LLM SEO Clearly Explained

DEEP DIVE

For all its ivory tower dryness, academia gets one thing right: rigour.

When researchers make claims, they cite sources. They acknowledge limitations. They build on what came before instead of positioning every insight as one-of-a-kind.

Marketing discourse? Not so much – especially when it comes to SEO’s “death” (and rebirth, and death again).

So instead of adding another opinion to the pile, I spent the last month “reviewing the literature” – i.e., reading dozens of articles, studies, and expert analyses on how LLMs are reshaping search (full reading list here).

I also sat down with people building the infrastructure, like Maxime Dolores, founder of Doppler (think Google Search Console for LLMs), to understand what's really happening under the hood (interview here).

The analysis grew large enough that I'm breaking it into three parts, which I’ll release in between regular editions of this newsletter:

Consider this your crash course in LLM SEO. Let's start with the fundamentals.

LLM SEO is the practice of optimizing your content so AI tools like ChatGPT, Perplexity, and Google’s AI Overviews can easily find it, trust it, and use it in their answers.

The goal is the same as traditional SEO, but the mechanics are fundamentally different.

With traditional SEO, you're trying to rank content on a search engine results page (SERP) so that users see your link and click through to your website. This creates unique opportunities for marketers to “outrank” competitors and drive clicks via compelling titles and meta descriptions.

With LLM SEO, you want your information to appear inside the AI-generated answer itself as a citation or a brand mention. There's no results page to climb, no headlines to A/B test. The AI either uses your content or it doesn't.

Right now, you'll see many names for this same practice:

To be clear, these are all the same thing.

Just as SEO covers optimization for Google, Bing, YouTube, and other platforms, LLM SEO is a catch-all term for optimizing content to appear inside one or more AI tools, which are increasingly being called "AI answer engines."

These AI answer engines include:

📌 I'm calling this out because, unlike traditional SEO, LLM SEO won't stay confined to a handful of search engines. It's already spreading across countless tools and will eventually be baked into everything – from productivity apps to smart home devices.

Kevin Indig recently called this "the fanning out of search," and he's right. We're not just optimizing for a few platforms anymore. We're preparing for a world where AI answers our questions literally everywhere.

If you’re a marketer, this makes having a firm grip on search more important than ever.

Not every AI answer engine works the same way, and the way an answer is generated affects what content shows up.

Broadly, there are three ways these tools generate answers:

1. Training-Data Answers
These answers come purely from the AI’s existing training data. There’s no live web search happening. The response is based on what the model “remembers.” Ask Claude or ChatGPT "What is digital marketing?" and they'll pull from knowledge baked into their training.

🧠 Curious what they’re trained on? You can download huge slices of the internet yourself. FineWeb is basically the web as a giant text file – publicly available for anyone who wants to tinker.

2. Search-First Systems
These pull fresh information from the web in real time. Perplexity and Google's AI Overviews work this way. They crawl current sources, then synthesize what they find. Queries like "latest Google algorithm update" trigger immediate web lookups.

3. Hybrid Systems
These combine both approaches. ChatGPT with Search enabled and Google's Gemini decide on the fly whether to use training data, fetch new info, or blend both. A broad question might tap memory; a timely query pulls fresh data.

Different questions trigger different answer modes – sometimes within the same tool. Understanding these mechanics helps you optimize content for each scenario. (More on tactical approaches in part two of this series).

Now that we know how AI answer engines pull information, it’s important to clarify what “ranking” actually looks like inside an AI-generated answer.

In practice, there are two ways your brand can show up:

Brand Mentions

This is when the AI includes your brand name in its response but doesn’t cite your content directly.

For example: “Popular project management tools include Asana, Trello, and Monday.”

Your brand gets name-checked, but there’s no link back to your site and no guarantee the AI used your actual content to form the answer.

Direct Citations

This is the gold standard. Here, the answer explicitly references your content as a source, usually with a clickable link or a numbered footnote.

Think of the citation numbers in Perplexity or the “Sources” chips in Google’s AI Overviews. This version gives you both the brand mention and the chance for referral traffic.

The main reason this distinction matters is because it directly changes how you track and interpret AI traffic.

In other words:
✅ Mentions = awareness
✅ Citations = awareness + traffic potential

Next, we’ll break down exactly how to monitor all three: mentions, citations, and the clicks they (sometimes) drive.

Once you understand the difference between mentions and citations — and how citations can actually drive traffic — the next step is setting up a way to monitor these signals properly.

Here's the three-layer approach I recommend:

Before you worry about mentions or citations, confirm that AI crawlers can actually access your pages.

How to do it:

No crawls = no citations. This is your baseline check.

Start with what's measurable: actual visitors clicking through from AI tools.

I've written a step-by-step guide for setting this up in Google Analytics 4 and Looker Studio, complete with a free dashboard template.

*Caveat: Not every AI tool passes clean referrer data, and referral traffic only shows clicks, not total brand exposure.

This is where it gets interesting… and where dozens of startups and big SEO platforms are scrambling to solve the same puzzle: How do you track where your brand appears inside AI answers – and for which prompts?

Here three tools I’m finding interesting in this space 👇

Profound – The deepest AI visibility tracker I've seen. Monitors both brand mentions and direct citations, tracks share of voice, and shows actual snippets from ChatGPT, Gemini, Perplexity, and more. Heads up though, Profound charges $499 a month USD.

Ahrefs Brand Radar – Currently tracks your brand presence in Google's AI Overviews, ChatGPT and Perplexity. Shows mentions, impressions, competitive share, and market reach. You can watch their demo here.

Doppler – Positioned as "Google Search Console for LLMs," this tool caught my attention with its competitive intelligence focus. Beyond tracking your own mentions and citations, it shows where competitors appear too.

The catch? None of these tools track prompt volume yet. We can see where brands appear and which queries trigger mentions, but we have no idea how many people are actually asking those questions. It's like doing SEO without search volume data.

Since we’re all flying blind here, I sat down with someone on the absolute frontier of this space: Maxime Dolores, founder (and builder) of Doppler, a tool designed to help teams track their LLM visibility.

What we cover:

You can watch the full chat on YouTube here: Watch the interview →

That’s a wrap on Part 1. In Part 2, we’ll dive into how to actually rank in AI answer engines – with practical tactics brands are testing right now.