform. What marketer has time to ramp up on the new thing Anthropic or OpenAI throws out? We were just talking through a project with about 500 pages to migrate, which is on the smaller side, and we're careful about anticipating the time on the client's resources. They'll ask, can't you just use AI and map it all over? And I'll ask, are you going to hit the release button without reviewing anything? With 500 pages and a new brand exercise, those should be rewritten. There's a lot of thinking upfront.
Iceberg projects: AEO, GEO, discoverability
We call them iceberg projects. Everyone talks about what's on top, but it's the stuff underneath. With AEO, GEO, and discoverability, there are conversations about the overall site architecture that's ideal to support it. That may be part of why there's a pause in migration right now: if someone is doing well getting found in AI search, they don't want to migrate for fear of losing it. Squiz's architecture gives you a variety of options for how you construct your sites. From an infrastructure perspective, our offerings sit in three layers.
Squiz Content Intelligence
We recently launched a product, about a month and a half to two months ago, called Squiz Content Intelligence. It handles the upfront work before someone even gets onto your website. We audit all the content across your site, and it doesn't matter what platform you're on. It has two parts.
Accessibility and AI discoverability are the same problem
First, it looks at accessibility and where you stand, because if your site isn't accessible, you're blocking the crawlers from AI search engines from even understanding your content. The AI search crawler technology is built on the same technology that powers screen readers and assistive applications. So accessibility becomes arguably even more important as a business function than it was before.
How LLMs read your content
On the flip side, if you've got 500 pages, 5,000 pages, anywhere north of about 150 pages, you need to understand how large language models interpret your content and how they lift certain messaging out of it. When we implemented our conversational search tool, we had to understand how the large language models that power it were reading the content, to get good answers to the questions people asked. We had consultants doing that manually for at least a year, pouring in hundreds of hours, to make sure a question on a university's website pulled up the right answer. It turns out that if you write your content and do accessibility on your own site to serve good on-site conversational search answers, it does the exact same thing for AI search out there, whether it's ChatGPT or Google AI mode.
So we look across all the content on your website and break it into topics. We analyze whether you're writing consistently, whether you're consolidating like content on like pages, and whether you're specific about the details you want to show up in search. Across five metrics we identify where you have conflicting information across pages, and which one is right. Because if AI can't figure out which answer to trust, it does one of two things. It hallucinates and hedges, or, more likely with AI search, it has an index of the top five to ten for every question and goes to a competitor's site or an answer on Reddit, because that feels more trustworthy than your own website. Insights from the structural side of content help both your findability in AI search and how conversational search on your own site turns up answers.
Prioritized fixes
Next, we prioritize which fixes will have the biggest impact on showing up in AI search answers. We tell you, here's the problem, this is high priority, and here's a suggested fix. You can take the suggestion, update it in your CMS or DXP, and watch it get better. Down the line, if you're working within the Squiz DXP, we're working toward helping you automatically update that content throughout the platform. That's the future state: using AI to smooth over the technical friction that happens with, here's a recommendation, now go to this page, log in, find that paragraph in the editor. Reducing those steps reduces overhead and gives marketers more time back to do what they're meant to do, which is marketing and telling stories.
Rebuilding the platform to be AI-first
This ties into rebuilding the platform to be in line with how the robots are talking. On the structural side, the industry is changing all the time, and the ways AI search interprets your content keep shifting. Think about traditional SEO, where Google constantly tweaked the algorithm, weighting a meta description over a heading, or FAQs showing up in snippets. That's still changing within AI search. The one constant is that if it can't understand your content, it skips over it and goes somewhere else. So you can start writing and managing your content from the perspective of how clearly large language models will interpret it, and that helps you show up in more answers.
SEO, schema.org, structured content
On the tech side, we have all the tools for good SEO. Accessibility helps SEO. Strong schema.org metadata and structured content blocks help SEO, and all of that in turn helps AI search visibility, because these engines use real search indexes. The structures are about not putting blockers in place on the technical side of the site. If a page is getting built, it should be built in an AI-first way. We've put great investment into making each part of the DXP able to be controlled by AI, including agentic processes that help you research a topic, create a draft, and pull it back in. There's a lot of opportunity to be creative as the abilities of large language models and generative AI come into their own. At the same time, it's about using AI to find the patterns in the content you already have, helping you improve and optimize it so you have a better chance of showing up, and reducing the work on the front end for your users and content authors.
Edge rendering, AWS, Cloudflare
The third part is the investment in infrastructure. Squiz is built with deep partnerships with AWS and Cloudflare, so a lot of the personalization work and component rendering is done at the edge. You get a hybrid benefit you'd typically associate with headless: it develops a static version of your page, pushes it to a CDN, and serves it at the edge. We're doing parts of that within our platform, but you also get dynamic personalization and AB testing while keeping the speed benefits of publishing at the edge, because of our relationship with Cloudflare. The technology to support dynamic personalization when you're pure headless poses a challenge, and that's one of the pluses of the Squiz DXP.
Personalization, AB testing, and the CDP
I wanted to get into personalization more deeply, because we haven't covered it enough on the show. For folks who don't know, one of the unique aspects of Squiz is that you have a CDP, a data platform associated with the content that holds information about your customers.
Our customer data platform integrates with our integration platform as a service, so you can pull in your own CRM data, email marketing, an advertising platform, or whatever else. It links that data into profiles for each user that lands on your site. If you already know who the person is, because you have the data in your CRM or they've logged in or submitted a form, you can get hyper-personalized. You could even personalize down to a greeting by name, but don't do that, it's creepy, unless you're Amazon or they're actually logged in. At the same time, if you don't know who the person is, you can still personalize even if they're anonymous, or dark. You might not have their name and email, but you have a better chance of converting them by inferring their interests from data signals in the CDP.
Our approach to personalization and experimentation in the DXP right now is AB testing. We're not doing full multivariate testing, but you can run a couple of variants, and do that AB testing in conjunction with your CDP segments. You could say, if someone is here from Toronto, here's the AB test that goes with that. You could get specific down to a neighborhood, or say all of Canada, or Toronto versus St. John's. It gives you, as a content person, strategist, or marketer, the ability to decide your personalization strategy. I have a lot of thoughts on this because, as VP of Growth Strategy, I'm implementing it on our own site right now. The flexibility to decide how best to personalize, whether swapping out images, swapping full components, or directing someone to a different page based on a segment, lets you build it into your overarching marketing strategy in a way that works for you, rather than being told you must personalize a certain way. That flexibility is key to staying nimble as customer behavior changes.
The 3,400 questions test
I'm conscious of timing, and I've got about 20 more things to ask in four minutes. We should reconnect for a demo day on Content Intelligence. My recollection is you gave us the chance to play with it a while back. One test we ran had it assemble all the questions people would ask to find the content on our site, which informed how to change the content to answer those questions. A really intelligent approach to crafting content.
That's key to how we tell you what's wrong with your content. We use a model to generate the questions. We have a fairly small corporate website ourselves, and for a topic with 12 pages our tool will generate around 3,400 questions to ask that content, to see where it gets tripped up. That's how we monitor whether the model, and AI search, is getting confused by the content. It doesn't matter anymore what questions someone asks ChatGPT. What matters is what questions ChatGPT asks your website after that.
Query fanout explained
That's because of how query fanout works. You can ask one question, but it's broken up into 40, 50, 60 questions you never even see. So you have to make sure your content answers all of them to give yourself the best chance of showing up in an answer. Our approach is to categorize the content across your site, understand and infer some context around it, and then generate thousands of questions per topic to fill all those gaps as best we can. As things change, that updates. It's interesting to walk through and realize it's 3,400 questions, and read through them.
How to find Squiz, and a free AI visibility report
How do people find out about you and get their hands on the product? Head to squiz.net, and you'll see a link front and center to Content Intelligence. We can run a minimal AI visibility report for anyone for free, and show you where you stand. We'll even set it up against a few competitors if you want to see how they're doing. It's a snapshot of how we approach this versus what you get with the full product, and then we can do a demo and dig in. We also have a couple of videos on the Squiz YouTube with mini demos where I walk through how we applied these changes to our own website. We'll put the Squiz link below on this video in the YouTube channel.
What's next on the roadmap
We've got a lot of exciting things happening and a lot of case studies recently published on our site. We're working with great universities, government agencies, and professional services groups coming online, both with Content Intelligence and across our full DXP, using our search and conversational search products. Stay tuned. There's a lot of cool new application of AI within content structures, strategy, and approaches to websites that will excite people. The goal is to make all of this easier for everyone who has to deal with mountains of content on their websites.
Wrap-up
That's awesome, and we're just past time. Joel, thanks a lot for making time to join me here. Thanks to everyone watching now or in the future. I'm Chris, CEO and founder of Dotfusion. We're a Squiz partner, and we implement, migrate, and modernize large-scale websites and modern content operations systems. I really recommend reaching out to Joel and the folks at Squiz to have a conversation about the product. We'll leave the contact information below. We'll see you for episode 47 in the near future.