A Guide to Web Guide: Our Hybrid Search Future

When Google launched AI Mode, Liz Reed (VP & Head of Google Search,) wrote that:

AI Mode uses our query fan-out technique, breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf. This enables Search to dive deeper into the web than a traditional search on Google ...

While “multitude” may be an exaggeration, AI Mode visually confirmed that Google ran additional searches and analyzed multiple SERPs (this message seems to be gone now). Typically, this was in the ballpark of a half-dozen searches. For example:

This is the question we’ve been exploring at Moz over the past few months. Based on Web Guide headers, could we start to piece together the types of query fan-outs Google might be performing and generate our own fan-out queries?

Using observed headers, we set out to understand how the fan-out process might operate. This involved a fair bit of deduction, and the end result isn’t a model of exactly how Google works, so much as a framework for understanding the broader process.

Our research eventually settled on a 10 fan-out taxonomy, illustrated below:

These 10 fan-out types are loosely organized as a journey. What is the searcher trying to achieve, and what are the steps they might take along that journey toward an eventual action? Let’s dive into each one and look at some examples.

As we move toward this hybrid world, it can help to move away from traditional search queries and towards questions (or, in LLM lingo, “prompts”). So, all of the following examples are based on the question: “What are the best wireless mice for gamers?”

1. Semantic fan-out

In an AI/LLM world, there are many ways to say roughly the same thing. Semantic fan-outs are queries or prompts with different words, but similar meaning. For example:

• “Which gaming mice perform best for players?”
• “What are good cordless mice for gamers?”

2. Entity fan-out

Entity fan-outs expand on an entity — a person, place, thing, or brand. For example:

• “Are HyperX mice good for FPS players?”
• “Is Pulsar a solid pick for gamers?”

3. Follow-up fan-out

Follow-up is probably self-explanatory — what is the searcher likely to ask next? The next two fan-out types are variations on follow-ups, but some general examples are:

• “How long do wireless gaming mice last?”
• “Are wireless mice good for competitive play?”

4. Attribute fan-out

An attribute fan-out is a follow-up focusing on a specific feature or attribute, such as:

• “How important is battery life in gaming mice?”
• “How much DPI matters for gaming mice?”

5. Anticipate fan-out

What about the follow-up after the follow-up? Anticipate fan-outs look two or more steps ahead, sometimes even at post-transactional questions. For example:

• “What sensitivity settings help new gamers?”
• “What accessories help wireless mouse users?”

6. Factual fan-out

The next three fan-out types dive into the informational intent space. Factual fan-outs focus on just that: facts, data, and specifics. For example:

• “What polling rates do pros often choose?”
• “How often do gamers replace wireless mice?”

7. Tutorial fan-out

Tutorial fan-outs dive deeper into a topic, such as how-to guides. The three informational fan-outs are strong candidates for content marketing. Examples include:

• “How to extend battery life on gaming mice”
• “How to set up a new gaming mouse?”

8. Perspective fan-out

Perspective fan-outs focus on human perspectives and opinions, including discussions and forums, which Google leans on heavily. Some examples are:

• “What mice do streamers often recommend?”
• “Why do some gamers swear by wrist rests?”

9. Comparison fan-out

The last two fan-out types move down the funnel toward an action or transaction. Comparison fan-outs compare two or more items or concepts. For example:

• “Which lasts longer: optical or laser mice?
• “Are premium mice really better than budget mice?”

10. Transact fan-out

Finally, Transact fan-outs move into a clear commercial or transactional space. The searcher is ready to take an action, such as a purchase. Examples include:

• “When do gaming mice usually go on sale?”
• “Where to buy mousepads for wireless mice?”

Note that these 10 fan-out types aren’t exhaustive or mutually-exclusive. Our goal is to provide a framework to understand why and how fan-out happens, and illustrate how mapping fan-outs can help marketers better understand the searcher’s journey.

Putting all 10 together

Let’s revisit our earlier visual map and plug in some of the examples we’ve covered:

Here’s the good news — hybrid search is at least partially organic, and Web Guide leans heavily on core Google search algorithms. Moving toward a hybrid future, here are three things that we have to keep in mind:

1. Search is a conversation
Whether it’s refining a query and clicking on results or a literal back-and-forth with an LLM, search is a conversation. Natural language is our default mode as humans, and the shift toward conversational search has been happening for well over a decade.

2. We’re going to have to let go
No matter how well you craft your content, headers, or meta descriptions, Google is going to summarize and interpret them more and more. Like you, I’m sometimes frustrated by this, but it’s important to focus on moving the searcher in a mutually beneficial direction, even if that means giving up on trying to control every step they take.

3. Ranking is going to get messy
Hybrid search and Web Guide, in particular, still have rankings, but consider two things. First, if a given fan-out better serves the searcher’s needs, they might prefer that section of results. In other words, result #6 could be the most relevant if it’s the first result in a fan-out that most aligns with a searcher’s intent and needs.
Two, and possibly more alarming, the results within any given fan-out are probably coming from the top organic rankings of that sub-query. In other words, result #7 could actually be result #1 or #2 on a different query. That means we’re going to have to compete on multiple sub-queries to rank on any given Web Guide results page.

You’ve probably heard marketers talking as if a Gemini-only, AI Mode future is inevitable. I think that’s extremely unlikely, for a lot of reasons, but let’s focus on the two big ones.

1. LLMs aren’t good at everything

Even OpenAI has admitted that hallucinations are a feature of LLMs, not a bug. Search-grounding and retrieval-augmented generation (RAG) can help, but they don’t solve the underlying problem. LLMs are probabilistic language simulators, and they have many limitations.

Even without hallucinations, LLMs are poorly suited to some tasks, such as navigational and local searches. So, you already see Google adding links and features to AI Mode, essentially turning it into a hybrid search result.

2. Generative AI costs a lot of money

We know that OpenAI is bleeding cash, and early estimates suggested that LLM prompts cost Google and Bing at least 10 times as much as traditional search. While infrastructure is improving and costs are going down, there’s also pressure to launch newer, bigger, and more capable models, which drives costs back up. Google immediately rolled Gemini 3 into AI Mode when it was released, for example.

At the same time, AI Mode results have been harder for Google to monetize, disrupting Google’s long-term efforts to use SERP features to drive searchers down the funnel toward queries where ads and shopping results are more effective. Less revenue at substantially greater cost is simply not sustainable.

Note that this isn’t a binary — we could land on a hybrid approach that’s somewhere in-between Web Guide and AI Mode, or we could end up with logic that routes some searches to AI Mode (letting it do what it’s good at) and others to Web Guide.

The narrow answer is that I think Web Guide and other varieties of hybrid search are good news for the SEO industry — they’re grounded in organic results and the core algorithms we already know and love (or know and hate, depending on the day).

The broader answer is that — whatever form Google search takes — humans want to use natural language, and machines are improving at delivering natural language. This is also good news, and it’s a path we’ve been on for over a decade.

Whether the future leans more toward traditional organic results or more toward LLMs and AI-driven results, we need to understand the broader search journey, the paths people take toward action, and how to serve their needs at multiple stops along those paths. Web Guide is showing us the signposts along the way.