---

About Simon : LinkedIn

FULL TRANSCRIPT BELOW

Shamanth: I’m excited to welcome Simon Thillay to Intelligent Artifice. Simon, welcome to the show.

Simon: Thank you for having me.

Shamanth: I’m excited to have you, Simon. And I also want to thank you for your patience because we actually recorded without hitting the record button prior to this and realized a little bit too late that we weren’t actually recording.

Shamanth: So thank you for jumping on again. I’m excited to have you because you have a front row seat into how AIs are changing organic visibility, and that is incredible just because things are changing so fast, so quickly. So talk to me about what are some of the most important elements that are contributing to app store visibility with AI today?

Simon: Sure. I think the first thing to understand is that AI visibility for apps is basically changing surfaces, as much as it’s changing who gets surfaced. Just last week, Google officially started speaking about a new way they are presenting search results in the Play Store. It’s called Ask Play.

Simon: And essentially what we’re seeing is that now Google, for instance, is just making sure that you see certain results are generated with AI, or results are resorted by AI. It’s not changing the fact that you’re heading to the Play Store to search. It’s not changing the fact that people will download your app through the Play Store.

Simon: But now there is this question of, is a store, the Play Store in this case, only showing a list of results that look essentially like a well-designed list of links where you would see the app icon and title, or do you already start seeing such results are more adjusted to semantic fields? How are they displayed?

Simon: So I think that’s the first thing to understand: when does AI intervene in the discovery process. We’ve also seen that when it comes to the app stores, there’s no obvious AI feature in the search results of the app store just yet.

Simon: But we do see that there are, at least for some apps, traffic coming directly from ChatGPT, from various LLM apps, that you can find by searching through your App Store Connect console. And that suggests that people are also maybe pre-searching, using AI, which apps they might want to take an interest in, and then heading to the store to download. So I think these are the changes on the consumer side.

Shamanth: Right.

Simon: On the business side, what’s important to understand is when AI is used and what foundations power the AI results when your app is going to be mentioned not just by an old school search engine, but now by an AI-powered engine. And so for these, I think it’s important to not panic.

Simon: Kind of good news, AppTweak, the company I work for, has done already some studying, and we’ve seen that the most cited domain in ChatGPT answers for app-related queries is the App Store. So that means your ASO is not going to waste.

Simon: It’s now powering AI visibility. It’s not the only source, but it is part of it. And maybe something that’s important for people who, like me, have been doing ASO for a long time is to hear that maybe the biggest short-term change is that, with AI definitely liking to have more text to learn from and to reuse when generating answers, the long description of the App Store, which famously was not having much value in ASO in the past, now has the option of powering how AI learns about your app.

Simon: Potentially getting to improve your discoverability by making sure you rewrite those long descriptions with AI in mind, which has a few tips and tricks to uncover.

Shamanth: Impressive. Yeah, you’re right that in older ASO, the long description almost didn’t matter. And if I understand correctly, because it has so much text and LLMs are reviewing a lot of the text, it’s good to have the long description be optimized for AI search. And talk to me about what it means to optimize it for AI search, because you have to write it for AI, what are some of the best practices or elements of writing for AI?

Simon: So I think the marketer still wants to say, “Don’t write for AI. Write for humans using AI.”

Shamanth: Yeah.

Simon: But essentially what that means is two things. One is writing for humans who are going to be using AI means writing knowing who you’re talking to. Something that we see is becoming, is coming back as a trend for marketing, is don’t just own the channel.

Simon: Also make sure there’s a general understanding in every different marketing effort you do, that you know who your audience is, you know the context in which they’re interacting with your product, and that essentially you can clearly write for different user personas.

Simon: We don’t have to write for a single person anymore. It’s not entirely new for ASO because Apple and Google have already given us custom product pages and custom store listings for the past few years. But in general, it’s still this idea that what you write should be about who you write it for and what they care about.

Simon: And then the how you write it is the part where you write for AI. Here it’s all about knowing that writing with a clear structure is something that is having true relevance. I will say there are general sort of SEO tips for structure that I think everyone probably heard about already. Things like bullet points, structured paragraphs, etc.

Simon: These general text formatting good practices help AI read even better than a human would feel like it’s easier to read for them. The maybe slightly less known tip is to write while being aware of entities.

Simon: It’s basically this idea that when AI reads a text, it’s not an old search engine which is just going to do character matching to say, “This is a keyword and this is a keyword, and I’m going to…”

Simon: And so when someone types a keyword in their search query, I’m going to try to match that keyword with a corpus of text that uses it.” With AI, AI actually tries to identify entities as the middle step, where if someone is searching about something, it asks: what is the important keyword? What is the entity that they’re actually interested in?

Simon: Which entities relate to that one, and which are the sources that mention the core entity as well as the adjacent entities in what they write, so I can understand I am actually going to provide relevant answers? And so writing for AI is writing like a structured text, but it’s also writing a text that clearly helps identify which are the entities in the text.

Simon: A text that also makes sure it’s clear that an entity is an entity. And what I mean by that is acronyms, abbreviations. Something I’ve heard a lot recently is don’t speak about the same feature using different names. Make sure that everywhere you use the same name, because that’s how that name is eventually identified as an entity, etc.

Simon: So these are, I’d say, the two tips and sub-tips. Write for humans, meaning write with specific personas in mind, write with context in mind. And then write for the AI in use by the humans by making sure you write structured text and you write where AI can clearly identify the entities you are talking about.

Shamanth: Right. In many ways, what you’re saying is just keep things as simple as possible, keep things consistent. If you call it ASO once, don’t call it app store optimization. Maybe that’s not the best example, but keep things consistent just so when AI sees something, it associates it with your product rather than being confused, if I understand that correctly.

Simon: Yes.

Shamanth: Right? Yeah. And how would you think about content-based AEO? Does that matter at all for apps and products? And so background being, you know, with SEO, it was never clear about whether, you know, if you wrote a bunch of blog posts, was it actually helping or not. With AEO, I don’t know. You tell me.

Simon: Yes, it does matter. In fact, I think that outside of specific verticals where web presence has never been a priority, AEO today feels to me as a topic that’s being co-opted by SEO much faster than by ASO. And the reality of AEO is it’s both, it’s more.

Simon: AEO is about the fact that LLMs have access to the open web. The open web includes your own website. It also includes community platforms, authority platforms. So you can think easily of Reddit, Wikipedia, and at some point YouTube.

Simon: YouTube, the challenge is there’s much fewer texts, and until LLMs make a systematic effort to actually review the video content, I wouldn’t prioritize it as much. But still, it’s this idea that AEO is about the fact that the stores are also on the web.

Simon: And so AI is learning about apps, not thinking about am I directing people to a website, to a community platform, or to a store. They’re just thinking about: identify this is an app, and so whenever someone is going to be asking me about what’s a good app for a certain purpose, I should make sure I give a relevant answer based on the purpose, but I also include the fact that I should be referencing an app.

Simon: The ASO component is what I was mentioning before. Usually, when ChatGPT or other LLMs are requested to provide an answer that includes an app, they prefer to go towards the App Store when they know there is a good answer in there. But then if they don’t know that there are apps about a certain topic, they’re still going to look for the rest of the open web.

Simon: And so content efforts, whether they’re traditional SEO with building and writing a blog for your company, or whether we’re talking where SEO meets community management and questions like how can we have a say as to what Wikipedia is saying about our company, about our product, can we be present on various platforms, I said Reddit, but we could talk about Trustpilot, we could talk about so many different places.

Simon: All of that plays a role in AEO, and I think the best way to get started consistently is not to make AEO a problem of one function in the company, but rather just to say, “We want everyone to have a stake in AEO,” because AI is starting to appear in Google AI Mode, in the Gemini app, in the Play Store, just to cite the Google ecosystem.

Simon: And then of course other LLMs are being used through other ecosystems. So it’s a general marketing issue. It’s not a one domain specific issue.

Shamanth: Yeah. Yeah. Yeah, I think that definitely makes sense. So what you’re saying is the first priority for AI is the App Store, but if there isn’t so much of an app’s intent, it’s still going to look at the broader content ecosystem, right?

Shamanth: Yeah. And something I also found fascinating about something you said earlier in preparation for this call is that you’ve done a lot of study of AI products and how they rank in ChatGPT, and you had some very interesting learnings, and I would love for you to share those.

Simon: Yeah. So for listeners, I think one thing that’s important to understand is how to audit your app’s AI visibility. And here the first thing to understand is that AI visibility should not be measured at the prompt level, but rather at the intent level.

Simon: Because essentially if you make the test, ask ten people to research a certain topic using whichever LLM of their choice, and ask them to provide you with not just the results they got, but also the prompts they used, you’ll see that even when the instructions seem quite specific, people don’t write the same way. So a single prompt is not going to cut it.

Simon: You shouldn’t measure your visibility based just on how AI talks about you and your competitors one way of writing the question. Instead, what you need to look at is, if I write the question in multiple different ways, but all about the same intent, how often do I get mentioned? And how does that compare to others?

Simon: And then also, how many intents do I cover versus competitors versus potential? And this is where you and I talked about one of the early experiments I made using AppTweak, where I always like to look at how ChatGPT is performing on ASO and AEO.

Simon: And what I found was that ChatGPT has tremendous presence in terms of breadth of intent. But when you look at a certain intent, it can be outperformed by comparatively much smaller companies. And my favorite example is Grammarly, because Grammarly has done a strong effort already in terms of how they position their app.

Simon: Not just as a practical feature-led app to fix your grammar, they’ve also already worked on establishing context. And as a result, if you’re asking ChatGPT for app recommendations in terms of which apps can help you write business emails, Grammarly might be cited.

Simon: And in fact, what we found is that Grammarly is mentioned more often than ChatGPT for that specific purpose.

Shamanth: Yeah, and that is so fascinating because even though ChatGPT owns the platform, Grammarly has put in more of an effort in positioning themselves for niche audiences that ChatGPT just hasn’t. And therefore Grammarly is positioned as the best for business audiences, the best for writing emails, for proofreading, even though objectively ChatGPT might give you good proofreading.

Shamanth: But the way Grammarly has positioned itself has made such a huge difference. That is so interesting and so fascinating. Yeah. And so the other thing you talked about was, if I understand correctly, you said the way to measure the impact of AEO is to measure the intent. Did I understand that correctly?

Shamanth: Can you talk to me about that?

Simon: Yeah. The idea is you need to measure your visibility. You can measure it in terms of, let’s say, rank, like how high in the answer is your product being mentioned.

Shamanth: Yeah.

Simon: Before you look at rank, the first thing would be to look at the coverage of the intent. If you write a certain prompt a certain way and are mentioned as the top app in a list of five, let’s say, you might feel very happy with yourself.

Simon: But it turns out that if we wrote the prompt differently, not only would your app not be mentioned as the first answer anymore, it would not be mentioned at all. For the nine other prompts, let’s say, for a total of ten prompts analyzed when looking at a certain intent, this would be more likely to be worrisome.

Simon: Because you can’t predict how people are specifically searching, typing into ChatGPT. You can somewhat anticipate what topics they’re asking about, intents. And this is what AppTweak has built because AppTweak has built essentially its own system leveraging the ten plus years of ASO data to identify, per app category, per app vertical, what are the likely intents that people are searching for.

Simon: And then for these intents, what are multiple prompts that we can query regularly to start measuring visibility. And then we start measuring on three different dimensions. Coverage being the first and the most important early on. Rank, once you start having a solid coverage on a certain intent. And last but not least, sentiment.

Simon: Because AI does not create links, AI creates text, so you might get mentioned in a more positive or negative way in certain answers. And so if you’re the third answer to a prompt that essentially says competitor number one is the people’s favorite in the category, competitor number two is considered a solid competitor, and your app is also a valued alternative, though often criticized for price, then you might start expecting fewer conversions.

Simon: And so that’s why we also try to track sentiment because we understand that when looking to measure AI visibility, what we want to understand is: is an app mentioned, how often, how high, and how positively or not?

Shamanth: Interesting. Yeah. That is so fascinating. Also, just because it can appear to be a bit of a black box, but I think this is a very clean way to really make sense of AI visibility. Simon, this has been incredible. Also because like I said, this is all very much a black box, this is all very much emerging.

Shamanth: So thank you for sharing your perspectives, and thank you for sharing your insights. I’m sure this is going to be incredibly valuable for folks. This is perhaps a good time for us to start to wrap up, but before we do that, can you tell folks how they can find out more about you and everything you do?

Simon: Sure. For the company and the product I’ve alluded to, apptweak.com is our website, and it’s well referenced. In fact, usually AI does a good job at referencing it as well. So you can learn more there. And for people who also just want to stay in touch for the various findings I try to share regarding ASO and AI visibility, I think LinkedIn is always a good place.

Shamanth: Wonderful. And we’ll link to all of that. So now this is a great place for us to wrap up. Thank you so much for your time.

Simon: Thank you again for having me.