Episode 40: Machine Experience (MX): Optimizing Websites for AI Agents and Search Bots

Chris Bryce: All right, we are live. Awesome. Well, thanks to everyone who is taking some time out of their day on this, uh, beautiful Thursday for what is now episode 40 of, uh, Beyond the CMS, where we really discuss all things content, operations and CMSs and the, the future really of content and content production and publishing.

Chris Bryce: And, uh, we're going into our third year here and we've got. Uh, really interesting, super special guest Tom Cranstoun with us today. Um, who we can also call the MX guy. Tom, welcome and thanks for joining us. Um, pleasure to have you on the show. Uh. Just before we get into the deep dive here, I do wanna let everyone know that's tuning in now.

Chris Bryce: Uh, feel free to ask us some questions. I'm gonna do a little side monitoring of the LinkedIn chat and we'll do our best to, to get back to you live in the show or we'll follow up shortly afterwards. And for folks who are watching this. In the future, either listening on Spotify, Apple, all the podcast places or on YouTube, put some comments somewhere and, and we do keep our eyes on the boards here and we wanna do our best to follow up, but um, we've got some super cool discussions today.

Chris Bryce: Um, Tom, why don't I just let you introduce yourself because you'll probably do a much better job than I will, but it's just a real pleasure and thanks for making time outta your day. I know you're on an island in the Mediterranean somewhere right now, so thanks for taking time out to join us.

Tom Cranstoun: Uh, hi, Chris.

Tom Cranstoun: Yeah, my phone. Sorry. Hi, Chris. Yeah, thanks for, uh, inviting me back on the show. Uh, it's the second time and I really enjoy our chats here. Yeah. I'm Tom Cranstoun. I've worked in the AEM world, Adobe Experience Manager. I've spent a long time building large websites for large corporations across the planet.

Tom Cranstoun: Um, and I've been working with AI for the last two years. Uh, a couple of years ago I realized that AI is better suited for reading, uh, content rather than creating content. Creating content creates a lot of risk, but reading content is what AI is now doing. It is visiting people's websites, summarizing, uh, comparing, making recommendations, ordering, listing, doing all sorts of things with your content and sending it to the users.

Tom Cranstoun: However, most websites today weren't written for AIs to, um. Read them. They were written for people. And what you do for people is completely different, what you need to do for machines. And that's what I've been working on, how to make it better for the machines.

Chris Bryce: Yeah. And I know, uh, part of that collection of things you've been working on has been some audit tools and you've been kind to share some of that with me and look forward to talking a little bit about that as the show goes on, but maybe.

Chris Bryce: Let's kind of break it down. Um, our typical listeners and watchers to the show are CMOs, VPs of marketing. There's a marketing bent and sometimes the terminology and the technology is moving faster than folks who have stuff to do during the day, um, have to keep up with, but how do you kind of lay the groundwork for some of the thinking that you're doing?

Tom Cranstoun: Well, yeah, you, a couple of years ago I started to think hard about this process. You, you see. Screens that are designed for eyeballs with things like toasts appearing and floating up to the top of the screen or accordions that flip, uh, open in shots and you've got, uh, uh. Different, uh, micro animations happening on your screen, which are great for people, but when, uh, an AI comes and it scrapes the HTML off your site, it, it doesn't look at the CSS, it doesn't look at the rendered pages.

Tom Cranstoun: It looks at the HTML that you're sending down, or perhaps it takes a screenshot and sees something at one second into your, uh. Presentation on the web, not the actual content that's hiding behind it. So it gets a, a stilted view of, uh, what your, um, your, your, your website looks like. Your content, which means it may miss your messaging.

Tom Cranstoun: It may miss the reason for building the sites. You may miss the sales. I call these invisible users. They're invisible to your analytics and they're invisible to your, um. Your sales, they just don't turn up. Have a quick look at your website, go, no, not for me. Move on. Find somebody else's website. Somebody who's got a bit more semantic meaning, who's got a bit less micro interactions, who are doing things not with JavaScript, but with HTML and JSON and.

Tom Cranstoun: The, the metadata that you need in the background. So having nice, flashy websites attract human eyeballs, but as Adobe has proven last Christmas, the, the agents are really hitting the websites now. There was a 700% increase in the amount of agents hitting Adobe hosted websites, uh, last Christmas. So there's a lot going on here.

Tom Cranstoun: And if you. Why isn't designed for them, you're gonna miss the traffic.

Chris Bryce: Yeah. We've, we've been figuring out the right language to use to communicate this to our customers and our prospective customers. And one of the analogies we have is kind of like an iceberg where humans are reading the top of the iceberg and there's kind of pretty pictures and, and copy, but the bulk of the activity is trending towards bots reading all of this stuff.

Chris Bryce: Um, and. It's not difficult to get your site prepared for the bots either. There's a bit of a practice involved in that. And do you wanna talk a little bit about maybe what's the low hanging fruit on kind of at least following best practices to be creating content that aligns with what the human sees for the bots?

Tom Cranstoun: Yeah, there's a couple of things here. One is that, yeah, when, when you design for humans, you, you, you have to be aware of the cognitive ability. Humans can only hold five or six things in their head at once, so you, you present the information piece by piece by piece, but unfortunately the machines want it all at once, going piece by piece by piece slows them down.

Tom Cranstoun: So you, you, you have to design two different views, which are presented simultaneously. The full information for the machines and the step by step for the humans, and make it clear which is which. It's also. A bit like, uh, not like the Americans with Disability Act and, uh, accessibility things in Europe.

Tom Cranstoun: If it's made easy for somebody who can't see the screen to be able to understand what's happening on the screen, then you're actually making it easier for the computers to understand what's going on. So if you treat, uh, a, a, a computer like a person with a disability, then. You know, we should have fixed this 17 years ago when, when the disability thing came out.

Tom Cranstoun: But, uh, it's now a, a, an imperative to get the structure right, to get the framing right, to get the, the metadata right, and to get the HTML right. That's a simple practice to sit down and work out what's missing to help, uh, these invisible users. These people who are. Glancing at your screen and moving, moving on.

Tom Cranstoun: So that, that, that's part one. The second part is to look at your micro interactions and the steps that are happening in your, uh, websites. Uh, you, you, one of the other things that I think about is the, um. Headless architecture. I know that you focus a lot on headless work in Dotfusion, Chris, but headless is sometimes a, a barrier to agents because the agents won't necessarily execute your JavaScript and complete the actions.

Tom Cranstoun: They may just pull down your HTML and just look at the HTML. Different agents have got different capabilities. There are one and a half million models at the moment on Hugging Face. And in three months, another million will be added. It's so fast. And Hugging Face isn't the only repository of models. Any corporation with, with sense will be building their own models in-house and keeping it secret from the world.

Tom Cranstoun: So you can think there's probably two and a half million models out there already. Any one of which may come and look at your website using a foundation model. OpenAI, Grok, uh, Claude. Then they have got superpowers. But if you've got a scanner that's scanning, uh, a barcode, it might have a model on board. It might only have a few million parameters. It might not be very clever. So you have to design your websites to work at all ranges of models. All kind of, uh, LLMs, not just the, the top of the market ones.

Chris Bryce: That makes a lot of sense. And just for our viewers who are getting familiar with this concept of metadata and schema data, can you just break that down into kind of layman's terms? What, what are those, what does that mean?

Tom Cranstoun: Yeah, it is metadata, metadata, data about data that doesn't help people. Uh, I, the best way to describe this is it is like footnotes. Footnotes in a document. We all know you're reading the document, you get a little number and down at the bottom of the page, it gives you a wee bit more detail, perhaps a link to a website. That is actually metadata.

Tom Cranstoun: It's telling you something that you might not need to know, all of the details to understand what's going on, but if you wanna dig a little bit deeper, then the footnotes will tell you more. And when you're working in the academical world, people usually create piles and piles, and you find out that the footnotes are sometimes bigger than the document.

Tom Cranstoun: But in, in normal writing, in normal books, if you're reading fiction, you don't get footnotes, but, but in technical books, yeah, if somebody's made a statement and they want to give you the provenance and tell you what it's all about and more detail, a little footnote at the bottom of the page. You can read it if you want. You can ignore it if you want.

Tom Cranstoun: Personally, I read every single footnote I come across, and, uh, I'm creating my own books, which explain the, the, the machine experience, the MX, which is what I'm calling it. The MX guy is different from UX, which is user experience. We talked about the user experience. Short attention spans, give things to them slowly. Machine experience, infinite attention span, give it all at once. These two things have to blend together, and metadata is like a footnote. It's additional things you don't really need to know unless you really need to know.

Chris Bryce: Yeah. One of the examples I use, um, to communicate what metadata is, and many people are familiar with, when you take a photo on your mobile phone and you look at it later and you look at the info and it tells you the aperture setting, the time of day, where you were, it's all that data associated with the photo to give you some clues on what it's about, from a more of a technical perspective almost.

Tom Cranstoun: Yeah. It's, it's like a footnote for your picture.

Chris Bryce: Exactly.

Tom Cranstoun: Yeah, exactly. Everything has got metadata. If you look at, uh, a PDF, it's got who authored it, how many images are in it, how many words are in it. Those are all hidden inside the PDF.

Tom Cranstoun: If you look at, uh, a scripting language for a program, HTML for a website, they have comments, and those comments are actually metadata. They tell you something extra that you don't need to see on the screen. You don't need to know to run the program, but it gives you extra things like who wrote it, when it was written, why it was written, and these are forms of metadata.

Chris Bryce: So key point is, uh, for those of you who don't know, every, let's call it page of your website should have some associated metadata that describes to a, a robot or an AI bot what is it that's on that page. Technically, it could be structure, it could be how it's scaffolded, but you also mentioned schema data just to make things a little more spicy. What is that?

Tom Cranstoun: Oh, that's the technical definition from, uh, well, it was Google and Microsoft, setup schema.org many, many, many years ago, and its intent and its use is for common things like pricing. There's a structure that you apply to pricing. Uh, a very good case in point, uh, I was looking for a holiday.

Tom Cranstoun: Ooh, I'm currently on holiday, I do a lot of this. I was looking for a holiday last year, and, uh, I got Claude in this case to search some venues for me, some destinations. It was a cruise, and it came out with a bunch of cruise companies, and one of them came out at a price of $200,000 for a one week's cruise in the Dan Europe, whereas everybody else was around about the $1,800, and I thought, 200, it's pretty. What's gone wrong here?

Tom Cranstoun: It had created a table of cruise companies' prices and hadn't structured them according to any internal guidelines. It had just gone off, grabbed the websites. Now the Europeans use commas and dots a different way around from us, uh, Anglo-Saxons. We use two comma zero zero to mean 2,000 and zero zero to mean and no cents. So the Europeans do it the other way around, they say 2 0 0 comma zero zero.

Tom Cranstoun: So Claude got confused by different websites from different companies. I was looking for a cruise on the Dan Europe, which is a European destination, and I had British companies and European companies mixed up in the table, and Claude got confused between the pricing from two different cultures.

Tom Cranstoun: The way you get around that is that you use schema.org, which decorates the pricing in a fixed manner. schema.org says, and it's firm, that you must always use decimal points to separate the major currency from the minor currency. You must always use, uh, no commas to split up the, the uh, hundred thousands of whatever you're talking about.

Tom Cranstoun: So it's a structured way of ensuring that you put that into the document in a machine readable format. It doesn't get displayed on the screen. So that, that's what schema.org is all about. There are other things that happen. We all know about the American date format and the European date format where we swap the day and month around, that causes confusion.

Tom Cranstoun: But if you use schema.org and micro data, there is a rule that says that you put it in year, month, date order, and then the machines can look for those elements rather than the prose elements that you have on your page. You put the prose up for humans to read and the schema.org for the machines to read. You put both on the page and it keeps both happy.

Chris Bryce: Cool. So we've got your human data, copy, images and things that make you feel good and give you some information. We have metadata associated with the page and what's on it and how the bots actually quickly assess what's happening. And then we have schema data, which puts structure to patterns of data that you'd see on many other sites.

Chris Bryce: So it can kind of harmonize all of that at once. And now, um, we've been tracking Google and Shopify's kind of, uh, conglomeration around putting together this idea of a universal commerce protocol, which is somewhat related as well, but I believe what they're up to is they wanna provide a standard for the bots to deal with commerce and commerce workflows. Do you wanna share a little bit about what, what you know about that?

Tom Cranstoun: Yeah. Yeah. It, it, it starts off, well, it's a bit, and Web MCP is in a similar vein, which is another, uh, standard which has come from Microsoft. And these are the things that are helping the agents to operate in, in, um, uh, a coherent and, uh, aligned manner so that they all do the same thing in the same way at the same time.

Tom Cranstoun: But before they can get to that point, they have to decide that they want to do UCP or Web MCP with your website. So they have to understand your website enough to be able to say, yes, this is a candidate, let's go for it. There are, well, I, I think that there are five stages you have to go through before you get to the sales point or the CTA point.

Tom Cranstoun: You, you're not talking about an agent coming and buying things off of you. It doesn't happen just like that. You have to be discoverable, which means that you have to be in a thing called the Common Crawl. Uh, all of these agents are trained on a, on a body of work, which is a majority of information from the open internet, things that haven't been excluded by robots.txt, things that have been included by your sitemap.xml, things that are historical in nature. Many, many years ago, perhaps you published a website once.

Tom Cranstoun: Uh, there's this tool that goes through the internet and creates a massive, uh, body of everybody's website and pours it all together, taking the HTML and the text from the HTML. It doesn't look at the CSS, it doesn't look at the JavaScript, it doesn't look at llms.txt. It just takes your HTML and structures a view of everybody's website that it can find that is publicly shareable.

Tom Cranstoun: Now, this gets updated once every month and gets fed through the LLMs perhaps once a quarter, maybe once a year. It's a very, very slow process. It's the historical process that the, the training is done for the, for the machines. So the machines get given all of this information, and then the companies spend billions reading this, structuring it, extracting information, and building the backgrounds, the non-live stuff that the machines know about the world, the internet, and it's not just HTML pages.

Tom Cranstoun: They take books, they take magazines, they take print. They take any kind of human communication that they can and suck them into this giant crawl. I've written an article about the, the Common Crawl on my website, but it's a massive piece of work. About 44% of the Common Crawl is in English and no other language reaches above 6%. So if you're a German user, you are better off prompting machines in English than you are in German.

Tom Cranstoun: And if you're a Canadian user, you're much, much better off prompting the machines in English than you are in French. Uh, the machines have limited other understanding. It was all written by tech bros in America. Come on guys, we have a problem there.

Tom Cranstoun: But so I, I recommend prompting and using AI in English gives you a much, much, much better result than going for, uh, a language variant. Especially since the language variants themselves, uh, if you take German, you've got Swiss German, you've got German German, you've got, uh, you know, got different Germans. If you go to French, you've got Canadian French, French French, you've got Caribbean French. So, and Spanish is even worse of course, because of all of the varieties. So when you go to a minority language, you're one step off.

Tom Cranstoun: Now, I'm saying discovery, that's the first point, Common Crawl. Then you've got to be citable, you have to have enough good information in your historical information before, uh, the machines can cite you in a question without them having to go off, search and find you. So there's that. Then the third point is you've got to be available live with pages that return HTML that can be read by the machine, not just a a, a view script that has to be interpreted.

Tom Cranstoun: Then you have to have enough structure on the page for a live result. There's no point in a machine giving you the price for an Apple Mac that was two years old on a website way, way back. It needs current pricing, up to date, and new machines are coming out, you know, new Macs coming out all the time as we've just seen this week.

Tom Cranstoun: Uh, so the pages have to be live, up to date, searchable, scannable, and readable by machine. And when you get to all of those points, you can then consider the universal commerce protocol. If you fail in any of the ones leading up to them, it, it, it won't matter, because you won't be in the commerce marketplace if you've failed anywhere else in this type.

Tom Cranstoun: Commerce is glue on the top that makes the machines line up and understand. Using the commerce protocol means that they have got a way to get standard data back from your website, and that standard data must still be MX compatible. If it doesn't, it may, you may give it an answer, but the answer has to be congruent. It has to be decorated with metadata, it has to have the correct structures for pricing, et cetera, or else the machines may still hallucinate. So the, the UCP doesn't fix a problem, it orchestrates the machines. You still have to do the machine experience to fix the problem.

Chris Bryce: Yeah, and it's from an agency perspective, it's making the vision and the sales process quite difficult because we have so much to explain. It's kind of like we're building these F1 cars now where most people just think they look cool 'cause the aerodynamic package and delivery looks sweet. But the amount of engineering that goes into a highly performant website is insane. And it's, it's a long process to justify the value.

Chris Bryce: When in the marketing mindset, it's like, let's get this done and out there like we've always done for 20 years. And, uh, so it's kind of discussions like this that I hope people get a chance to sit down and listen to and just to start to get the idea that there's a lot of engineering under the hood.

Chris Bryce: Um, and you know, what do you, what do you do first and how do you prioritize it to a budget? And you gotta get something done. And you know, at the end of the day, our customers are like, they're expecting this perfect thing that's infinitely thought out. Um, yeah. But anyway, I know you've spent your, your life in kind of AEM land and you know a fair bit about content platforms. How are you seeing the impact of platform in all of this, AEO, discoverability?

Tom Cranstoun: Well, yeah, that's a good one. Um, one of the things that I, I, I cover in my book is the idea of, um, it's, it is this sovereign data, the fact that you own your data. If you're selling on Amazon, Amazon has got your reviews, and they hold onto your reviews.

Tom Cranstoun: And if you decide to switch platforms, which, okay, switching from Amazon to Alibaba may not be on the cards, but you know, switching platforms, uh, you leave your reviews behind, you've, you've lost out on things that you have built, you've built trust up with your users, and you've got a lot of things in there. These things actually belong to you. They don't belong to Amazon. They were created on Amazon and the platform owns them, and the platform is using that as a lock-in to keep you on their platform.

Chris Bryce: Yeah, that's what they're trying to do.

Tom Cranstoun: Now, my books and my philosophy is that this is sovereign data. It belongs to you, it doesn't belong to them. It was created because of your interaction with your customers. It may have been hosted on a platform. I see the future where the CMS is responsible for taking those reviews, internalizing them, and publishing them as metadata with provenance and, uh, the author's name and where it was created and why it was created and when it was created, and publishing it in an open way such that the platform doesn't matter anymore.

Tom Cranstoun: You own your sovereign data and it travels with you, and it's not locked in a CMS, it's not locked in a platform. It's actually at the publishing point in the HTML that you serve to your customers, to your bots, your agents. There's no point in having a webpage that says, I'm good on Amazon, because the AI isn't gonna go and have a look at Amazon after it's looked at your webpage.

Tom Cranstoun: Uh, so you need to keep this as part of your enterprise asset layer. I call this EAL, and this enterprise asset layer is the things that you own. The what that you do that is currently spread. You may have a bit on Twitter, a bit on X, sorry, a bit on Amazon, a bit on wherever, you need to hold this as your sovereign data and make that part of your offering to the machines.

Chris Bryce: You brought up an interesting point about the history of your, of your assets, of your domain, and, um, for the past, a little over two years now, we've been actively researching how do you actually get found in an AI search, or GEO, or AEO, or what have you.

Chris Bryce: Uh, what I was trying to do was establish a theory that I could be somewhat argumentative against the amount of people on, let's say LinkedIn, for example, that have the secret to getting found. Um, and there's a lot that goes into getting found, and there's a considerable amount of luck at this point, frankly, but there are things that you can do.

Chris Bryce: So one of the starting points is define what do you wanna get found for? We still have customers that say, our SEO sucks. And we're like, well, what do you wanna get found for? Like, how do you know it sucks? Like, what are you actually addressing? Um, so for us, and back to your earlier point, we build large scale headless websites as our primary product.

Chris Bryce: Um, and you're right, there are all kinds of tricks with Next and React and making sure that your JavaScript is presenting nice, clean stuff. Um, but we had a target for kind of best headless agency in Canada, so, so a bit of geography in there, clear focus, a bit kind of niche, long tail. Um, so, I don't know, after about a year or so, we start coming up in front of Chat and Claude and those guys for that keyword.

Chris Bryce: And I'm like, well, okay, we've been following a protocol, but why is it actually that you have us rank first? And Chat said back to us, it said, well, you've been around for 25 years and you say on your website that you're the best. What is going on here? I thought it was our Wikipedia sighting and our Clutch reviews, and that, as I discovered later, is foundational to that thing that it got back to us.

Chris Bryce: It kind of made it as simple as that, but, um, yeah, you're right, there's a lot of, uh, structural work to do, and what's really made a difference for us is building schema, and mostly schema, where our metadata was always pretty good, but really unique schema, and something that not a lot of folks have mentioned to me, we go back and update our blog posts that we had posted from three or four years ago.

Chris Bryce: Uh, and there's kinda a bit of a movement of, like, are we getting penalized by having stuff up there that was a little bit keyword stuffed from five years ago? Now it's just this lurking kind of like space garbage stuff sitting around. So instead of destroying them, we've gone in and updated them. So here's a fresh one we talked about.

Tom Cranstoun: I think that's a good idea, Chris. Yeah, I need to do that too, my stuff is well out of date because you talk about what AI was two years ago, it's completely different now.

Tom Cranstoun: Yeah, so my blog posts are, yeah.

Chris Bryce: Okay, good point. Very good point. Yeah, keep things fresh, keep them interesting. Not just, not just the machines, but keep people interested.

Tom Cranstoun: Yeah, yeah, exactly.

Chris Bryce: Um, maybe there could be some confusion, and you could help with some clarity around, like, even the difference between GEO and SEO and how are they alike, how are they different?

Tom Cranstoun: Ooh. Yeah. Yeah. So a surprise question for me, a very surprise question, and I'm thinking hard here. I mean, SEO is, um, yeah, it's, it's how you get your stuff into Google, is what SEO's all about, and that is a historic process. Whereas GEO is how do you get your stuff now into an AI, which is urgent and immediate.

Tom Cranstoun: So there's a difference there. I see SEO as being like the Common Crawl we discussed earlier, and the, uh, GEO being like the live web search that you do now. So SEO tends, I, in my mind, is setting down historical anchors that say who you are and what you did, whereas GEO is always who, what you are doing, rather than what you did.

Chris Bryce: Yeah, interesting way of thinking about it.

Chris Bryce: Um, hey, this is a problem with this show where all of a sudden we run outta time when we really get going. Um, but I'm cool to stay on because I really wanted you to, uh, share a little bit more about your, your MX Audit.

Chris Bryce: And I just thought that, I don't know if you're open to discussing that, but if you are, I think it'd be really cool. Um, also parts of the show get re-syndicated as little snippets, and I think this would be a really great one if you can just tell us about, uh, what do you call it anyways, the MX Audit?

Tom Cranstoun: Yeah, I call it MX Audit. Everything I'm doing is MX something, that's why I'm the MX guy. Machine experience. The Machine Experience Handbook is the book I'm publishing on the 2nd of April. Uh, I have a free chapter which I'm gonna pass around to anyone that wants it. Uh, I also have MX the Protocols, which is coming out in July, which is really in depth, uh, really, really in depth. It's a thousand pages. The Handbook's only 200 pages. The protocols is huge.

Tom Cranstoun: And MX the Audit, what we do here is I have, uh, used my book to train a machine to look at people's websites, and it does it in a scripted manner. So it takes a look at your website and does what, uh, Screaming Frog might do for SEO, it does it for machine experience.

Tom Cranstoun: So it takes a look at your website, it looks at the HTML you serve, and it looks at the HTML that's rendered, so it can catch the difference between a simple website or a server side rendered website and an Adobe EDS website, et cetera. It looks at both stages, and it can tell what a simple AI will see and what a complicated AI will see.

Tom Cranstoun: So it looks at those things, it looks for micro interactions, it looks for accordions, it looks for videos that autoplay, it looks for all of those things that may confuse a machine. It also looks for the key anchors, like your robots.txt, your llms.txt, which is a thing that isn't popular at the moment, and I know why, and I hope to get this changed.

Tom Cranstoun: But the, um, it looks for the key markers for things that will help LLMs, machines, I should say. So it, uh, looks at your JavaScript, it looks at your JSON-LD, and it builds a picture of what your website looks like. It then compares the human view with the machine view and generates a report that says how you can improve and what should be done, the quick wins and the slow wins, and it's tailored to your website.

Tom Cranstoun: My audit report, also, while it's at it, only samples 10 pages from your website unless you want more done. A 10 page sample is, uh, enough to get a feel for what's going on. If you wanted to get your full website done, you know, a large site like Nissan, it's got thousands of webpages. So you need to, uh, engage the machine lightly, find out what the main problems are, then get a plan, and then move forward into solving each of the steps.

Tom Cranstoun: Um, yeah, it's, I think it's incredibly useful. Uh, when I first wrote it, it was negative. And, uh, so with a bit of feedback from some of my testers, thank you, Chris, it must really annoy on our site, man, but, uh, we got a lot of great valuable information from it. Yeah, it now looks, and gives a positive spin to everything that it does.

Tom Cranstoun: Uh, obviously people built websites for humans. Building it for machines is a completely different proposition. Uh, it's tied with disabilities, it's tied with WCAG, it's tied with SEO, but it's different, and it's a different way of looking at things. And to go in deeply, as you said earlier, uh, if you start writing and building your pages with proper metadata, that means you have to go back to the content source and decorate those too, which means practically you're rethinking the way that your business presents itself to the world.

Tom Cranstoun: And this is not a simple task, as you said. Uh, and to get it into and outta the far end means that you've got to get the source into the Common Crawl, the Common Crawl into the machines, the machines into the training, the training into the live production. That can take many, many months. So you start where you can, you improve the live stuff, eventually the historical stuff will improve.

Tom Cranstoun: Uh, there's, uh, it's a long tail, a very long tail. It's a long tail end.

Chris Bryce: From an agency perspective, um, it's interesting in that we've taken your concept, because I know that you have an agency kind of product line for your product as well. So any agency folks listening in, reach out to Tom. Um, so we're kind of putting your product as the, uh, kind of the fundamental audit of a package, which basically helps large organizations optimize their site for the robots, knowing it was originally built for humans.

Chris Bryce: And, but it's an interesting situation for us where we're pretty open and fully disclosed, but we're going back to our clients saying, wow, like when we look at this from a bots perspective, there's a lot of work to be done, and sometimes we have to take a little bit of heat about, well, why didn't you factor that in? And that's fair enough, and it's like, times change, and you know, we're getting ahead of it.

Chris Bryce: Um, now our clean SEO work certainly was helpful to get along the way, um, but you just find these kind of little holes everywhere. So we also have a very good process, after you get the audit from the MX Audit, we have a really kind of clear, almost sprint based workflow to start chipping at the low hanging fruit and getting things all cleaned up.

Chris Bryce: But, um, awesome. Uh, it's fair to say that the MX Audit is a repeatable thing, right?

Tom Cranstoun: A month later it'll tell you the differences between the two stages, and you'll see that you're progressing and how much more work you've got to do. It's very useful.

Chris Bryce: Yeah. Super cool. And look forward to working with you more on this stuff. Um. Hey, any final comments before we wrap things up here, Tom?

Tom Cranstoun: Uh, yeah, I'd just like to say, well, one thing that really, really, really made this all work is the CMS Experts Group. I really must thank Giannis and, uh, uh, Matt Garippa, because they're both involved in this. They got me into the AI HTML thing two years ago, before it was a problem. And I've been thinking and thinking, working and planning and talking to CMS experts. Uh, and it's been really, really good. So a shout out to those guys.

Chris Bryce: Um, awesome. Yeah, they do a great job of keeping, uh, that community strong. And, uh, that's where we met as well, so.

Tom Cranstoun: Yes, thanks, with the good vibes along the way.

Chris Bryce: Um. Okay, well, hey, thanks to anyone who's been listening in now for a while.

Chris Bryce: Uh, if you are tuning in, you're probably kind of keen on the subject matter. Um, and for folks who are watching this, uh, later on into the future, thanks for making some time. Reach out to us if we can answer any questions along the way. Uh, we're gonna be back in a couple weeks with episode 41.

Chris Bryce: Uh, Tom, have a wonderful holiday. Stay warm, enjoy the sunshine, and, uh, we're gonna sign off right now. Thank you everyone, have a wonderful day.

Tom Cranstoun: Thank you, Chris. Cheers.

Chris Bryce: Let me hit some buttons here. Okay.