Mobile & F2P | Marek Pasieczny | ASO Meets AI: The LLM Advantage | Digital Dragons 2026

First of all, thanks for having me here and thanks you all for joining. And we have today very interesting topic to to cover quite fresh to be honest. Uh if we take into consideration last changes on the Google Play Store and the App Store. But before we start, I have three question to you. The first one, how many of you joined the session seeing AI or LLM terms in the title just because of that? Okay. Yeah, that's typical. Thank you. Uh do we have here any ASO specialists, experts, team leads, practitioners? One. Yeah. Hi. Okay. And do we have any app developers,

app owners, people who actually develop an app? Nice. Nice. Okay. So, uh I will try to cover with this session all the subject that could be beneficial for you that you can make advantage of it. So, yeah, let's let's dive in. Uh shortly about me. Eight years in mobile gaming and ASO. I did a little bit of user acquisition with the budgets up to quarter of a million monthly. So this is something but yeah and now uh I'm currently focusing on AIdriven solution but not in the case of agenting AI more like a solution connected to visibility in AI uh and what's changes on the store recently. Yeah. uh kind of agenda but uh this session is about uh how LLMs are reshaping app discovery. It's about AI discoverability, visibility of our apps. Uh and uh the

key information that the user journey starts now not only on the in the app store but also outside of the app store. That's very crucial. This session it's not about agenting AI or any technical deep dives but I can recommend you these two experts that I worked with in the past. uh Agneska now currently at Zen she's brilliant switching from graphic design now from to building AI tools even coding cloud computing she's brilliant and Jakob Marovich from future play games a brilliant mind huge external ASO background also building agenting tools so if you want to do something about agentic AI you can connect uh them directly Yeah. So before we dive deep into

ASO connection to LLMs, I owe you an introduction to LLM. So let's start with uh minimum understanding of LLM and why LLMs matter for discovery. Now well LLMs um didn't came out of nowhere I would say this is not any you know like magic wand or wizardry or sorcery something like this many people may think about this like this uh it's a part of AI which is a border term in this in this case and it includes the terms like uh machine learning well no one deep learning and NLP natural language process. So what has LLMs from deep learning and natural language process? NLP can understand context, user intent and create answers. And if we add here a

layer of deep learning like billions of sentences uh articles across the web researches we are getting large language models that can understand intent understood needs and generate humanlike answers and get you can get into conversional convers conversional mode with LLMs. So basically speaking this a AI assistant that can understand your intent compare alternatives recommend some content create short list for you on daily basis these are LLMs chat GPT Gemini clot AI Google mode this is it so yeah something we that we call AIS it's currently LLM or AI assistance and looking at this task

We can conclude that it makes a very good starting point for app discovery because doing comparing alternatives recommended content recommended u some different apps it it might be good for app app discovery your app discovery okay so fact one uh LLMs are based on training data this is very important important information. Uh because you know maybe you overcome this issue with that you make some requests creating intent or creating prompt asking for some research like find me the the most fresh job offers for me and uh LM returned back with some outdated data like uh the job offers that doesn't exist anymore. Uh yeah and but in the newest models of Gemini chat GPT LLM can go into the web

and search for the fresh information but you know only if you ask to them to do so. So the good thing about it is that um yeah it can enter the internet of course but uh but uh sorry for that uh training data is crucial and uh LLMs are updated from time to time uh with billions of information across the internet. Um, that's why it's so hard for LM to get the the most fresh data. And how LM's changed the weight of searching. Um the answer is simple because users start to ask increasingly about something about you know their needs like LM please find me something

please compare some alternatives something like this instead of just blindly browsing or wasting their times like you know like we did for a long long time and these are two approaches I would say the old and new one, but the old one still exist. If you look at the prompts, this one and this one, they basically looks the same. Yeah. But the first one is uh it's a short query on average three to four words. It's a exact match uh limited by context and usually the results you will get from the Google search. You still need to dive deep into it source by source, website by website looking for information that is valuable for you. So it's a lot of time. On the second note, we have prompt

searches. So basically like I said um as I said it looks quite similar. Yeah. But on average prompt are longer 11 to 13 words and they allowed us to get more into conversational mode uh to get to deliver more context to LLM like JP or Gemini. uh because uh you know the more we get into this conversation the will start ask about more about more about more our needs our our issues we have yes and this make uh this conversation more intentful and the results that we'll get much are much more richer and we can we can get also some mentions andations like according to our brand so we can say that we are moving now slowly But it will speed up for sure from

keywords to prompt from just simple correlation to conversation. Yeah, like people to people like human to human. Uh yeah, another fact intent matters more than ever. This is a simple classification made by the guys from the tool answer the public. I will talk more about this tool later on but they try to classify uh the prompt into types like user intents to be honest. Yes. So we have four types of intents. 90% of them are usually informational or commercial like we are looking for some information in chat GPT. Looking for some comparison like like this example I put here. What are the best relaxing puzzle games? It's informational the comparison between two types of games for example. Um yeah also navigational how to find something. uh

and only 4% or on transactional not only focus on money but but also for example how to get a free starter parking some specific games. So dealing with premium apps or games. So basically the apps where we that are free to play but we want player to join and to monetize. Of course we should focus on transactional part. If you want to boost our revenue and our income coming from LLM mentions and stations this one and another fact about brand is strong. Your app brand need to be strong. And this is uh new research made by the guys from HRES a big big uh tool connected to search engine optimization but not only now to AI uh analyzes as

well and they made a research based on 75,000 brands what are the most influential factors for LLMs to be mentioned by them to the users. So yeah, brand in this uh example is the king. Brand web mentions correlate much more stronger than for example ad traffic or ad cost. Uh branded uncor so if you put a backlink in the article with the name of your brand it uh matters a lot and brand search volume. So how many times your brand uh is mentioned across the web the researched articles. Yes. So basically you can spend a lot of money for your mobile app or mobile game but still brand matter much more and you can get up to 10 more AI mentions by working only on your brand strange

and last but not least in this more of a knowledge part sentiment when we can say that Bran is the king or Indian is the king Then sentiment is the queen because sentiment is information about user trust. And if you go dive deep into any tools like hrefs like recommended by me Samrush or brand 24 the Polish one actually acquired by Samrush 100%. You can get to the information about sentiment. uh sentiment is uh one of the most influential factors for LLMs for AI assistant because it build trust and you can see there in this tools the type of for example prompts for your I don't know website mobile app etc and the sentiment and the higher is sentiment

for some uh specific prompts the higher is the possibility that that your mobile app will be mentioned there. So yeah, take care about the brand, take care about sentiment. It also applies to for example reviews on the store or reviews across the web. Yeah. So we are moving to ASO now. But still on more uh theoretical part, knowledge part than on use cases. We will move to use cases later on. Yeah, AB discovery begins outside of the store. still happens in the store but begins outside of of the store when you are dealing with LLMs uh and you asking LLMs for recommendation. Yeah. And uh still the like the comparation between like old way and new way that we are all familiar with the views coming from

uh the first one app store search the second one from Google play search. So basically we can type there some exact M keyword looking for the game and on the app store uh level we will get uh mostly the the view from Apple search ads the first first placement and second placement somewhere in between lies the organic result but uh yeah mainly due to Apple search as placement this results are polluted are polluted uh because you know by paying uh for being here it can be misleading for you for your intent on the Google Play site uh at the very beginning we are getting three sponsor ads much more tailored to to us to the intent of course but still these are not organic results

uh and then uh the organic results come yeah And at the second part we see uh the prompt that I uh created uh in uh perplex city to find uh mobile RPG games with some specific features. And basically it returns the information about one game called Albin online does that on Google Play is basically on sixth position considering the ads and it's not visible on app store level. So the ranking differs so much in this case. Also the alternatives for the recommended album online. Uh take a look also at the at the sources uh that are seated regarding uh this specific game like it's a product page of Alban online. So probably they do a lot of work to align with uh like LLM's requirements to be to be here and also a

direct link to Apple uh Apple store to the app store. So basically it can shorten the way coming from the LLM to the app store to convert the users. Yes, that's why I mentioned that app discovery begins outside of the store. Uh also Epic Games Steam Power. Yeah, big sources definitely. So that's the new way versus the old way still existing way. Yeah. And following up to this, the old way is more like a keyword first exact match and starts in the store directly. Yeah. user compared the results uh wastes wastes his time uh while browsing while sh searching and ranking was the goal mainly regarding the keyword always to be on top to generate installs to be on top in top three category for example. But now it changes. It changes to intent

to meaning going more into convers conversational mode. Uh yeah, this discovery starts here, not here. Yeah, LLM or AI assistant creates a short list for you. You can choose from this. So this is the ranking. Uh this is actually the the new ranking. this and this instead of this. So we can say that LMS are not replacing the app stores of course because the app stores are the main source of uh users. It closes the loop with conversion to install but are reshaping how users present their intent and how they are searching. uh and the visualization how does it work in practice let uh let's take into

consideration the problem that we used before best mobile RPG game uh yeah so LLM retrieving this information yeah and now what's happening it goes to uh the researching mode yes across the web the whole web on on this pre-training data uh LLM is going through for example Wikipedia, websites, blogs, comparison uh pages also some social media very important ones especially do these two and based on that LLM synthesize uh recommendation. So the best RPG mobile games to play in 2006 are and then is a user decision to go dive deep into it or maybe to continue to the store directly uh using the source. Yes. And here's the magic happens that we can basically convert users coming from LLM inside of the

store. So it begins outside of of the store coming to the store. But of course it's not that simple every time but it happens. uh what matter the most in this case uh I would say everything starts with an intent without intent LLMs doesn't start to work uh but knowing the intent the user needs what lies behind some issues on information that users want to want to get uh that that starts the process and the process it's like this that Uh it's about the clear app positioning very strong I would say because LM must understand what your app specifically does. Yes. Then it looks for some uh sentiment for some social proof in community mentions in reviews. Yeah. Checking if sentiments

is either positive or negative. uh also strong correlation between LLMs and those factors uh app store listing. So the text that you have on your regarding your app for example in Google play might be also beneficial. You can uh include some information about your app there. So it can it can be a supporting signal in this case. Visual assets not so much. We are talking about icons about screenshots you know. Uh so visual elements uh doesn't play that much role for LLMs but of course if you want to create your app store listing regarding some specific intent this graphic elements should be aligned with it. Yeah I will talk more about it later on. Store ranking what I mentioned and store keywords not so much. It's quite limited

uh it's quite limited factor for LLMs just right now. Okay. But it's not happening also outside the store because AI enters the store. And I would say that in one of the freshest updates you might miss it of course because it happens like the two three weeks ago or even one week ago like uh the end of April 2nd 22nd of April app store changed it algorithm espec especially for the US to be much more aligned in user intent on the app store. So we can expect more intent or prompt like searches than keyword searches for example. On the other hand, Google Play at the end of uh March introduced some core update not to the whole Google ecosystem not only for Google search and

search engine optimization but also what's happening on the store directly. So yeah, I encourage you if you deal with apps to go into Google Play Console and check your uh Google Play Explorer, Google Play search metrics looking at the volume of organics at the convention rate because something might happen between March 27 to April 8 and and [snorts] uh and continue from this date and also since the April 15. So definitely check it. Yeah. um how uh AI enters the store. So both stores were focused currently uh basically on 2025th and beginning on 2026 on their well planned features to deliver. One of the feature on Google Play was Google Play games level up introducing if you using uh Android device you can uh you can now use uh UTAP especially dedicated for you

uh yeah and this UTP change a lot because it's all about the user engagement uh showing your streaks in some specific games so showing some events that happening inside the games you play uh but uh people find out very easily and very shortly that a lot of AI solution were introduced and embedded to this Google Play games and UTAP espec especially because you know you might you might seen the recommended for UTAP on Google. Now this recommendation should suit you even more better than before because yeah uh this all information that includes your engagement and activity in in the game are processed by AIS in case of its recommendation. One of the feature that was introduced was the AI assistant called sidekick. It's a game overlay. If you play the game, it can help you to overcome some, you know, hard levels or

something like this, but it's not connected to L&M itself. Uh what changed also uh Google added ask play something that we know from Google search that we can typically ask about the ask Gemini about uh about the game uh directly on the storefront. and a very fresh change mid mid April um AI generated review summaries so users don't need to go into every single summary uh to check it to check for some common issues for game like game features something like this because before user enter the store listing or even if user has entered the store listing uh it happens that this kind of the liar of summarize uh review uh yeah it's h it happens yeah so

I would say like this kind of changes can be highly conversiondriven yes because users usually the ones coming from organic traffic are using for are seeking for social proof and this kind of social proof summarized social proof can be very and highly conversiondriven. Yeah. The most fun part use cases use cases in the world of AO ao and LMS. But before you start, you need to know users intent, the promps they are using. Yes, because if you don't don't know the prompts, you can't optimize for AI assistants like chat, GB, Gemini, cloud, etc. And I have two examples of the tools that come to the rescue in this case. The one is the AI visibility tool in app

tweak um introduced last month. Uh from what I heard it's quite expensive. They didn't mention what's what's the cost and what's the pricing. uh but what they did appweek has uh app DNA for all all the apps that that they have in in the tool. So then they can basically say this app is in this category but we can define a subcategory if this app like this and this and this app has this and this and this features. Yes. And based on that they created a groups of prompts and under every group you can uh check uh the prompts and you have here the very two important metrics intent visibility and sentiment like I said before intent is the king sentiment is the queen. So these are most one of the most influential factors for LLMs.

Um by saying that I mean um if your app has a strong brand with a very positive sentiment the probability that you will be mentioned for some specific front is higher. This this one is regards to the some apps connected to meditation something like this calm headsp space. Uh the second one is mentioned before answer the public that make a classification of intent. It's partly free so you can go there and check the prompts for some keyword for some websites. Typically this is an example of Clash of Clans. So it shows what types of user what type of prompts users use to check Clash of Clans. So basically these are user intents. These are users I don't know needs and issues

to be solved. Uh what's more they are showing the intent like I said before informational commercial navigational and transactional and also they have the ability to generate content right away. So take some actions based on that and what actions I will talk more about it later on. Yeah. And the core build a strong web presence. This is this is actually the key. If you know the prompts, you can start to build a strong web presence. Uh what way? Like one of the way is using the sources you have. If you have a landing page or product page for your app, not only the you know the store listing on Google Play on or the app store, you can uh you can generate content that answer the user's needs and user intent. You can generate uh some specifically created uh frequently asked questions answering user intent and

prompts there. Based on that LLM can enter your store, enter your product page or landing page, scrape this data and mention your app uh when user are asking you know LMS for something. Yes. Uh another example comparation page. So if you have a blog, you can create an article comparing your app versus another or buy an article or on any external source or use some review accelerators a captera G2 something like this where two kind of apps are uh are compared. Yes. So these are very valuable sources for LLM domain with high authority and also practical guides again uh showing how your app works in practice or create a support page uh answering users users intent. So yeah, target the longtain prompts in this all solutions.

And following up, this is an example of super supercell uh and what uh they actually do regarding their games. They created a support page for this is example of Clash of Clans, but they have it for every specific games in their portfolio. And I mentioned before you know the intent types informational, commercial, navigational, transactional what they actually do they divided the support into blocks. So you can see the connotation like information about account it's a more like aformational content. game play it's a navigational content again troubleshooting it's a both commercial andformational and purchases this four magic percent transactional content so basically based on that classification they created a support page that is easily scrapable for AI assistant to find information and

deliver this information to the users And another thing building authority. So not only the brand power, brand strength, sentiment, but also authority. It's all combined together. And the authority of your app is coming from being discussed across the web and being referenced by AI assistant by LLMs. Another example of Clash of Clans here uh talking about like that the Clash of Clans animated series will be produced by Netflix but uh most importantly Reddit and LinkedIn Reddit is the one of the most mentioned and seated source uh in LLMs uh in the U in United States. uh it's come very natural because uh users of Reddit uh talk at Reddit about the game itself.

They share their needs, solutions, their intents. So also the Reddit structure make it very u accessible for LMS to scrape. So it's a good source for JPT Gemini cloud to find there any information regarding specific apps based on user intent. And another example here from LinkedIn. Maybe you have heard about it or not, but LinkedIn quietly changed its algorithm recently. It can now uh assess your profile sentiment, what kind of post you share, how user interact with this post and based on that it can scrape information also to answer users needs and intent uh inside the tools. This is a nice uh nice example of United State users who shared some information uh regarding economics versus real life economics versus economics in Clash of Clans. So this is a basic post of the

student that get uh 65 uh reactions and five comments. And this one uh the of the source that was referenced to me by uh by Chad GPD. So I would say that I encourage you to go into LinkedIn to post some information about your app or game uh to make them valuable for the users. Don't make it from your uh company account or app account. just from make it from your personal account and see and see what's happening. Uh before you do this uh be careful to use too much AI for creating that kind of post because you know it's quite recognizable not only by by uh by people but also by by LLM itself. Um yeah and example of how you can re rewrite your

description on the store. This is an example um coming from the app Surf SharkVPN. They changed the long description recently and it looks uh sort of like a very like a frequently asked questions but some specific user prompts and intents are targeted here. So what they are reaching for this is uh this is the different example but what they are reaching for they they want to be seated by the uh LLM as a sources because this opens a gate to the store itself. It opens a gate to for users to enter the store into and to convert. it doesn't happen so much that LLMs uh cheated the the sources directly to uh to the app stores but in this case in case of the the app called Kataviki they did a very good job in describing or matching uh

their description with user intent to get the sources uh once again review and sentiment. Uh the one that I mentioned AI generated reviews. So the reviews are not only for the users users just right now but the reviews are for the uh LLMs as well. And inside the store, App Store, Google Play Store, there's a huge bunch of reviews that uh LLMs can make advantage of, can get into it, read the sentiment, read the user needs there. But also outside of the store, I heard that Trust Pilot is a great source of reviews and uh one of the most influential review sites for LLMs a lot. Uh so it's free to use as far as I know for responding to user reviews

and uh uh and building sentiment but also some external tools like Captura, like Pocket Gamer, Mini Review, the games not not that are not only the review accelerators but also a comparison page. So they deliver more information about sentiment band but also if users would like to compare one up to another. Yeah. Um and I mentioned that graphic elements doesn't play a much role for LLMs right now like screenshots or icon but still what you can do is to target your graphic assets to specific prompts. Just take a look at the example. what's the best app for tracking my runs with heart rate data and simple progress inside. So we can usually scrape the intent based on that and create a brief and create a screenshots based on that. These screenshots were generated by Gemini uh in less than 3 minutes based on uh this

prompts and information that I delivered to it. And I was very pleased with the results actually because uh these three screenshots uh respond to this prompts in a very good way to be honest. So this way you can influ influence your conversion rate to install. Following up on this, uh different prompts should lead to different pages. I know I don't know how many of you are familiar with the case of CPP or CSL uh custom product pages on the app store and custom store listing on the Google Play Store. Both are alternatives page pages that you can create for your app. So basically you can have a default app but if you want to create uh some store listing for specific uh specific features of your app you can do so. Yes, this is an example of uh of prompts and

what kind of for example TPP and CSL you can create uh based on features. So my recommendation is leave your default page on the app store and Google play store as a brand baseline to all the users that are searching for your brand are entering store uh looking for your brand but create custom product pages and custom store listings for specific features for specific intents for specific prompts to answer the user's needs. Of course, there's going to be not so much traffic coming from them, but you will be pleased looking at conversion rate to install. Yeah. Last but not least, how to measure it? It's not going to be easy. I would say uh this all traffic coming from LLM can be measured by direct or indirect signals.

Um talking about indirect signals for example you can track brand keyword uplift uh Google play console delivers such a data you can check it on search term level uh for the brand uh brand searches misspellings for your brand. So if you will start to optimize for LLMs and ASO start to look at if your brand uh installs uh increase or not. This this might be crucial. On the other hand uh app store search and Google play search also should be correspondent to uh to AI AI activity. This is a a little uh light in the blind uh on uh on the app store. I'm talking about this. You can look at the refer sources such as GPT cloud and Gemini. This graph represents that actually app store analytics enters

enters the AI era showing the app referral calling chat GPT. So if the if uh users based on their prompt will be redirected to the app store directly from for example to the to the app store from chatgpt you can see the traffic incoming from that source and uh last but not least conversion uplift all the all the my recommendation a proposed actions before uh should be trackable on conversion uh to install level. Yes, if you match the keyword intent with your actions, uh creating this special sites, uh building your brand presence and rank strength, you should uh expect much higher conversion to installs. So, it's not the attribution is not going to be perfect, but it's still measurable.

Yeah, summing it up. I want to leave you with five things at the end. Yeah, search is shifting from keywords to and definitely from short queries to longer queries to matching with user needs. Uh LLM shape app discovery before the store visit is happening across all the web. uh when LLM's mention on seated your app uh positioning of your app the unified way of presenting your app the brand signals brand strange yes and also the sentiment play a major role here and everything starts happens of course uh at the at the stores at the mobile stores but started to happening outside of the stores the whole app discovery so the teams that adapt early will have an advantage definitely and um I would say that uh yeah I have an insight from the recent webinar made

by appweek they make a research b during the webinar based on 200 marketers that joined the webinar who has the you know AI strategy for ASO on track who is preparing for it that kind of stuff so only sec six six uh six% of people have their action their strategy uh ready for ASOLMS 29% is building it so we can say one out of three is doing something actually so yeah adapting faster uh it's will be crucial thanks questions thank you very much Mark for this uh talk. Uh quick reminder, please use this QR code to uh rate this talk. It's very important for us. And yes, we still have a [snorts] few minutes for for the questions. So, if you have any,

please raise your hand. Okay. Hold on a second. I give you a mic. So, uh I'm sorry if this has been mentioned before, but what are the best tools in your opinion uh for uh prompt uh research?
In in Europe, you mean? I mean in general like
in general
uh for example if I was to uh do a keyword research for like regular uh search engine optimization I have like Google keyword planner at many many other tools but what kind of tools if any are there for LLMs?
Yeah I would say that u scraping it manually don't work for sure. Uh so don't do this because you will get a lot of single prompts and single intents. So we need you need something like accelerators of

uh intents the the the tool that will show you that the volume for this specific prompter intent is much bigger than one or two or five. Yes. So I I showcased the the two apps. The one is called AI visibility in appeek but it's going to be pretty much expensive. they released this tool for apps but this week or the next week they release also a visibility for games. The one mentioned before also answer the public you can check the pro check the prompts there and I I also I heard from the the people they have SEO agencies and you know just switching just right now to more LLM uh optimization that they are building the tools. So you can also be in a good relationship be in contact with some agencies uh yeah treating this subject LLM especially thank you

thank you for the question another one
I've got a two quite maybe uh connected one are they any disadvantages making changes to the to the descriptions too often. For example, um and um after what time we should see the difference like changing the description will not uh make a change in invisibility overnight but two weeks, month, half a year. Okay. Sure. Thanks for your question. Uh okay. So the case that I presented with Surf SharkVPN that they changed the description is quite new and people across the market do such a thing uh also in the recent two or three weeks. So it's very hard to evaluate just right now. I'm I'm going to do something like this this week in the current company

that I work with uh to also create a description that match all also the keywords like we did in the past but also the user intent from LLM uh also uh answering the user needs and something like this. So yeah, we will see. Very hard to evaluate. Usually it takes from the one months up to even 6 months. Uh regarding the volume installs you're getting to your app. If you have a ton tons of installs, the this whole indexing process will be much faster. So usually for the big apps, it's up to 30 days no more. [clears throat] But it take it can take up to three even to six six months to see the full picture. But yeah, it may be a it may be a faster just right now since the uh AI enters the store with the new core update. So uh just as I

said before that LLMs are based on pre-training data. It's all about when the next update of for example chat GPT or general clot will be made. So uh these LLMs will get this information from the web when you change something. So it it it might be faster just right now in the nowadays. Yeah.