How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov
Listen on your favourite platform
| Platform | Link |
|---|---|
| YouTube | Listen on YouTube → |
What Does “How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov” Talk About?
This episode of the James Dooley Podcast dives deep into the mechanics of query fan-out inside large language models, featuring a focused conversation between host James Dooley and SEO expert Sergey Lucktinov. The two break down exactly how LLMs like Google Gemini, ChatGPT, and Perplexity take a single user query and expand it into multiple background searches, explaining why more complex queries such as best massage chair for people under one thousand dollars generate far more fan-out queries than simple factual questions like what is the capital of France. Sergey details how Perplexity can generate up to forty fan-out queries while Google typically uses up to twenty, and how each fan-out query pulls roughly ten results, creating a candidate pool that can reach two hundred websites before filtering even begins.
The episode also walks through the multi-stage filtering process LLMs use to narrow down that large pool of candidate pages. Sergey explains that in the first stage, models perform a light metadata analysis covering page titles, meta descriptions, site names, and trust signals, with Google Gemini pulling directly from Google's own link profile and penalty data. This alone can eliminate forty to eighty percent of candidates. A second round of structural analysis follows, and only then is a full content extraction performed on the surviving pages. James and Sergey also address practical questions for SEOs, including whether ranking position in a listicle matters, how brand mention frequency influences inclusion, and why intent-based content optimisation is currently the most reliable strategy given that tools to directly reveal fan-out queries do not yet widely exist.
“Instead of optimising a website for a specific keyword, you focus on the problem you are trying to solve for the user. Then you optimise your content to answer all the potential questions that user might have.”
— Sergey Lucktinov
Who Are the Guests on “How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov”?
James Dooley is a well-known figure in the SEO and digital marketing space, recognised for hosting conversations that bridge technical search concepts with practical strategy. In this episode he plays the role of an informed interviewer, asking precise questions that translate complex LLM mechanics into actionable insights for marketers, founders, and content creators.
Sergey Lucktinov is the guest expert and brings detailed knowledge of how large language models process and respond to search queries. He demonstrates a strong command of LLM architecture as it applies to search visibility, covering topics like fan-out query generation, multi-stage content filtering, metadata trust signals, and the semantic weighting systems used by platforms like Google Gemini and ChatGPT. His approach focuses heavily on intent-based optimisation as the primary framework for getting content surfaced in AI-driven search results.
What Are the Key Takeaways From “How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov”?
Here are the key points discussed in this episode:
- The complexity of a search query directly determines how many fan-out queries an LLM generates, with Perplexity producing up to forty and Google typically up to twenty background searches for complex topics.
- Each fan-out query pulls approximately ten candidate results, meaning a twenty-query fan-out can create a pool of up to two hundred websites that the model must then filter down.
- LLMs use a multi-stage filtering process starting with metadata analysis, then structural analysis, and finally full content extraction, with forty to eighty percent of candidates eliminated in the first stage alone.
- For brand or product mentions in list-style results, frequency of appearance across multiple sources is a strong signal of credibility, while for strategy or explanation citations, clarity and structural quality of a single page can be sufficient.
- Because tools to directly observe fan-out queries are not yet widely available, intent-based content optimisation that fully addresses the user's underlying problem is currently the most reliable approach for LLM visibility.
“If a product appears repeatedly across multiple listicles, the LLM recognises it as a strong signal that the product is reputable.”
— Sergey Lucktinov
Is “How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov” Worth Listening To?
This episode is worth listening to because it offers a rare and genuinely technical look inside how LLMs actually process search queries, going well beyond surface-level advice about AI SEO. Sergey Lucktinov explains the fan-out process with concrete numbers and specific platform comparisons, making it possible for listeners to form a real mental model of what is happening behind the scenes when someone searches on Perplexity, Google Gemini, or ChatGPT. The filtering stages he describes, from metadata trust checks using Google's own link data to structural and semantic analysis, give SEOs and content strategists a clear framework for diagnosing why their content might be getting cut before it ever reaches the extraction stage.
What makes this episode particularly valuable is that it moves from theory to practice without losing precision. James Dooley asks exactly the questions a working SEO would ask, including whether you can reverse-engineer fan-out queries and whether position in a listicle affects inclusion odds. The answers are honest about current limitations while still providing actionable direction, particularly the emphasis on solving user intent comprehensively rather than chasing individual keywords. Anyone building content strategies for AI-driven search will find this episode a strong foundation for rethinking how visibility actually works in 2024 and beyond.
Who Should Listen to “How Do You Know What Fan Out Queries Were Searched? James Dooley Interviews Sergey Lucktinov”?
This episode is ideal for:
- SEO professionals who want to understand how LLMs filter and rank content beyond traditional keyword optimisation
- Content strategists and marketers building visibility in AI-driven search platforms like Perplexity, Gemini, and ChatGPT
- Founders and brand managers trying to understand how their products or companies get included or excluded from AI-generated lists and recommendations
- Digital agency owners and consultants who need to explain LLM search behaviour to clients and develop updated optimisation frameworks
Where Can You Listen to James Dooley Podcast?
You can listen to James Dooley Podcast on all major podcast platforms:
- Apple Podcasts – Search for “James Dooley Podcast” in the Podcasts app
- Spotify – Available on Spotify for free
- Amazon Music / Audible – Listen through your Amazon account
- Overcast – For iOS users who prefer a dedicated podcast app
- Pocket Casts – Cross-platform podcast player
You can also subscribe using the RSS feed: https://feeds.transistor.fm/james-dooley-podcast
What Are Listeners Saying About This Episode?
“The breakdown of the two hundred candidate pool and how it gets filtered in stages was genuinely eye-opening. I had no real understanding of how much content gets eliminated at the metadata stage alone before any actual reading happens. This completely changed how I think about technical SEO for AI search.”
“Sergey's explanation of why frequency of brand mentions across listicles signals credibility to an LLM was the most useful thing I've heard in months. It confirmed a strategy I was already testing and gave me the language to explain it to clients properly.”
“Short, dense, and no fluff. The comparison between Perplexity using up to forty fan-out queries versus Google at twenty was exactly the kind of specific detail I was looking for. James asked the right questions and Sergey actually answered them rather than giving vague generalities.”
James Dooley: Hi, today I’m joined with Sergey Lucktinov, and what we’re going to be talking about is query fan-out and how you can understand what query fan-out searches are being performed in the background when you make a search in large language models. Sergey Lucktinov: Hi James, thank you for having me. So when query fan-out happens, we need to consider the context first. What kind of topic are you trying to search for? Is it something simple, like what is the capital of France, or is it more complex, like what is the best massage chair for people under one thousand dollars? Depending on the complexity of the query, the system figures out the intent of the search. It determines what the user is actually trying to get from the LLM, and then it performs either a very small fan-out or a much larger one. Another factor is the model itself. Some models use fewer fan-outs, while others use many more. Perplexity uses the largest fan-out, sometimes up to forty queries. Google typically uses up to twenty. Most other models sit somewhere in the middle. James Dooley: So when Google performs up to twenty query fan-out searches, is there any way of knowing what those fan-out queries actually are? From a keyword research point of view, it would be useful to know what queries are being used so they can be worked into content. Is there a way of reverse-prompting this or seeing it directly? Sergey Lucktinov: There are some limited ways. For example, when you run deep research prompts, you can sometimes see hints of the fan-out process. But the way I usually approach this is from the other end. Instead of optimising a website for a specific keyword, you focus on the problem you are trying to solve for the user. Then you optimise your content to answer all the potential questions that user might have. So rather than analysing individual keywords, you analyse intent and try to fulfil it fully. Over time, I’m confident tools will appear that allow us to see actual fan-out queries more clearly, but right now intent-based optimisation is the most reliable approach. James Dooley: So for anyone watching who might not fully understand this, if someone performs a complex search like best gaming chair under one thousand dollars for elderly users, is the model breaking that into multiple searches such as best gaming chair for elderly people, best gaming chair under one thousand dollars, and similar variations? Sergey Lucktinov: Exactly. Using the massage chair example, the system generates a set of related queries. These might include best massage chair for pain, best massage chair for sleeping, and similar variations. It then fetches results for each of those queries. For each fan-out query, it typically collects around ten results. So if there are twenty fan-out queries, you end up with a pool of around two hundred candidate websites. From there, the analysis begins. In the first stage, the LLM performs a light analysis using metadata such as page titles, meta descriptions, site names, and quality signals. For example, when Google Gemini is used, it pulls trust data from Google itself. That includes link profiles, penalties, and general trustworthiness. ChatGPT performs a similar process but uses different internal quality and relevance signals. After this first pass, anywhere between forty and eighty percent of candidates are removed based purely on metadata. Then a second stage begins, where the system performs light structural analysis. It checks page structure, stability, and clarity. Another large percentage of sites are removed at this stage. Finally, a full content analysis is performed on the remaining candidates, and the model compiles its final answer from that reduced set. James Dooley: That makes sense. So once the pool is narrowed down, does inclusion come down to a consensus score? For example, does being mentioned more often increase your chances? Or does ranking position matter, such as being position one in a listicle versus position ten? Sergey Lucktinov: It depends on the type of query. If you are trying to be mentioned as a brand or product in a list-style result, then yes, frequency of mentions matters. If a product appears repeatedly across multiple listicles, the LLM recognises it as a strong signal that the product is reputable. Different websites also carry different semantic signals. Some sites may have high authority but weaker semantics or performance. Others may have excellent structure and clarity but lower authority. Appearing across a range of sites with different strengths increases the likelihood of inclusion. For content-based inclusion, such as when you want your strategy or explanation cited directly, the situation is slightly different. In that case, clarity, structure, and ease of understanding are extremely important. Publishing the same strategy across multiple platforms helps, but even a single high-quality page with strong structure, speed, and semantic clarity can be extracted by the LLM. So multiple mentions help, but they are not always essential depending on the goal. James Dooley: Perfect. Hopefully anyone watching this now has a clearer understanding of how query fan-out works within LLMs. Sergey, thank you very much for joining me today. Sergey Lucktinov: Thank you. Thanks a lot.
Creators & Guests
Host
James Dooley is a UK entrepreneur.