Marketers of Technology
00:11:36 · Malte Landwehr
[00:11:36] S0: Yes, gladly. I mean, what do we do specifically? We define together with them a set of prompts. And then every single day we simulate a human opening something like ChatGPT or gemini.google.com in their browser, typing in that prompt, getting the answer, looking at the sources, identifying the brands and these answers. It all goes back into the software. So step one is you have a reporting, you know where are you appearing, where are you not appearing, and what source are the LLMs using. Then we also download all of these sources and check, is your brand mentioned in the source? Are your competitors mentioned in the source? And then the most easy thing we can give you is we can tell you, hey, these are content pieces that your competitors published that the LLMs are constantly citing. Please consider publishing something similar if you believe it fits to your brand. And we don't just do this for your website, we also do this for LinkedIn, Instagram, Reddit, like anything we can find. We also tell you, hey, there are these third party sources where your competitors are mentioned and the LLMs are citing it and you are not mentioned. So that can be maybe you are missing the G2 profile page, or maybe all of your competitors have an interview on Forbes and you didn't have one. And we make you aware of that. Or maybe there's a listicle from Zapier where you would fit in perfectly, but you are not listed. All of this information we give you. And then we also go one step deeper and recommend to you how to rewrite your existing content. Because when LLMs download your website to maybe use it as a source, they don't take the whole document. They take chunks of texts and score each of them and only use those one or multiple of those chunks. So if you have a paragraph where it says, because of the above mentioned reason for these types of users, the third solution is the best. If you take that out of content, this actually just contained zero information. So then we would recommend to you, hey, maybe reformulate this. Please write it in a self contained manner. Please specifically mention the entities you are talking about.
00:13:46 · Malte Landwehr
So you could, for example, say based on reviews on Reddit and G2, Peec AI is the best software to track your LLM visibility globally because they support Naver AI briefing, or however it's pronounced in China, DeepSeek in China, Qwen in China, full stop. So if an LLM takes this out of context, it still contains all of the information. It is written in a declarative, authoritative manner. So it's not saying, oh, maybe this could be the best solution. It also gives a reasoning. It's just a random statement like this is the best, but it actually gives reasons why it's considered the best. And it also has verifiable facts because an LLM that is doing some level of reasoning can then do a search and check, oh, does Peec AI actually support Quen in China? Does Peec AI actually support DeepSeek in China? And then it can hopefully find information on the internet that this is true. Because one big thing with influencing LLMs is they are almost always doing a web search or a grounding process or rack, however you wanna call this, and they pull in sources and they are looking for consensus. If five sources say the same thing, the LLMs are very sure that this is true. If only one source claims something that none of the other sources can back up, it's much, much less likely to make it into the final answer.
00:15:06 · Andres Figuera
[00:15:06] S1: Okay. I understand a lot better now. And Malte, then from my understanding, during the customer's journey here, you provide the reporting, you let them know what they should change, but you don't go in there and change it for them. You just provide the recommendations, and then they decide if they want to follow that or not.
00:15:25 · Malte Landwehr
[00:15:25] S0: So we have an in app agent, and you can connect something like Webflow via MCP. And then you could take one of our recommendations and tell our in app agent, please implement this. We don't necessarily recommend to implement everything blindly without human review, because you should always think about, does this fit my brand? Are there may be accessibility or legal requirements in my specific market, my specific region? Is this still properly my brand voice, etcetera, etcetera? But in general, yes, we can implement things also for you. And we also, we have clients where I know that they have an automation set up in Claude, where every morning they just take all of the actions. Like that's the name for our recommendations feature. They take all the recommendations from Peec AI and just tell it to implement it. And it's connected to their WordPress or their Webflow. And I also have test websites that are running like this. I always tell people, look, we put a lot of effort into making sure these are good recommendations, but please double check it. Like there can always be a recommendation in there that maybe doesn't fit your brand, or let's say there's a recommendation to build an MCP, you shouldn't just build an MCP, right? Like, please think about what you want to offer. But in general, yeah, we can do also the implementation part.
00:16:51 · Andres Figuera
[00:16:51] S1: No. But I do I do appreciate that you give them that little disclaimer because it's made with AI, and I imagine, and then it's been through its rigorous process of testing and making sure that everything is going well. But AI can hallucinate sometimes with what you're mentioning, it's the brand image or their voice or how they want to be portrayed online that they really need to check that. And that way you're also not held liable for providing some bad advice that somebody blindly followed. So I really like how your guys are doing it right there. And, Malte, now shifting the perspective a bit more into the future, what would you like to say that you were able to accomplish towards the end of two thousand twenty six? To the end of this year. That's that's right around the corner.
00:17:31 · Malte Landwehr
[00:17:31] S0: So I'm very confident that we will pull in a lot more interesting data like Google Search Console, for example. I'm very confident that on the agentic side, there will be massive improvements. Like we already have an in app agent, we have skills, you can have custom skills, you can have context about your company. But I believe that there's more we can do for you in an automated fashion and also convince people to put some of the prompts they used to put in Claude to put into our agent instead, if it's SEO or AEO or GEO related, because we just have more context there. So for example, our in app agent has access to our own research database and can always use all of the learnings we've had by running those experiments, reverse engineering, a lot of things, and use that for making decisions and making recommendations. And of course, there are many more systems we can integrate with to funnel the output of our work into. I think those are some of the biggest things we will do until the end of the year.