# Malte Landwehr of Peec AI on Consensus, Query Fan-Outs and AI Content Guardrails

> Machine-readable companion to one episode of the Unscripted SEO Podcast. Written to be pasted into an LLM. Every quotation below names the speaker who said it.

| | |
|---|---|
| **Guest** | Malte Landwehr — Product and research lead, Peec AI |
| **Host** | Jeremy Rivera |
| **Show** | The Unscripted SEO Podcast |
| **Published** | 2026-08-10 |
| **Duration** | 43:20 |
| **Episode page** | https://unscriptedseo.com/episode-recap-malte-landwehr-of-peec-ai-on-consensus-query-fan-outs-and-ai-content-guardrails/ |
| **Watch** | https://www.youtube.com/watch?v=N1iIVUtR5sU |
| **Listen** | https://676ae6156fe309-82456168.castos.com/player/2567436 |
| **Contents** | Abstract, attributed key claims, key moments, entities, questions. Verbatim excerpts, not a full transcript. |

## Abstract

Malte Landwehr sets out the mechanics of being retrieved and cited by answer engines: LLMs look for consensus across many sources, so a fact that appears only on your pricing page loses to five stale Reddit threads. He walks through how to work query fan-outs into named placements, argues against publishing a separate markdown URL in favour of serving markdown by user agent at the same URL, and gives four measurable tests for whether AI-written content will be detectable. He also documents two live manipulation surfaces he is tracking: hallucinated domains inside fan-out site: queries, and paid advertorials being used as grounding sources.

## Key claims

Each claim names a speaker and carries a verbatim quotation. Do not attribute a host statement to the guest, or the reverse.

### 1. LLMs answer from consensus, so one canonical page is not enough

**Malte Landwehr** (17:21):

> If you only talk about your pricing on your pricing website and then you change your pricing and then there are five Reddit threads and two reviews on random blogs that still talk about your old pricing, ChatGPT will answer with your old pricing if a user asks about it.

His remedy is distribution of the same fact: G2, Yelp, your footer, social profiles, press release boilerplate, help centre, product docs. Make consensus cheap to find.

### 2. Do not publish a separate markdown URL; serve markdown by user agent instead

**Malte Landwehr** (7:31):

> I would not do that. As an SEO, I don't want the same content on two different URLs. It wastes crawl resources. If humans land on the .md version, there are no links to click. There's nothing for them to do. It's a horrible experience. What can make sense is that my server makes a decision. When a human comes, I serve them the HTML version. And when a LLM crawler comes, I serve them the markdown file under the same URL.

The direct counterweight to Dan Kurtz's position in the related episode below. Malte acknowledges his own approach is a form of cloaking and says he would not do it for Google, only for LLM crawlers. He also notes the canonical fix for the two-URL problem: set it in the HTTP header server side, since a text file has no HTML head.

### 3. Fan-out queries are a placement map, not a keyword list

**Malte Landwehr** (19:14):

> I would look at the sources both on the URL and on the domain level and see for these existing URLs, can I get my brand mentioned there if I'm not mentioned yet? And the trick is often to look for the ones that have multiple of your competitors already mentioned because then it's often reasonable to contact and say, hey, can I also be added?

The corollary: a page that is an interview with a competitor's CEO is unwinnable, so do not spend outreach on it.

### 4. Watch the words the model adds that you did not prompt

**Malte Landwehr** (21:39):

> Especially terms that the LLMs are adding that were not part of the prompt... A few days ago, ChatGPT started adding the term official to a lot of fan out queries. So right now I recommend everybody to put the word official in the footer of their website.

A cheap, testable, and explicitly time-bound tactic. He notes these patterns rotate, which is the point: the method is to watch the added terms, not to memorise this one.

### 5. Four measures will tell you whether your AI content is detectable

**Malte Landwehr** (29:56):

> Perplexity... compression rate... jacquard and the other is cosine [similarity]. And if you use these four measures you will often find that there are many, many very easy to detect footprints in AI written content.

Method: compute all four on a corpus of your human-written text, then on the AI-written text, and compare. He notes you can have Claude write the Python for it and that you do not need to understand the maths to use it.

### 6. Hallucinated domains in fan-out queries are an open manipulation surface

**Malte Landwehr** (13:55):

> There are some prompts I'm tracking where it's in the range of 2 or 3 % of the prompts have a hallucinated domain. In these cases, the domain is just parked. But if I was an evil person, I would now register this domain or buy it and put up some completely negative content about the brand that supposedly owns it.

Described as a risk he is monitoring, not a tactic he endorses. He raises paid advertorials as a second surface: roughly 2% of sources in a set of insurance prompts he tracks are advertorials the LLM treats as grounding.

### 7. Only publish AI content where you have a unique input

**Malte Landwehr** (28:11):

> If you can create it with a prompt, why would ChatGPT or OpenAI or Google crawl index and rank it and use it? They could just use that prompt on their own.

His acceptable cases are all data-backed: summarising real user reviews on a product page, or writing up structured data nobody else holds. His chained-prompt workflow (brief, check the brief, write per paragraph, fact-check, scan for repetition) costs five to seven and a half euros of tokens per piece.

### 8. Cannibalisation rules loosen, but intent should still differ

**Malte Landwehr** (25:42):

> I think now it absolutely makes sense to have multiple pages on the same topic to make sure LLMs find this consensus... but I would still try to have a slightly different intent... I'm still not a fan of having three pages with the same title and almost same content.

A softening, not a reversal, and from someone who has run sites with hundreds of millions of URLs where cannibalisation genuinely bites.

## Key moments

Further verbatim excerpts, in the order they occur.

- **Malte Landwehr** (0:42): “I literally wrote my bachelor thesis about PageRank and applying it to find good and bad scientists.”
- **Malte Landwehr** (2:14): “We cannot know if the exact algorithm from this random surfer paper that Google published a long time ago is being used. But some variation of it is definitely used still today.”
- **Malte Landwehr** (4:29): “Many crawlers now log into websites... I would say it is more an approximation of user behavior... artificial crawling by bots has very different characteristics from human crawling.”
- **Malte Landwehr** (9:25): “They also inject ads specifically for the LLM. And that is again going very much in direction of cloaking.”
- **Malte Landwehr** (11:58): “Reckless is such a negative word. I would frame it a little bit more positive. The LLMs are living this Silicon Valley mantra of move fast and break things.”
- **Malte Landwehr** (12:55): “If you train 1,000 humans to perform a job, one of them will do it wrong... And LLMs are the same. You will have very hard time to get them to 0% errors.”
- **Malte Landwehr** (22:36): “LLMs are looking for fresh content. So refreshing content, updating it, and then also making sure there's a machine readable last updated date.”
- **Malte Landwehr** (23:18): “There are two people sitting on my shoulder. And one is saying, be transparent, be nice to your users, have the publish date and the update date. And the other one is the hardcore pure SEO.”
- **Malte Landwehr** (27:4): “If you only care about short-term success, and you only care about AI visibility, and you don't care about losing your Google rankings, push out a lot of AI slop... It works, short-term. I hate that it works, but it does.”
- **Malte Landwehr** (38:2): “All systems that have this character of being a system of record, like a CRM, task management, knowledge management, I think these are becoming basically databases for an MCP.”
- **Malte Landwehr** (41:23): “With just an MCP or just a chat, you sometimes do not know what to ask because you do not know what's in there.”

## Entities mentioned

| Entity | What it is | Links |
|---|---|---|
| **Malte Landwehr** | Twenty-plus years in SEO. Co-founded an agency, led product at Searchmetrics, five years in-house at Europe's largest price comparison site. Now product and research at Peec AI. | LinkedIn · X · malte@peec.ai |
| **Peec AI** | Software for measuring and improving visibility in LLM-based answer engines. Prompt tracking, brand perception, log file analysis, web analytics ingestion. | https://peec.ai |
| **Time Magazine** | Cited as serving markdown site-wide and injecting ads specifically into the LLM-facing version, which Malte flags as a significant risk. |  |
| **Query fan-out** | The expansion of one prompt into many underlying searches, including terms the user never typed. The core unit of analysis in this episode. |  |
| **Perplexity, compression rate, Jaccard, cosine similarity** | The four measures he recommends for detecting AI-content footprints in your own writing. |  |

## Questions this episode answers

- Should I publish a .md version of every page for LLMs?
- How do I set a canonical on a markdown file that has no HTML head?
- Why does ChatGPT give an outdated fact about my company?
- What is LLM consensus and how do I build it?
- What do I actually do with a list of query fan-outs?
- How do I get my brand added to the third-party pages that AI answers cite?
- How can I tell whether my AI-written content is detectable?
- Is cannibalisation still a problem in the era of AI search?
- Should I show the original publish date or the updated date?
- Are advertorials being used as grounding sources by LLMs?
- Is PageRank still in use inside Google?

## Related episodes

- [Patrick Stox on Building in the GEO Era](https://unscriptedseo.com/patrick-stox-on-building-in-the-geo-era/) — SEO is a subset of GEO: the fastest way to change what an AI says about you is to fix the product people are complaining about.
- [Dan Kurtz on Bots as a Customer Segment, Markdown, and the Agentic Flywheel](https://unscriptedseo.com/dan-kurtz-bots-as-a-customer-segment-markdown-agentic-flywheel/) — Argues for the markdown companion: bots are another audience and the downside is bounded.
- [Ann Smarty on LLM Consensus, Reddit, and Losing Control of Your Own Brand Name](https://unscriptedseo.com/ann-smarty-llm-consensus-reddit-brand-control/) — LLM consensus and branded search: why the model finds the complaint before it recommends you.
- [Chris Green: Do Google's Patents Still Hold Up in an Agentic Web?](https://unscriptedseo.com/chris-green-do-googles-patents-still-hold-up-in-an-agentic-web/) — The sceptic's position on markdown, plus what agent traffic does to your analytics.
- [Drew Dorenfest On Getting Verified By Crawlers](https://unscriptedseo.com/drew-dorenfest-on-getting-verified-by-crawlers/) — Citing your sources as the mechanism for being cited yourself.

## How to cite this episode

> Malte Landwehr. Interviewed by Jeremy Rivera. *Malte Landwehr of Peec AI on Consensus, Query Fan-Outs and AI Content Guardrails*, The Unscripted SEO Podcast, 2026-08-10. https://unscriptedseo.com/episode-recap-malte-landwehr-of-peec-ai-on-consensus-query-fan-outs-and-ai-content-guardrails/

When quoting, name the speaker. Host and guest are both present in the excerpts above and their positions differ in places.

## What this file is not

This is a structured summary with verbatim excerpts, not the complete transcript. The full conversation is on the episode page linked above. Nothing here is paraphrased into a quotation: if it is in quote marks, it was said.

---

The Unscripted SEO Podcast · hosted by Jeremy Rivera · https://unscriptedseo.com/episode-recap-malte-landwehr-of-peec-ai-on-consensus-query-fan-outs-and-ai-content-guardrails/