Transcript: Get discovered: Adapt your marketing for the AI search era
Go back to the main session page.
Stephanie Small: Hello, everyone. Thank you for joining us for today’s Tech Forum session. I’m Stephanie Small, associate product manager at BookNet. Welcome to Get Discovered: Adapt Your Marketing for the AI Search Era.
Before we get started, BookNet Canada acknowledges that its operations are remote and our colleagues contribute their work from the traditional territories of the Mississaugas of the Credit, the Anishinaabe, the Haudenosaunee, the Wyandot, the Mi’kmaq, the Ojibwa of Fort William First Nation, the Three Fires Confederacy of First Nations, which includes the Ojibwa, the Odawa and the Potawatomie, the Métis, as well as the unceded and ancestral territory of the Musqueam, Squamish, or Tsleil-Waututh peoples. These are the original nations and peoples of the lands we now call Beeton, Guelph, Halifax, Thunder Bay, Toronto, Vancouver, Vaughan, and Windsor. We encourage you to visit the nativeland.ca website to learn more about the peoples whose land you’re joining from today.
Moreover, BookNet endorses the calls to action from the Truth and Reconciliation Commission of Canada, and supports an ongoing shift from gatekeeping to space-making in the book industry. The book industry has long been an industry of gatekeeping, and anyone who works at any stage of the book supply chain carries a responsibility to serve readers by publishing, promoting, and supplying works that represent the wide extent of human experiences and identities in all its complicated intersectionality.
We at BookNet are committed to working with our partners in the industry as we move towards a framework that supports space-making, which ensures that marginalised creators and professionals all have the opportunity to contribute, work, and lead.
If during the presentation you have any questions, please use the Q&A panel found in the bottom menu.
And now, let me introduce our speaker, Ariel Hudnall. Ariel is the managing director of Serif, a literary marketing agency specialising in book publicity and advertising, author platform development, and storytelling strategy. She holds a Master of Publishing from Simon Fraser University, where she also teaches digital book marketing as an instructor. She’s on the board for the Canadian Book Club Awards, and over the last 10 years, has worked or volunteered in the industry as a beta and slush reader, developmental and acquisitions editor, book designer, advertiser, publicist, and marketing consultant. She lives in the Cariboo region of BC, the ancestral lands of the Nazko, the Lhtako Dene, and the Carrier-Tŝilhqot’in peoples.
Thank you for joining us today, Ariel. Take it away.
Ariel Hudnall: Thank you so much, Stephanie, and thank you to everyone who’s joining us today. I’m truly excited to be joining Tech Forum again to talk about discoverability and how that intersects with this new age of AI search. We’ll go ahead and get started.
So, Stephanie gave me a very nice introduction here, but I want to quickly mention where I work, which is Serif. Serif is a literary book marketing agency that works with all of North America. We do publicity, advertising, both digital with Amazon and with print publications. We work on social media, and most importantly, we help publishers and authors find the best strategy for their books, both online and direct to readers. So, I’m bringing all of that knowledge and our long history with clients to bear in today’s session.
So, discoverability has changed a lot in the last three years or so, but particularly with the advent of AI search. There’s a number of acronyms we use here, which is AEO, GEO, AIO, and, of course, our old favourite, SEO. But AI in particular has disrupted how we discover things, and it’s affecting you already, whether you know it or not, and whether you’re using AI or not.
Today’s presentation takes a neutral stance on AI. I’m hoping to speak to people that are using AI or not using AI. Largely, these recommendations come from meeting the changes of the industry, whether you decide to use these tools or not. There’s ways to combat potential drops in discoverability you may be experiencing.
So, some of the things that have already changed include, of course, the influx of AI-generated content. We see this especially on Amazon, which is unfortunate for many publishers and authors because it can be anywhere from 70% to 90% of your entire book sales channel. The current estimation puts AI-generated content on the book pages of Amazon at around 52%. So, 52% of all books on Amazon are AI-generated, which is a lot of noise to get through when you’re trying to publish a book.
Secondly, search has shifted from kind of keyword-focused to query-first. So, basically, recommend me some books about being a new mom or things like that. So, search has shifted to query-first. What this means is that your keywords and all of your descriptions — so, book descriptions, your author bios, which are another form of description, website information, even blog posts — need to be thinking about how people discover content through questions, and particularly how AI looks for the answers to those questions.
So, there’s going to be some magic making between how we think humans ask questions and how we think humans answer questions, but also how AI interprets questions and how AI amalgamates answers. If you want more information, particularly about keywording and metadata descriptions, please go to the first session from this series presented by Tricia McCraney of Biblio. It happened two weeks ago. I believe a link will be dropped in the chat shortly. Because there’s lots of wonderful information there that would be relevant to all publishers.
Finally, metadata now favours the early. If you want to own the completeness of your book’s information, you need to be the first to fill the blanks. Don’t let AI or a different retailer or someone in the media misinterpret what your book is by not having that information available first. So, complete all of that ONIX metadata, and also make sure that your website is as robust as it can be.
Some other risks that are rising with this impact of AI is that — of course, I think it’s everyone’s fear — IP, your intellectual property, can be scooped, scraped, and duplicated at record speeds. We’ve seen a concerning increase in particular about kind of the correlation between a book becoming more popular and the quantity of imposter books that appear at the same time to try and siphon some of your profit away through a fake book.
This is not restricted to the book industry. Just last week, I was looking for a movie to rent on a digital…like Apple TV. And there were three imposter Odysseys on the movie pages that all came out in 2026, and none of them were the blockbuster. So, we’re not the only ones being targeted, but it is something we need to be cognizant of, especially as we start to see specific books doing better. Make sure that you are keeping an eye on your IP, tracking your IP.
Okay. The other important risk that could happen here is more mundane in the day-to-day. Unclear positioning, not having the right wording, the way of talking about a topic can bury content before it even has a chance to be seen by an audience. A really great example of this, which has come to light recently, is, for example, on Instagram…Instagram automatically assigns a topic, one topic, to your account. And if you are somebody who’s publishing and doing deep content on many different topics because you publish many different content types, Instagram may be misaligning your account with different spaces.
So, for instance, as a marketing agency, when we do social media, we often will either fall into the Bookstagram community. So, people who love books and just want to talk about books ad nauseum. And people who are looking for marketing services. So, authors talking about, here’s the best tips for X, and here’s information about Y. And I have to decide, as a marketing agency, which version of us is more valuable in the long run for our goals on social media. The same goes for the content you’re posting, either as an author or publisher or bookseller. You need to decide and coordinate your content to reinforce the topic that you’re talking about.
Where this comes in with AI is not just the algorithm, but also because social media content is now indexable by AI engines and chatbots. So, every caption, every piece of content is more important than ever. And they should all be reinforcing the primary goals and messaging that you hope to bring to the public.
The final risk here, of course, is one that is well-known. AI can misinterpret and hallucinate if you are not careful with the word-choicing that you make.
Okay. I’m going to go ahead and go to the next slide now.
So, I’ve talked a lot about search, but AI’s influence isn’t just going to affect search. There’s many other ways that it’s going to be impacting supply chain, demand, discovery, that isn’t just through our search bars. So, a really great example of this is, outside of publishing, e-commerce is changing rapidly. If anyone has tried to buy things online, you will certainly have noticed a shift. And I think, in publishing, we are often siloed into the way we do business, but it’s important to keep a bead on what’s happening outside. Because there’s lessons we can take from other industries that could help all of us make more income.
So, number one, of course, is mobile wallets. So, this is being able to pay with Apple Pay, being able to pay with Google Pay, any other one-click thumbprint, don’t have to pull out your credit card or give people your credit card information, really important for security. So, it has quickly become the gold standard for online ordering. And it’s going to become pretty critical infrastructure across the board, I imagine.
The next one is Google’s UCP system, which is Universal Commerce Protocol. This is already existing to some extent across the internet, but it is now integrating AI agents. So, rather than thinking about this as something that is irrelevant to you because you’re not going to be using AI tools on your commerce site, think instead about people, let’s say an educator who is quickly looking to build a class list of textbooks. They might go to an AI agent and say, I need four textbooks on this topic for an upcoming semester because they’re, building their classroom list, and the agent will pull up four listings. If they’re not applying them for a classroom, just for themselves, it becomes a one-click, buy all four. So, it becomes less friction, easier to get to the thing you need without having to do a bunch of research. The agent is [inaudible 00:15:47].
And so, agents have been adopted fairly quickly by the consumer base. Anecdotally, even in my own family, people who are very cognizant of politics around AI are still using it in this particular way, whether they know it’s AI or not.
So, with the UCP system, which I have kind of animated here, you can see someone’s asking for a very specific product. The AI thinks about it. It gives you an example. You can click through. It asks you to create an account with a single sign-on. So, this is the next piece that’s incredibly important for new e-commerce. We see this with Bookmanager.
So, for example, I create an account for my Bookmanager here in Quesnel. I can use the same account to buy from a bookstore in Vancouver or Halifax, or gifting for somebody else across the country. That’s basically a single sign-on without the connector to Google or Facebook. That’s more typical. These are considered better in long-term situations because, again, security improvements and ease of use. People don’t need to remember 50 keywords. They just need to remember the single one, which is usually protected by thumbprint.
So, e-commerce is rapidly changing, and our industry hasn’t kept up. The large amount of e-commerce that happens within industry platforms, on individual websites and so on, are fairly secure, but not ready for the AI shift. Okay.
All right. In a hypothetical scenario, maybe you are somebody who only cares about physical spaces. I’m a local publisher, and my relationships with my bookstores are everything. I’m not as concerned about Amazon. In fact, I refuse to sell on Amazon. These situations occur and are completely valid. But it’s going to change, too.
So, as an example, we have been using ISBN codes since the 1970s, and they were revolutionary for helping with the book supply chain. But they are limited in the amount of information they can hold. And if you’ve ever been a book designer, you know they can take up quite a bit of real estate on the back end of your product.
There’s been a big shift, especially with new EU policies that require to know, where did your paper come from? How is your ink sourced? All of these things to think about environmental impacts were the shift to move to 2D barcodes.
You may have actually already seen this around your grocery store shopping with toothpaste boxes where the QR code is very small. This is to help take up less real estate, add more data. But it has an extra element that I think is really important for publishers to already be thinking about, which is customers can scan that 2D code and get a bunch of information. You could send them to a landing page. You could give just directly to kind of a metadata sheet if it’s for a bookseller, etc. The key here is it allows you to have more ownership over the information that is being sent.
And conversely, in the AI era where AI is interpreting everything online, in an in-person space, you can set the record straight. So, 2D codes is potentially another thing to be thinking about. Okay.
So, does this mean everyone has to use AI? Is that really the solution here? Well, we can go ahead and do a thought experiment.
So, Bynder, a marketing survey company, which I’ve linked to in my slides, which you will get after this presentation, conducted a survey. They asked a thousand people to look at two articles. One was AI-generated and one was written by a human. Forty-five percent of these participants couldn’t identify the AI. Fifty-five percent could. And interestingly, because AI use is more rampant in the U.S., U.S. Americans were 10% more likely to identify that AI. So, people are, as consumers, really learning to spot AI. And what does that mean?
So, at first glance, the survey results are fairly interesting. Fifty-six percent of the participants preferred the AI-written article. So, does that mean the AI-written content was better? We can’t really make the judgment from the survey, but what we do know is that this result was only when the participants were not told which article was AI.
When asked about how they feel about content they suspect to be written by AI, the respondents were much clearer. Fifty-two percent will become less engaged if they suspect AI was used. So, we’re going from 56% preferring the AI written article to 52% saying, basically, if I clock AI use, I’m 52% more likely to distrust it and move away from it.
So, the key word here is suspect, because maybe some of you have also had this experience. You might not be using AI at all, but you’re getting comments, this is AI-generated. So, suspect is the key word here.
So, where are the trust gaps when you’re thinking about using generative AI? I put these in order of better trust to least trust. So, clarity is the one place where currently AI purportedly excels more than humans, and that’s in organization, outlining, and structure. So, basically, 82% of a different study with the International Journal of Science, Strategic Management and Technology found that AI-organised content was clearer, easier to understand, helped them get to their answers more quickly. So, excels in this particular space.
However, in terms of actual trust — so, do they believe what the AI is telling them — only 50% of participants did. So, that kind of goes back to that 52% when they know it’s AI, they’re less likely to believe the outcome.
For engagement and particularly CTAs, which is calls to action, human content performed better. Seventy-eight percent of human content was more engaging, more likely to result in a click-through. But it was 52% for AI. So, it works, but only marginally. And I think, really important here is this is one data set by itself. When we’re thinking about this in the context of, for instance, an ad where you only have a 3% click-through rate to begin with, AI’s additional barrier changes that number, too. So, they correlate in many ways. And you should expect that, in some cases, it may improve your chances, but in many cases, it may not.
The worst performances for AI came down to authenticity and emotional depth, which is unsurprising. When it comes to being authentic, human content wins. Less than perfect is better than AI. Typos, better than AI. For emotional depth, also incredibly important. As you’ve probably guessed, AI’s emotions are fairly shallow in the sense of it can perform happiness to an average, but it doesn’t understand give me happiness with a bittersweet undertone or something like that. So, emotional depth can be in particular a hard thing for the AI to replicate in a way that consumers find natural.
Okay. There is one more stat that I think is actually more important, and that’s how customers are using AI to understand what to buy. So, in fact, through a study with Reuters, 54% of users that have been recommended a product by AI, because AI is crawling everything, are more likely to convert them to the product that it suggested. And additionally to that, they were 53% more likely to spend more time on the website where that product was being sold.
So, if you’re being indexed and surfaced by a chatbot to say, this is the book to get on this particular topic, not only is that consumer more likely to click through, you have a less of a barrier to entry. They’re also going to maybe spend a little bit more time on your website, learning more about you and what you publish. So, thinking about product pages and are those related books at the bottom really related? Is there a way that you can, for instance, join the mailing list on every product page? That kind of thing. So, thinking about that kind of stuff can really help with further conversion.
But ultimately, what all of these stats are saying together is that the trying to be human with AI is the issue. Consumers like AI to have that impartiality, even if it’s biased and sycophantic. When they feel like AI is being unbiased, they’re more likely to respond positively. Okay.
So, where does suspicion start? And I’m using suspicion very carefully here because it’s not just when you’re using AI, it’s also when the consumer thinks you’re using AI. And this is happening a lot, especially for, conversely, people who have been really good at their marketing game pre-AI, had great photos, excellent copy, down to a science. Now, it almost increases the likelihood that people think you used AI because it’s so good compared to what they’re used to seeing.
So, suspicion in this case starts in the uncanny valley, which I have given a description to here. But it’s basically why we don’t like wax figures. They feel alien or not quite right, the sheen of skin and so on. And that’s what makes us have that visceral reaction to things that don’t feel normal.
Common triggers when using AI-generated material, of course, book covers, especially when text and content is produced by AI. AI-generated event graphics and flyers, very easy to clock, but also very important in doing book launches and things like that.
Friendly AI chatbots. So, there’s a big difference between, hey, Gemini, can you search this for me, kind of thing and using VoiceGPT that has the kind of ha, ha, ha, yeah, right, let’s talk about that, where it laughs and breathes and has different intonations in its voice that can start to move into the uncanny valley for a lot of folks. Same with, for instance, AI audiobook narration is a really great example of where, for some, it elicits this kind of uncanny valley response. But for others, it doesn’t. And it may depend on the programme you’re using.
Text that doesn’t fit expectations. It feels much longer than it should be. The structure is just a little too clean or repetitive. It’s very forward with its praise of you. It’s trying to win you over. And similarly, emails that elicit deja vu. So, I learned about your book, and I was so excited by rehashing the book description. Nothing new mentioned.
Finally, if you’ve used AI to retouch your author photos or created character art, these can elicit an uncanny valley response. And most importantly, if you use AI for any of these things and it’s clocked as being AI, you have now the burden to prove that the rest of the project was not AI-generated. So, if your cover is AI-generated and it’s clocked as such, most consumers will assume the interior of the book, all of the manuscript, was also AI-generated, and not by, like, maybe it was only 30%. They’ll think the entire 100% manuscript was AI-generated.
Okay. So, let’s talk about reducing that suspicion, whether you’re using AI or not. How do you achieve that impartiality that the machine loves and consumers love? So, the first is to avoid AI for key marketing visuals. So, book covers, social media graphics, event flyers are the most common ones I see. If you are someone who publishes or writes genre fiction, make sure you’re not doing AI-generated character art either or backdrops and things like that.
For key marketing copy as well, you want to avoid this for book descriptions, outreach emails, socials, and newsletter copy. And there’s certainly a level of assistance that AI can provide. But I would argue that it’s more for the brainstorming stage, and everything actually set to pen needs…or to keyboard needs to be written by a human.
Some excellent use cases that we’ve seen at Serif, working across many genres and with many types of clients, is identifying keywords and as well as how AI speaks about specific topics. Interestingly, I’m not talking about having AI write you a paragraph on what is… I don’t know. I was just talking about this earlier.
So, epilepsy. What is epilepsy? What are the topics and keywords I need to use? Rather, when you’re in the planning stage for writing a book description or putting together a manuscript, ask AI, what is epilepsy? And see what kind of information they produce. What are the keywords they’re using? What resources are they pulling? Because all of that will tell you what is the priority content that the AI is prioritising. So, why is that resource being used and not this resource? All of that.
And it can come down to the wording you used in your query. And that can help you understand, okay, as an example, we’re working on a metadata audit for a client where their book is about the history of elder care facilities. And if we just search elder care, that seems like the primary keyword, but actually it’s going to send you to all the journals for current elders about to enter care. So, it’s not a good keyword fit, especially for AI, because it’ll only surface modern stuff. So, now, we need to start thinking about, okay, all of our keywords need to have a historical element. History of, for example.
Other use cases for AI, of course, organization, restructuring, adjustments for clarity. So, maybe you write something and you say, I need to turn this into an article or a blog post. Give me some recommendations on structure. And again, don’t use the output structure, but use it to adapt your editing process. Really important because more and more places are banning the use of AI for outlining. So, don’t use the actual outline, but take lessons from it for your first primary edit. Okay.
The next one, of course, is ideation. I find this incredibly useful for things like social media posts, shaking up the newsletter of your content a little bit. If you have an ongoing blog, ideation for like, okay, we have a book coming out that’s really important. Give us 10 article ideas for this particular topic. And then choose four. So, it can be very helpful, especially because often when we’re working at a fast pace with low resource or time, we start to move into rote.
So, every book, we do a graphic of a testimonial, and we do a graphic of the cover by itself. And the captions are always the same. That might feel boring to you, or it starts to feel pretty auto from your end. If it’s feeling like it’s kind of rote already for you, I guarantee your followers are already checked out. So, ideation can be a good way to inject your process with new ideas to break you out of slumps.
And finally, probably one of the most important ones actually is accessibility improvements. If you’re a team that has never had time to do alt text, now, you can. So, do it. There’s no excuse now to not have alt text. So, make sure that you’re using it in the ways that allow you to be more accessible to a wider audience. I’ll note here, of course, you always edit your outputs because it might be wrong. So, just make sure you’re not auto and not checking. Yeah. Okay.
So, what are some other ways we can improve discoverability in a technical way? The first and probably most important thing is you should take a look at your robots.txt in the back end of your website. Every website has one. It’s a little file that exists alongside your website design. And it’s just called robots.txt. It’s literally the instructions for robots that visit your website, whether they’re allowed to look at your content, which ones are allowed to look at your content. And it’s just really important to get it set up.
There are a number of resources online that will walk you through it step by step. But the key here, which I’ve added here, is that in particular with the major AI bots is you want to make sure that OAI search bot is not banned from your website because it’ll actually block you from chat, indexing, and surfacing, and sometimes from search.
So, there’s a number of bots. Google alone has 200 bots that crawl the internet. A lot of it is obfuscated and private. You don’t know how the systems work. But people much smarter than me have written much more useful guides on the topic, and can give you basically a copy-paste to put into your robots.txt. But that’s the very first thing. Can robots call your page? And if you don’t want them to, ban them.
Understanding the difference between being crawled, indexed, and surfaced. We’ve put descriptions for this in a blog post for BookNet Canada that gives a lot of information about the terminologies of various aspects of AI and discovery. But basically, crawled means that a user put in a query, and AI did its quick little scan of the internet. And your website seemed to be related to the query in some way. So, you may appear in that particular answer to that particular user, which is great. But that’s a one-off. Indexed means that the AI bot crawled and surfaced your answer. But for whatever reason, maybe the user was very satisfied with the answer or didn’t have follow-up questions. The bot decides to index your content, which means it adds it to its core data set. This can be different from training models. It’s more like it just indexes the actual information, and then later may surface it in future answers.
So, it’s a three-step process. Being crawled alone is fine. If you don’t want your information to be used in training data, there’s several bots that you can just block that won’t affect your ability to be included in answer engine bot traffic. So, get to know these three terms, and then set up your robots.txt as you wish.
The next important step is to improve the completeness of your individual product metadata. So, this is particularly on the back end. And again, look at the previous session on this because it really goes into depth there. But you can also update your product page infrastructure, including if you’re an author or you publish on specific topics with a blog, adding more ways to get referenced to your website.
So, going back to, for instance, the elder care book, it’s a complex topic, and it can be…it can require multiple keyword alterations. And you can improve your ability to be crawled by having, for instance, a blog post that talks about that history in a different way, targeting a specific keyword or a specific audience or specific questions so that maybe this didn’t work, but this did. And then you get indexed. So, it’s not just about product pages like Amazon, though very important. Your website, if it has more data, more robust answers for what an AI is looking for, you can bet that your page will appear above Amazon. If someone is looking for something like, I need a recommendation on resources for epilepsy, that is how you can be higher than Amazon.
There are, of course, other ways to be higher than Amazon, but they’re hard to beat, which is price and ratings. So, if you have a better price than Amazon, which you have to kind of play chicken because their bots are checking for different prices all the time, you can be placed higher than Amazon or higher than Indigo. A good way to do this actually is to have a special edition on your website that’s not like a traditional retail thing. But it’s less expensive than the other varieties because right now, AI does not differentiate between format. So, that’s a way to do it.
Secondary is ratings. I know most publishers don’t put ratings on their own website, and there’s an obvious reason for that. But if you do have a reliance on Amazon, on Indigo, on any other retail site, make sure when you’re doing outreach, when you’re getting your authors to activate their communities, it’s not just about talking about the book online, it’s posting reviews. Some communities are really good about that. I would say, in the lit-fic, nonfiction space, poetry space, people are not as good at just at least rating the book rather than writing a review. Every little bit helps. And rather than just being about now consumer trust, it’s also about AI trust. And so, higher ratings with lower price is the gold award for AI crawling you.
Okay. So, let’s get into some more tips for book discoverability. The most important thing here is that, basically, all search now is query-based. So, rather than someone just typing a couple of words in Google or in Claude or Gemini, they’re typing usually a full question, and people are getting more comfortable with writing very distinct prompts with extra details, kind of like narrowing the window of answer they’re looking for.
So, for instance, if I’m looking for a book, I’ll be like, I’m looking for a book for a seven-year-old with a reading level of, you know, whatever. It’s a very specific type of search, rather than just books for girls, which would have been Google pre-AI.
So, some of the things that we can do to help books in that kind of query-based discovery is thinking about, well, what are the types of queries people are putting into AI, and how can I add that first to metadata or internal architecture? So, on your own website, having tags for books that are like, you know, books for middle school, whatever. That’s almost like a filter search. So, some of those can include, again, for younger readers, likely gender. So, books for boys, books for girls. While I am certainly hesitant to move into binary for gender, that’s kind of how consumers still function. Like even with Facebook ads, you can only choose between two genders, and it has to assume the buyer knows what kind of gender they’re searching for.
Secondly, situational purchases. So, books about anxiety, books about epilepsy, books about the history of Ontario. All that kind of metadata is really important because, typically, when people are looking for books, they’re either looking for themselves or for somebody they want to buy a gift for.
Key topics, especially when controversial. So, a good example right now, unfortunately, is residential school denialism. So, thinking about how do you situate your book into topics, even if you are not in agreement with the majority answer. So, if I was doing metadata for a book about politics that was about conservatism versus liberalism, I would mention both. You don’t have to say which one it’s for.
Bought together focus. So, really important. Make sure your series and comps are really clean and that metadata back in. You want to make sure your series are all linked to each other across every retail channel. And it’s super clear, for instance, what number it is in a series can be one of the first things that gets lost in a series metadata.
And then, of course, prioritised keywords. It’s not enough to just list all your keywords. You need to make sure that they’re in the order of importance because certain retail channels have a line at which they cut.
So, for external, so this is more on your website with social media, your newsletter, book descriptions, all of that stuff. Make sure you’re also just naming in your description your sub-genre. So, if you are writing a poetry book or publishing a poetry book, what kind of poetry is it? Have you named it, or is it just poetry? Wild, right? Does it have a school of thought that it comes from? Is it reminiscent of an Oscar Wilde versus Murasaki Shikibu? Very different schools of poetry.
Secondary benefits of your book, especially for nonfiction. What does it help with in terms of classroom settings, therapy settings, secondary audiences that are using it as a tool or an instrument of education? If you’re writing in genre, think about your tropes. What are the types of characters? A good place or resource for this is to go to tvtropes.org to discover the archetypes that you have included in your book, whether that’s characters or plot devices.
Increasingly, especially in the genre space, readers want to kind of know, what are the tropes? Because they like to read a lot of the same thing when they’re going through a period of interest. It’s almost like a hyper fixation. The return of, if you liked X and Y, you’ll like this book. That’s a great thing to include in the description, especially if it is a true marriage of two types of books together.
Critical keywords shouldn’t just be in your metadata. They should be in your descriptions. And particularly, the most important keyword for your book needs to be in the first line, not necessarily the description. I call it a log line. It’s usually that bolded first sentence at the top of a description. Highly important. Okay.
And, of course, format. So, tell me about the product, especially if I can’t see it. Is it illustrated? Are the pages textured? Is it full colour? How many illustrations, etc.
Okay. This is largely just a summary of what we’ve talked about today. I’ve provided it in the slides just to have a helpful one thing to send to teams. The main thing I’ll mention here is the caveats.
So, for keyword identification, AI is really bad at generating a list of keywords. So, if you say, give me 50 keywords on a book with this description, you’re going to find a lot of junk in there. What’s better to do is ask the queries you’re hoping your book will be surfaced in so that you get the actual keywords AI has used to describe or provide an answer. So, that’s how I would recommend keyword identification, particularly when you’re first starting building your metadata.
Secondly is the key copy structure. So, again, going back to that querying what you think is going to be asked, you’re going to see the order of importance that that AI engine is presenting the material in. And that can be a good indicator of how you should write your book description or structure your nonfiction textbook.
Okay. But for those who wish to not be found or to hide specific content because you’re under IP attack, you can ban crawlers in your website in the same place, robots.txt, but that can affect your regular search. So, if that’s a place you would like to go in terms of scorched earth, you are absolutely able to do that. But that’s the place you would do it.
If you’re worried about artwork like interior spreads that you want to use to promote the book being scraped, and the illustrator’s IP and livelihood getting scraped, you can also place artwork, your team photos, and other illustrative elements through scrambler apps like Nightshade and Glaze. Good examples to the side here. Those cats look weird as heck.
Next, you can code your ebooks to reject crawlers outright. Again, please go to the first session. Tricia shared how to reject crawlers in your ebook content. And you can also consider additional barriers, including like server side blocking. So, blocking specific types of crawlers, IPs, etc. You can rate limit, which reduces the amount of times a crawler can come to your website in a certain period of time. And again, going back to protecting your popular IP, make sure you report imposter content. And the most important places to look for that are Amazon, YouTube, in particular, audiobook adaptations of books. It’s a weird loophole. And Google Books.
Okay. That’s everything. So, I’m excited to jump into questions.
Stephanie: Thanks, Ariel. That was great. My mind is teeming with all of this information. So, we are going to start the Q&A portion now. So, just a reminder for anyone who has questions, please submit that in the Q&A panel. And we’ll dive right in. So, if you don’t mind, I will feed the question to you, Ariel, and you can answer as you are able. Okay?
Ariel: Sure.
Stephanie: All right. So, first question would be, how can our audience today adapt book descriptions to make them easily discoverable in this AI age?
Ariel: It’s a great question. So, under the understanding that a lot of teams already put a lot of effort into metadata and don’t have a lot of time to spare, the number one thing you can do is really understand what the most important keyword is for a book, and make sure it’s in the first line of that book description field everywhere. This is so critical because, for instance, Amazon’s actual search bot, like when someone’s searching for books on whatever, it’s looking at the first 85 characters for free. You don’t have to do extra advertising or anything like that. It doesn’t necessarily have to be in the subtitle or the main title. But knowing what that primary keyword is that’s actually going to drive people to a relevant book, so important to have it in that first line.
Don’t make up your keywords. Make sure that there’s a data process involved in that. I’ve done presentations on this in the past, but to TLDR it, you would go to different locations to test the keyword. When you put that keyword into Amazon, are the right books showing up? Or is it a completely different field, like that elderly care example? You can go to Google Trends to see how this keyword has performed in search density over time. If you have a Google Ads account, you can use their Keyword Planner tool to explore versions of a keyword to see which one is the most important. And then you can use Search Console for your own website.
So, there’s a lot of different, mostly Google, tools to test whether a keyword is useful or not.
Stephanie: Interestingly, some of the tools that people are already used to using. So, that’s some great advice. Okay. Here’s another question. What is the single biggest lever in this crazy age of inundating with information that an author or a publisher or anyone else in the book supply chain can use to increase visibility for their books or themselves?
Ariel: Another really awesome question. So, other than putting that keyword in your first sentence, it’s a really hard thing for, I think, a lot of people to do. Book descriptions should feel like any other product description on the website or on the internet, which is to say, get rid of the purple prose. Make it understandable. If you’re using a lot of effervescent, kind of coded language, all that kind of stuff, it reads prettily to a human, of course, but a machine cannot understand it. And so, it’s like, okay, so, we’re talking about flowers and then we’re talking about this. It can get very confused, which is what encourages or increases the rate of hallucination and misinterpretation.
There’s many examples of it, but the funniest one to me is the algorithm, for over a year, believed that I was a 20-something hipster in need of beard care products. So, it can get quite wild. And I poison my AI search all the time because of the work I do. So, not surprised. But be clear, use real, simple language, not…don’t dumb down things. Depth is still available and encouraged, but use real language. Don’t obfuscate meaning.
Stephanie: Great. Clarity is key. I say that all the time. Okay. We’re running out of time, but I think I can squeeze in one or two more. You mentioned earlier in the presentation that users spend 53% more time on websites where the product that AI recommended is being sold. Now, is that true of publisher websites as well, or is this mostly Amazon?
Ariel: It’s hard to beat when it’s like, for instance, through that UCP portal where it’s like, give me a book on whatever. It’s going to send people to the highest trusted website, which is going to be Amazon. Where publishers can win is if someone’s looking for resource, or they’re looking for a series on, that kind of thing. Because your website can have way more metadata than Amazon ever shows. And that can be through articles that kind of expound on more of the interior of the book that maybe isn’t visible from any retail space. Or it can be just really positioning it as the resource on a topic. Feel free to beat your chest a bit and crank up the marketing language on your websites. But it really comes down to resource.
Now, I have seen publishers beat Amazon in a product location when their pricing is better. So, I’ve actually seen it and screenshotted it. It certainly exists. But I wouldn’t try to beat Amazon on the looking for a product space because Amazon or Google are going to win that argument. But if it’s about resource, if it’s about learning, all of that, they’ll get a lot more.
Stephanie: Okay. Unfortunately, I think that’s all the time we have for questions. If anyone has really burning questions that we didn’t get a chance to answer, send us an email [email protected], and we will try to put you in touch with Ariel to answer those questions.
Ariel: Absolutely.
Stephanie: So, thanks again for joining us, Ariel. Before we go, we’d just love it if everybody in the audience could provide feedback on this session. We’ll drop a link to the survey in the chat. So, if you could take a couple of minutes to fill it out, that would be fantastic.
We will be posting a recording of this session. And we’ll email you a link to it as soon as that’s available. So, keep an eye on our Tech Forum website as well, bnctechforum.ca, to look for upcoming events and recordings as well as to explore content from our previous sessions.
And then lastly, we just want to thank the Department of Canadian Heritage for their support through the Canada Book Fund. And thanks to you all for attending today.
Ariel: Thank you, everyone.