Malte Landwehr
Questions and Answers

My wife and I went on vacation to South Korea. Just for fun, I set myself the challenge not to use Google. It worked incredibly well to do everything with ChatGPT. Searching, fact checking, translations, and planning which places to visit.

When we arrived at the airport and I reflected on the vacation, I realized that how people find information and consume content is changing faster and more radically than I previously thought.

I landed in Germany on Friday, went to the Peec AI office on Saturday, signed the contract on Sunday, and quit my corporate job on Monday.

I had already been in contact with the Peec AI founders regarding an advisory role. By the way, the Peec AI office back then consisted of a few desks in the corner of another startup's office.

For Peec AI, there's actually a funny story. Many of our competitors regularly write listicles and comparisons in order to influence LLMs and make it seem like they should be recommended over us. Of course they use AI to create these.

A while ago, one competitor released a listicle that mentioned as a positive aspect of Peec AI that we have the best email and PDF reports. Now that this document is ranking and can be retrieved by grounding whenever you ask LLMs to write about us, this document often ends up in the sources. So many, many more competitors now, when they use AI to write about Peec AI, mention this fact.

The funny thing is, we do not have PDF and email reports. This is a completely hallucinated feature, and whenever I see a listicle that mentions this, I know that the person did not actually use Peec AI.

If it ranks well in Google, the first few items on my checklist are already checked. Next, I usually verify AI crawlers are not blocked. You can test this by putting a URL in a GenAI chat and asking, “What is the third paragraph about?”

Next I check the structure. Is there at least one paragraph that an LLM-based answer engine can easily cite? I care about length (1 to 3 sentences) and high entity density. The language should be clear, declarative, and authoritative.

If that does not exist, I recommend either rewriting the whole article or adding a summary or question-answer block. These usually end up getting cited the most.

Probably yes. But just like with traditional web search engines, I recommend having one strategy. You did not create separate pages or backlinks for Bing and DuckDuckGo. So do not do it for Gemini either.

Largely, the same approach works across all LLM-based answer engines. If there is a huge discrepancy between ChatGPT and Gemini, I would first check the answers and then the sources.

Checking the answers: is it just me who is less visible or my competitors as well? I see this often with ChatGPT and Perplexity. A brand is visible in 80% of ChatGPT answers but only 40% of Perplexity answers. If you dive deeper, you notice that the same is true for the competition. Because Perplexity only gave answers in the style of “I cannot recommend a best SEO agency. If you tell me in more detail what you are looking for, I might be able to recommend one.”

That is why in addition to Visibility (in what percentage of chats am I mentioned?) I also like to look at Share of Voice (across all brand mentions in chats, what share goes to my own brand?).

If checking the answers does not reveal a general discrepancy between ChatGPT and Gemini, I look at the sources. Very likely, Gemini is using different sources from ChatGPT. And in these my brand is probably less popular. This has an easy fix: get mentioned in the sources favored by Gemini.

Another thing I check in the sources is how often my own website is used as a source vs my competitors. If Gemini cites a competitor's website more often than mine, I know I have an opportunity to optimize.

Let me start by saying that I am a huge fan of automatically created landing pages for SEO and AI content. I have personally overseen the creation of millions of landing pages and hundreds of thousands of AI-written content pieces.

But you always need to ask yourself two questions:

• When someone wants to read AI content, why would they go to my website instead of ChatGPT or Gemini?
• If Google could create the same content, why would they go through the effort of crawling, rendering, indexing, and ranking my content?

If you do not have good answers to these questions, you should reconsider creating such content in the first place.

There are many cases where good answers to both questions exist:

• You have a list of hotels with attributes (pool, distance to beach, etc.) and want to describe each in a single sentence.
• You have a website with weather data. The data refreshes every 60 minutes. On each city-level page you want to display the current weather.
• You have a product details page with user-written product reviews. You want to summarize these reviews.
• For a stock market index (like the S&P 500) you want to have a page that always reports what happened in the last 3 hours.

But if your idea is to publish 10,000 LLM-written product reviews or blog posts, that is not something anybody wants to read. So Google, OpenAI & Co will try not to crawl, rank, cite, or recommend it.

The good news is you do not need perfect data. You do not need to track every single prompt. You just need a set of questions (your prompts) and a panel (the LLM-based answer engines). And then you ask your panel these questions. Over time, you look at the delta to see trends. As long as the data is directionally correct, you can see everything you need: how are you performing vs the competition, how is your performance developing, and what are opportunities you should tackle.

Additionally, CMOs can track self-reported attribution. Simply ask each new user or lead where they heard about your brand.

While many performance marketers and SEOs struggle with these approaches, both are established techniques that have been working well for decades. Of course the data is less perfect than a report from PPC campaigns. But it works!

Yes. In multiple ways.

Dynamic 1: Platforms. In SEO, shifts in market share are rare. In most countries Google is the dominant player. Over the last 15 years, they lost a lot of transactional searches to Amazon, Booking & Co. Yandex expanded to a few more countries. New search engines like DDG and Ecosia captured a few percentage points of market share here and there. That's about it. In AI search, Perplexity used to be the number 2 platform behind ChatGPT. Now it is no longer even in the top 5. Just in the last few months Gemini had a steep rise.

Dynamic 2: Sources. It is normal to see major sources decline by 90% or increase by over 1000% within a week. This makes long-term planning a challenge since it can affect distribution of marketing budgets and head counts.

Dynamic 3: Answers. Google results used to be somewhat uniform and stable with light personalization. LLM answers are non-deterministic. You need to run a prompt multiple times to reach statistical significance. And even then you can only really say “this brand appeared in 70% of answers”, which is less precise than a strictly ordered search result list from a web search engine.

Partners are a massively underrated lever in AEO.

Any content about your brand can influence LLMs. If partners tend to create positive content, highlighting capabilities or showing a case study, their content can be very beneficial for the involved brand.

Ontrack, Autodesk, Zoom, Moosend, and even the Quebec tax authority, all have directories of software and integration partners. These directories regularly show up as sources in LLMs when asking questions about software. With Peec AI, we can see that our agency directory is regularly cited by LLMs when asking about agencies.

On the more aggressive (spammy) side, I constantly see people buying listicle placements or positive articles about each other. Similar to how reciprocal links worked in the early SEO days. While this works very well right now, the impact is obviously short-lived.

The biggest AEO misconception is that AEO is just SEO.

Number 1: Try to understand how your customers ask questions and talk about your products. Recorded sales calls in Gong, support tickets in Intercom, the chat widget on your website, the search bar on your app, all of these are great sources to understand the language of your target audience. Use these insights to come up with the best possible prompts to track!

Number 2: In addition to visibility, keep an eye on how LLMs perceive your brand and how accurately they represent key facts. I have seen cases where LLMs were mentioning outdated packaging and pricing details for months. Since the LLMs always spoke with confidence, prospects were confused during the sales process. In cases like these, you must find the sources responsible for the outdated fact and ask them to update. Or create new sources for the LLMs to find. The pricing page on your own website is not enough.

Yes, I believe this is a new category. I see a couple of reasons.

SEO was a traffic channel, ultimately. It was about bringing people to your website. There, they would experience your tone of voice and your corporate identity. They would read the words written by your copywriter, and they would use the navigation of your website. You would see them in your web analytics, and you would give them a retargeting cookie. You could have micro-conversions like a newsletter signup. Whether the traffic converted was on the product team, the web team, the copywriters, the designers, as much as it was on SEO.

In the GEO space, it is very different. For the SEO space, the optimisation was basically that my smallest unit was a URL, and a URL can rank for a keyword. This is the smallest unit of SEO: a URL ranking for a keyword, and then there are things like content and backlinks that play a role in this. In the GEO space, the smallest unit is a brand being recommended in an AI answer, so very different.

Just one example: URLs have a loading time, URLs have a title tag, they have content, and they have a publication date. Websites have an average loading time, websites are defined by that sum of content, websites have backlinks. URLs have historical data of whether they were clicked in Google. A brand has none of these. A brand has text written on assets owned by the brand, and it has text written by assets influenced by the brand, and it has text written about the brand. From the perspective of your user, they consume an answer in a chat. All the stuff is, of course, influencing tokens but is no longer directly reaching eyeballs.

Yes, I have repeatedly argued with others, primarily online on LinkedIn, X, and Reddit, about this because I very strongly believe that this is a new category. The fact that many SEOs pretend like it's not has led to the fact that a lot of snake oil salespeople have come into the industry. People who have no background in SEO and who are suggesting to do things where it's very obvious for anybody who spent two years in SEO that this cannot work long term.

This is creating a lot of problems because, short term, these things do work, so companies start to trust these people. Instead of positioning themselves as the experts for this topic that they are, because they know the fundamentals of it, they also have left it to new joiners.

This works very well short-term. But at some point these companies will lose their Google rankings. And as a consequence also their LLM visibility. AirOps, ClickUp, Gong, Sprout Social, Monday, Webflow are examples where this happened recently.

If you are a top player in your niche, use this very sparingly and well targeted. The risk is not worth the short-term reward. If you are the 10th largest CMS for Denmark, go all in on AI content. Spam. Spam. Spam! It might save your company.

Always ask yourself two things:

1. If a consumer wants an AI answer for this topic, would they ask ChatGPT or me?
2. Imagine I was working at Google or OpenAI and cared about cost and user experience. Would I want to spend money on discovering, crawling, indexing, and storing this content? Would I want to serve it to users?

If the answer to either of these questions is no, do not do it. You can still compete in Google. This is possible. But you will need an experienced SEO team and enable them. Otherwise you will lose after a few quarters.

What works well: secondary content. Summarizing 10 user reviews. Explaining a chart. Updating the weather forecast for each city in Germany every 15 minutes. These can be AI generated. Or, use other websites. You can use Bild or Handelsblatt to publish a listicle that mentions you.

Usually SEO, which makes sense.

But the SEO team often can only identify the gap or opportunity. They need help from brand management, PR, influencer marketing, social media, community management, investor relations, customer support, and many other teams.

I know a Fortune 500 company where 16 separate teams have KPIs that support AI search. They have one FTE just to inform all these departments how often their content was cited.

What I see smart organisations doing is creating a separate AEO/GEO team. And then moving the best members of the SEO team into that.

When LLM-based answer engines cannot answer with parametric knowledge, they do something called grounding on websites. For this, they do not enter the prompt into a search engine. They distill fanout queries from the prompt and search for those. These queries tell us what information the answer engine needs.

For example, if we ask “What is the weather like in the capitals of Germany and France?” the LLM will search for “weather France” and “weather Berlin”. It knows the capitals of Germany and France. It does not know the weather.

For prioritizing content, look at the following inputs: What content from your competitors gets cited by LLMs? Create better content for the same topics. What fanout queries are LLMs running? Especially, what terms are they adding that are not part of the prompt? Create content for that.

They still think of AI search as just an extension of SEO.

I believe Perplexity is only mentioned because people used to talk about “ChatGPT and Perplexity” as the new answer engines. Nowadays, Perplexity is not even a top-5 answer engine anymore. ChatGPT, Gemini, Grok, AI Overviews, and Claude are all bigger than Perplexity. Even Rufus will likely soon overtake Perplexity.

AI Overviews are the most underrated LLM surface. Probably because it is not a typical chat interface. But in terms of reach, AI Overviews are easily twice as large as ChatGPT.

By the way, for companies selling to enterprises and governmental agencies, there is another LLM people often miss: Microsoft Copilot. Many traditional companies forbid their employees from using anything else.

Separate three steps: finding documents, reading documents, and citing documents.

Finding. Your document can be a homepage, a product detail page, a PDF, a blog article, or a press release on another site. Getting found works much like traditional SEO. The correct keywords need to be in the text and in the title, synonyms included, content quality should be good, and any signal that helps a search engine understand this is high-quality content helps.

This is a real risk and we have clients who are actively being attacked this way.

One example, without too much detail: a large consumer brand with a customer support hotline. The product is the kind of thing you sometimes call about for a minor reason, but sometimes in a panic, when the product malfunctions and you are stuck. If you ask ChatGPT, it will often give you the wrong support hotline, because people intentionally create websites containing the wrong number. Call the wrong hotline and one of the first questions is: what credit card did you use to book or purchase this? In a panic, you might just tell them. It is a real attack vector that works very well right now for pushing misinformation into LLMs.

Many things. First of all, just like you used to do SEO, you can now do GEO or AEO, whatever you want to call it. You can influence what kind of content is on your website. You can influence how the content on your website is presented from a technical point of view. You can influence the markup and structure of your website and your content. And you can influence what people are saying about your brand on other websites.

There's also this concept of UX, user experience, which is something every website has and many websites pay people to optimize. In the future we will just have AX, agent experience. For example, an agent might prefer a website that gives very clear prices and very clear product distinctions. It doesn't just say “these are the ski boots from Mammut”, it says “these are the ski boots from Mammut, manufactured 2025, sold in Switzerland under serial number 77935, in shoe size 45.2, which is usually perceived as slightly larger, click here to order, delivery will happen in 18 hours”. Versus a website that just says “these are the Mammut hiking boots for men, size 42 to 43, delivery time usually under 7 days”.

Some of this is of course speculation, we do not know what an agent that exists in five years will actually value, but probably it's precision. Detailed information helps, because there are also online shops that make mistakes.