Studying the "adventure holidays" potential Query Fan-Out

To better understand how Google’s AI Mode expands a query like “adventure holidays for families”.

I compiled and analyzed multiple sets of “fan-out” query data... That is, the related brands, websites, and topics surfaced in AI-generated responses.

These sets came from five different pulls, each representing a snapshot of what Google’s generative systems consider relevant to the original seed query.

By comparing the overlap between these lists, examining the types of entities included, and identifying the most frequently mentioned brands.

I wanted to see the variability in responses, both directly from AI Mode and the potential hallucinated responses when I asked for JSON files on the Query Fan-Out data.

Differences in the data

One of the clearest takeaways from comparing the five fan-out sets is the variation in brand and content representation across each list.

Despite all lists being generated in response to the same seed query “adventure holidays for families” they show moderate to low overlap, with similarity scores ranging from only 7.8% to 15.6% between any two sets. This variation is informative in itself.

Some lists lean heavily on tour operators (e.g., Intrepid, G Adventures, Backroads), while others surface a broader mix of brands.

• Small blogs and family influencers (e.g., The Mom Trotter, Crazy Family Adventure)
• Legacy publishers or aggregators (e.g., Frommers, Tripadvisor, Reddit)
• Niche regional DMCs and long-tail operators (e.g., Say Hueque, Exotic Voyages, Greece Insiders)

This brand diversity suggests that the source corpus used in AI Mode queries may be shifting, or influenced by different user profiles, query rewrites, or regional settings.

Overall, 70 different travel brands, blogs, and Reddit were mentioned. Even though some were mentioned more than others, there isn't a really big standout mention leader in the dataset.

• backroads, 3
• intrepid travel, 3
• ke adventure travel, 3
• exodus adventure travels, 2
• adventure life, 2
• the mom trotter, 2
• travel babbo, 2
• trafalgar, 2
• crazy family adventure, 2
• earth trekkers, 2
• g adventures, 2
• thomson family adventures, 2
• road scholar, 2
• austin adventures, 1
• the traveling child, 1
• world of travels with kids, 1
• global basecamps, 1
• collette, 1
• vacation kids, 1
• say hueque, 1
• full suitcase, 1
• croisieurope, 1
• outdoor families magazine, 1
• the travel mom, 1
• sky vacations, 1
• family vacation critic, 1
• oars, 1
• oars whitewater rafting, 1
• mommy poppins, 1
• brendan vacations, 1
• exotic voyages, 1
• greeking.me, 1
• ciao bambino!, 1
• the adventure people, 1
• lights on africa safaris, 1
• wandermust family, 1
• pack up + go, 1
• greece insiders, 1
• 2 travel dads, 1
• reddit, 1
• learning escapes, 1
• charlie the traveler, 1
• varanasi excursion, 1
• family travel forum, 1
• tripadvisor, 1
• wayfairer travel, 1
• travel with kids, 1
• exodus travels, 1
• explore worldwide, 1
• responsible vacation, 1
• family adventures in the algarve, 1
• family thailand adventure, 1
• family fun journeys, 1
• thrifty family travels, 1
• family iceland adventure, 1
• family adventure holidays, 1
• acorn family holidays, 1
• across south america, 1
• adventure alternative, 1
• realistic asia, 1
• responsible travel, 1
• odynovo, 1
• the family adventure company, 1
• frommers, 1
• abercrombie & kent, 1
• tauck, 1
• costco travel, 1
• explore!, 1
• national geographic family journeys, 1
• global family travels, 1
• pitstops for kids, 1

TF-IDF*

Words like “family,” “adventure,” “holidays,” “travel,” “kids,” and “destinations” dominate the visualization, pointing to consistent thematic anchors. Alongside these core terms, more nuanced phrases such as “rafting,” “safari,” “multi-generational,” “Costa Rica,” and “teenagers” hint at the types of experiences and traveler profiles AI systems are likely optimizing for.

`This distribution reinforces the idea that Google is expanding the query not just by topic, but by lifestyle, age group, and activity type, and reflecting diverse user intents and content verticals.`

A variety of entities

Some lists are tightly structured around entities with clear websites and distinct product offerings, while others blend these elements.

• Generic concepts like “Family Iceland Adventure” (not a clear brand)
• Media channels like Travel With Kids (TV) or Reddit threads
• Placeholder-style or undefined mentions (e.g., Thrifty Family Travels or Family Fun Journeys)

This inconsistency implies that not all fan-out results are equally well-grounded in verified entities.

Some may be soft matches inferred from content themes rather than exact brands.

Content format bias?

There’s a notable divide between lists that prioritize:

• Commercial, structured travel offerings (Backroads, KE Adventure Travel)
• Experience-driven, first-person storytelling (Earth Trekkers, Travel Babbo)

This reinforces the idea that AI Mode is not purely commercial in its expansion of the query, but instead pulls in informational, editorial, and community-driven content, depending on the context.

User context shaping

Even within a single session, differences across fan-out pulls could reflect changing LLM responses, divergent grounding sources, or user context shaping. A single concept like "family adventure travel" is not treated statically as it drifts toward themes like:

• Destination-focused journeys (Costa Rica, Galapagos, Borneo)
• Age-specific needs (toddlers vs. teenagers)
• Style of travel (rafting, luxury, multi-generational, educational)

This reflects the non-deterministic, intent-driven nature of generative search and how it is assembling a constellation of possibilities rather than a fixed result set.

How can we use this data?

Even though the data across sessions varied significantly, sometimes dramatically, for the same core query, it doesn’t mean the exercise is pointless. In fact, the variation itself is part of the insight. It shows that Google’s AI Mode is not fixed to one index or interpretation. Instead, it dynamically explores different angles of intent, entity grounding, and user context with each generation.

That kind of variation can be frustrating if you're looking for consistent answers. It also demonstrates how Google is testing and refining what might be helpful to users based on various scenarios.

For brands, SEOs, and content strategists, this gives us something useful to work with. It tells us that showing up in different content formats, whether commercial, editorial, or community-based, and in niche or long-tail expressions of user need, is increasingly important.

Whether you're focused on traditional SEO or building visibility within LLM-generated platforms, this type of query expansion gives us a look into what the system sees as relevant, not just in terms of keywords, but in intent, tone, and usefulness. The more we understand these signals, the better we can shape strategies that match how users are served in this AI-led search environment.