From Rankings to Visibility: Why the SEO Playbook Needs a New Chapter

In this week’s episode of Voices of Search, we spoke with Malte Landwehr, CMO and Chief Product Officer at Peec AI, about what it actually looks like when search behavior changes faster than the systems built to measure it.

Malte discovered SEO at 15, spent five years at Search Metrics, five more leading SEO at Idealo (one of the world’s largest price comparison platforms), and has spent the last several months reverse-engineering how AI models source, ground, and serve information to users. His trip to South Korea last year, where he went nearly two weeks using ChatGPT instead of Google for everything from restaurant recommendations to shopping, was the moment it clicked: this shift is happening faster and cutting deeper than most of the industry is willing to admit.

Today, he broke down why grounding is the most underappreciated concept in AI discovery, what ChatGPT Shopping is actually pulling from, and what a meaningful measurement framework looks like when clicks and search volume are no longer reliable signals.

Key Takeaways From This Episode:

• Grounding is where SEO still matters most: it’s the process by which AI models pull web search results to inform their responses, and most teams aren’t thinking about it clearly enough.
• ChatGPT Shopping primarily scrapes Google Shopping results via a third-party provider rather than pulling from official merchant partnerships, as the media narrative suggests.
• Self-reported attribution is currently the most reliable signal for understanding AI-driven discovery. The gap between click-based attribution and what users actually report can be staggering.
• Consistency of brand description across every web property is the single most important thing smaller brands can do to improve how AI models represent them.
• Automated content at scale has a predictable arc: visibility goes up, then it comes down. Malte has never seen a long-term exception.
• Fan-out queries are the new unit of optimization. These are the multiple search queries AI models generate from a single user prompt, and understanding them changes how you think about content strategy entirely.

A Two-Week Trip to South Korea Changed How Malte Thinks About Search

Malte’s South Korea trip wasn’t a planned experiment. It started as a personal challenge to see if ChatGPT could replace Google for two weeks, and turned into a professional inflection point. Every restaurant, every purchase, every tourist decision ran through ChatGPT. No food poisoning. No tourist traps that didn’t exist. No meaningful failures.

“This change that is happening with AI search is going to come faster, and it’s going to be more severe than I thought,” Malte said. “That’s when I decided I have to be part of this change, not become a victim of it.”

His years at Idealo sharpened that conviction. Idealo spent 15 years in a legal battle with Google over shopping features in the European Union. That fight produced results, but it also produced a lesson: fighting the direction of change is expensive, draining, and ultimately a losing strategy. 

When Malte saw AI search beginning to compress clicks the same way Google Shopping had compressed price comparison traffic, he didn’t reach for a legal brief. He started building measurement frameworks.

How AI Models Actually Retrieve and Source Information

One of the most important and least discussed mechanics in AI search is grounding. Malte gave one of the clearest explanations of how it works in practice.

When a user submits a prompt, the AI model doesn’t simply retrieve a single search result. It generates multiple fan-out queries, shorter intent-specific searches derived from the original prompt, and sends those to a search index. 

Depending on the model, that index might be Google’s API, Bing, Brave Search, or some combination. The results come back as a pool of candidate documents. Not all of them get selected. Of those selected, only specific text passages get pulled into the context window. That synthesized content, combined with the original prompt, is what generates the response.

Every step in that pipeline is a potential drop-off point for your brand.

“Maybe my website was never crawled by Google. Maybe it was not indexed. Maybe it was not ranking for the fan-out query. Maybe it was not selected as a potential source document. Maybe it was not cited. And maybe if it was cited, a part was cited that doesn’t actually talk positively about my brand,” Malte explained. “We need to start understanding this pipeline and understanding where it breaks.”

The practical implication: traditional SEO is still the foundation. Your content needs to be indexed, crawled, and ranking. But that’s no longer enough, as the optimization layer above it is new territory.

ChatGPT Shopping Is Scraping Google. Here’s the Proof

The public narrative is that ChatGPT has official partnerships with Shopify, Etsy, and major US retailers who submit product feeds directly. Malte’s colleague, Tom Welts, reverse-engineered a different reality: the vast majority of ChatGPT Shopping results are scraped from Google Shopping through a third-party provider.

The evidence is in the details. Fan-out queries ChatGPT generates for e-commerce prompts, when entered directly into Google Shopping, produce roughly 80% overlap with the products ChatGPT surfaces. The JSON payload contains identifiers that are recognizably Google Shopping identifiers, including one that points strongly to a specific scraping provider.

“I’m 100% convinced the vast majority is just scraped Google Shopping results,” Malte said.

For merchants, the implication is that if your products are in Google Merchant Center, they’re likely already showing up in ChatGPT Shopping, whether you’ve set up an official integration or not. 

Malte’s advice is to maximize reach and share your data broadly. For most merchants, that data exists elsewhere anyway.

The longer-term picture is a bit more sobering. As AI models get better at handling purchasing decisions autonomously, Malte sees a future where many online shops become fulfillment centers in practice. Holding inventory, managing returns, and absorbing risk, while the discovery and decision layer shifts entirely to AI. “It will at some point come down to price, and that’s always a race to the bottom.”

Your Website Is One Source. AI Models Pull From Dozens

AI search optimization is no longer just about your own website. The sources AI models pull from are diverse by design, and they differ by industry, audience, and query type.

Reddit matters. YouTube matters. In B2B, LinkedIn matters. For audiences skewing younger, TikTok and Instagram are already in the mix. Each of these represents a surface where your brand can be present or absent when AI models go looking for sources.

Malte’s advice on each is measured:

• YouTube requires genuine commitment, not a channel created to game Perplexity citations.
• Reddit requires understanding the platform’s deep aversion to branded content and being honest about whether engagement there fits your ethical approach to marketing.
• LinkedIn is a key source for B2B queries — if that’s your audience, your presence there matters.
• TikTok and Instagram are increasingly being cited for younger audiences, making them worth monitoring even if they’re not yet a priority.
• Major publications like Forbes and TechCrunch show up regularly, depending on topic and geography.

The more sustainable entry point is simpler: type the prompts that matter to your business into the AI models you care about. See what comes up. See what’s being cited. Then figure out where you’re absent and whether there’s a legitimate way to be present.

Brand Consistency Is the Most Overlooked Lever

For smaller and personal brands especially, Malte’s most actionable insight is also the most overlooked: consistency of description across every web property is directly correlated with how confidently AI models identify and represent you.

He’s seen multiple cases where someone searches their own name in ChatGPT and gets back a response that can’t distinguish them from a college basketball player, a hobby musician in Norway, or a third-division soccer player in Germany. 

The reason: their descriptions of who they are and what they do vary enough across LinkedIn, their website, press releases, and social profiles that no clear entity emerges.

The fix isn’t complicated. Pick a consistent way to describe yourself and your work. Use it everywhere. “Overnight, the LLMs understand that when you put your own name into them, you get the right answer,” Malte said. For larger brands, this is less of an issue. For everyone else, it’s the most direct lever available.

A Measurement Framework for a World Without Clicks

Malte organized his measurement framework by priority rather than just listing metrics, and that structure matters:

1. Self-reported attribution: Ask your leads and users where they heard about you. The gap between what click-based attribution reports and what users actually say can be enormous. Malte cited one case where click-based attribution showed 0% of leads from ChatGPT. Direct user surveys revealed 22%. That’s not a rounding error. That’s a fundamental blind spot in how most teams are currently measuring.

2. Prompt-based visibility: Build a representative set of prompts relevant to your business, run them consistently, and track visibility in aggregate across multiple prompts and multiple days. Never draw conclusions from a single chat. Patterns only emerge at scale.

3. Citation share: Of all citations across chats relevant to your business, what percentage come from your own properties? Track this against competitors. If it goes up, visibility tends to follow.

4. Source coverage: Of all the documents being used as sources for your prompts, how many mention your brand, and in what context? “If you get citation fair share and source coverage up, I guarantee you will get visibility up,” Malte said. “And I’ve seen many case studies where once visibility goes up, you also see more sales, more clicks, more revenue.”

On automated content, Malte specifies that the arc is always the same. Visibility goes up, then it comes down. “I’ve never seen it work long-term. If it sounds too good, it will come and bite you at some point.” And on the broader question of what SEO habit teams need to unlearn first, the answer is click-based attribution. “Think more like a brand marketer than a performance marketer.”

SEO Is the Foundation—Not the Whole Building.

The search behaviors that have governed digital marketing for 25 years are changing. Not theoretically, not gradually, but right now in practice. Clicks are declining. Search volume as a metric is distorting. The pipeline between a user’s question and a brand’s visibility now runs through grounding processes, fan-out queries, source selection logic, and citation mechanics that most marketing teams haven’t begun to map.

Malte’s argument isn’t that SEO is dead. It’s that SEO is the foundation, not the whole building. The teams that understand what sits above that foundation and start measuring it honestly are the ones that will have a clear picture of their brand’s health in an AI-first world. Everyone else is flying blind and calling it fine.

Voices of Search is a daily SEO and content marketing podcast hosted by Jordan Keone and Tyson Stockton. The show delivers actionable strategies and data-driven insights to help marketers navigate the ever-evolving world of search engine optimization and content marketing. New episodes air weekly, covering everything from technical SEO to AI discovery, featuring industry leaders and practitioners sharing real-world frameworks and proven tactics.

Subscribe to Voices of Search on Apple Podcasts, Spotify, or your favorite podcast platform. Follow Previsible on LinkedIn for updates and subscribe to the VOS YouTube channel for video episodes and clips. You can also visit the official VOS site to explore the full episode archive and submit your SEO questions for future episodes.