This breakdown is invaluable. Understanding which SEO practices carry over, which need refinement, and the entirely new AI-era work provides clear guidance for staying visible and trusted in the GenAI landscape.
I talk about the latest AI trends and insights. If you’re interested in how SEO and content strategies are evolving in the AI era, check out my Substack. You’ll find it very relevant.
The book is available everywhere Amazon sells worldwide (in English), Andrew. I’m seeing orders from many countries outside the US. Unfortunately, after speaking with many authors about their problems with offering pdf-based versions, that option (and kindle options) is not available currently. I apologize if this won’t work for you. :(
Hi Duane, I would appreciate it if you could reconsider offering a PDF / kindle e-book version. I am your subscriber from India. It is expensive to ship the book from Amazon US to India (pay more than the cost of the book). On top of that it is cumbersome to get it delivered because of stricter customs in the country.
I can completely appreciate the challenges you describe. The book has been released on all of Amazon’s local subsidiaries, globally. For the immediate future, I will not be releasing the kindle version of a pdf version, I’m sorry. Understandably this will make it more difficult for some people who want to read the book, but face challenges like you have described, and that is regretful. In the future, I may revisit these options.
Now that I’ve read it fully I’d say that I agree but a fundamental difference worth stressing more is that what gets cited in an answer machine is really not the same as what gets ranked in a ranking machine.
It has to be something that isn’t common or unchanging knowledge, otherwise what’s the point in citing it?
And so what kind of content is produced is ultimately more important than how it is packaged (although that’s certainly important too).
I’m going to have to respectfully disagree, Aviel. 😉 LLMs don’t care if something is unique or uncommon. What they fundamentally look for - what they are programmed for - is the alignment between the question and the answer. The tighter the alignment, the better. And part of that alignment comes down to your content, which may or may not be unique compared to others. But when all the factors of trust, recency, retrievability, alignment, etc are calculated, the “best” answer is what should be brought back. Citations don’t depend on uniqueness. In fact, why a platform includes or excludes the link is not shared by any platforms, so we have no current way of forcing that trigger.
I definitely see your point about alignment. But in my experience I don’t get citations for everything, even when alignment is presumably very good with some sources.
For example, in a recent conversation with Gemini about click tracking options (Supabase vs Vercel Analytics events) I got at first no sources, but very clear recommendations. I did get some later but only for tangentially relevant information.
Later I discussed creatine loading with Gemini and got a number of sources from presumably quality sites with very strong opinions/recommendation on the topic.
I ascribe the difference to what is well documented vs what requires sources (or “information gain”).
For Gemini to make a real recommendation it appears to want to back it up with sources if it doesn’t have complete confidence in it.
In other words, the point I was trying to make is that the content should focus on the things that are more likely to need citations rather than what is “common knowledge”. That is to say, they should focus on what is real information gain as seen from the model’s perspective.
This breakdown is invaluable. Understanding which SEO practices carry over, which need refinement, and the entirely new AI-era work provides clear guidance for staying visible and trusted in the GenAI landscape.
I talk about the latest AI trends and insights. If you’re interested in how SEO and content strategies are evolving in the AI era, check out my Substack. You’ll find it very relevant.
Hi Duane. Would love to buy the book but I'm not in the US. Is there a digital/Kindle version?
The book is available everywhere Amazon sells worldwide (in English), Andrew. I’m seeing orders from many countries outside the US. Unfortunately, after speaking with many authors about their problems with offering pdf-based versions, that option (and kindle options) is not available currently. I apologize if this won’t work for you. :(
Hi Duane, I would appreciate it if you could reconsider offering a PDF / kindle e-book version. I am your subscriber from India. It is expensive to ship the book from Amazon US to India (pay more than the cost of the book). On top of that it is cumbersome to get it delivered because of stricter customs in the country.
I can completely appreciate the challenges you describe. The book has been released on all of Amazon’s local subsidiaries, globally. For the immediate future, I will not be releasing the kindle version of a pdf version, I’m sorry. Understandably this will make it more difficult for some people who want to read the book, but face challenges like you have described, and that is regretful. In the future, I may revisit these options.
Thank you for your reply.
Yes, I searched on Amazon India for your latest book and it is not available yet. I look forward to its release.
I assume you mean on Dec 1, 2025 and not 2026?
Absolutely! Awesome catch, thank you! Or…was I just trying to give you LOTS of warning? 😁😆
Now that I’ve read it fully I’d say that I agree but a fundamental difference worth stressing more is that what gets cited in an answer machine is really not the same as what gets ranked in a ranking machine.
It has to be something that isn’t common or unchanging knowledge, otherwise what’s the point in citing it?
And so what kind of content is produced is ultimately more important than how it is packaged (although that’s certainly important too).
I’m going to have to respectfully disagree, Aviel. 😉 LLMs don’t care if something is unique or uncommon. What they fundamentally look for - what they are programmed for - is the alignment between the question and the answer. The tighter the alignment, the better. And part of that alignment comes down to your content, which may or may not be unique compared to others. But when all the factors of trust, recency, retrievability, alignment, etc are calculated, the “best” answer is what should be brought back. Citations don’t depend on uniqueness. In fact, why a platform includes or excludes the link is not shared by any platforms, so we have no current way of forcing that trigger.
I definitely see your point about alignment. But in my experience I don’t get citations for everything, even when alignment is presumably very good with some sources.
For example, in a recent conversation with Gemini about click tracking options (Supabase vs Vercel Analytics events) I got at first no sources, but very clear recommendations. I did get some later but only for tangentially relevant information.
Later I discussed creatine loading with Gemini and got a number of sources from presumably quality sites with very strong opinions/recommendation on the topic.
I ascribe the difference to what is well documented vs what requires sources (or “information gain”).
For Gemini to make a real recommendation it appears to want to back it up with sources if it doesn’t have complete confidence in it.
In other words, the point I was trying to make is that the content should focus on the things that are more likely to need citations rather than what is “common knowledge”. That is to say, they should focus on what is real information gain as seen from the model’s perspective.
I can see your point more clearly now, Aviel. 🤝