Basecamp Advertising’s own Search Console property recorded 21,100 impressions and 136 clicks in the 28 days to 20 September 2026, a 0.6 percent click-through rate across 675 queries and 247 pages. A large share of those impressions were never seen by a person.
Some of the query strings in that report read like this: “evaluate the ota company viator on does viator charge booking fees?” at 101 impressions, and “which safari companies in luxury safaris have the most effective advertising?” at 84. Another arrived as “my main motivations: kid-friendly. Does viator charge booking fees?” at 38 impressions. Nobody types those into a search box.
The query fan-out tour operators are now seeing in Search Console is a documented Google technique rather than a reporting glitch. It inflates impressions, drags site-wide click-through rate down, and leaves an owner reading a number that no longer describes demand.
What is query fan-out, and why does it matter to tour operators?
It is Google issuing several searches of its own from one person’s question. Google Search Central, in its guide to AI features and your website, states that both AI Overviews and AI Mode may use a query fan-out technique, issuing multiple related searches across subtopics and data sources to develop a response.
Search Console Help describes the same mechanism from the reporting side, saying that AI Mode groups the user’s question into subtopics and searches for each one simultaneously. One prompt can therefore produce a handful of machine-written searches, and a single page can collect an impression in several of them.
For an experience operator that has a specific consequence. The questions a planning assistant decomposes are exactly the ones your business answers: what a tour costs, what the booking fee is, whether it suits children, which operators run it. Your pages get pulled into the answer, but the person asking never arrives as a click.
Why did your click-through rate fall without anything breaking?
Because impressions grew in a place that rarely produces clicks. Search Console Help documents how impressions, position and clicks are counted, and for AI Mode it states that standard impression rules apply, that a click is counted when someone clicks a link to an external page, and that position follows the same methodology as a normal results page.
Two further rules in that documentation explain the volume. A follow-up question inside AI Mode is treated as a new query, so one conversation can appear as several searches. Google Search Central adds that sites appearing in AI features are included in overall search traffic and are reported inside the Web search type, which means there is no separate bucket to exclude.
On this site the effect is visible at page level. The strongest page, an explainer on Viator booking fees, carried 4,741 impressions and 14 clicks over those 28 days, against a site average position of 10.5. A click-through rate calculated across that mix averages two unrelated things, so treating it as a single health metric will produce the wrong decision.
How do you separate fan-out noise from real booking demand?
Read the query strings rather than the totals. Machine sub-queries have a recognisable shape: several clauses, instructional verbs such as evaluate, identify or compare, stated traveller motivations, and often your own page title repeated back inside a longer sentence.
Real booking demand looks different. It is short, it usually carries a place or a date, and it converts. Sort the last 28 days by impressions, read the top 20 strings by hand, and split them into those two groups before you change a campaign.
Then cross-check in Google Analytics. Analytics Help defines the AI Assistant default channel as traffic where the medium exactly matches ai-assistant, which is set when the referrer matches Google’s list of AI assistants. Over the most recent seven days this site recorded 86 direct sessions, 63 from organic search, 7 from AI Assistant and 2 referral, with chatgpt.com at 4, perplexity.ai at 2 and claude.ai at 1. Those are small numbers, but they are people who arrived, which is more than most of those impressions represent.
What Search Console will not tell you
It will not label which impressions came from an AI feature. Because AI Overviews and AI Mode data sits inside the Web search type, there is no filter that isolates it, so the split has to be inferred from the query strings themselves.
The same documentation notes that Search Console does not include data from experiments in Search Labs, and that position is recorded as the topmost position your page held, averaged across queries. A page cited prominently in one answer and buried on page two of another averages out to a figure that describes neither.
Treat the report as a visibility signal with a known distortion, and stop using it to judge whether pages earn revenue. That judgment belongs to the booking data, which is where the more serious problem usually sits.
Which numbers should an operator act on?
Tracked bookings, revenue by channel, and enquiry submissions, in that order. Impressions and average position are worth watching for direction, not for decisions, and neither belongs in a monthly report as a headline.
None of it is readable until the booking flow is measured properly. When checkout runs on a booking platform domain or inside an iframe, conversions get credited to the platform rather than to the campaign that produced them, which is the failure pattern described in Google Ads conversion tracking for tour operators. Fix that first, then judge channels using the approach in multi-channel attribution for experience operators.
A workable monthly set for an operator: tracked bookings and revenue by channel, enquiry form submissions, the share of bookings that arrived through a platform rather than direct, and clicks from search. Four numbers, all of which move for reasons you can act on.
What should you publish when the machine is doing the asking?
Pages that answer the compound question plainly in the first two sentences. A fan-out sub-query is usually specific and comparative, so the page that gets cited is the one stating the fee, the policy, the season or the limit in text, with the source named.
Google Search Central is explicit that no new technical work is required: there are no additional requirements to appear in AI Overviews or AI Mode, no machine readable files or AI text files to create, and no special structured data to add. The same guidance says a page must be indexed and eligible to be shown with a snippet, which makes indexing hygiene the only real prerequisite.
In practice that means writing the operator-facing facts down rather than implying them. Prices with what they include, cancellation and rebooking rules, age and weight limits, season windows, what happens in bad weather, and the platform fees a guest will see at checkout. That is the same material covered in AI Overviews for tour operators, and it decides whether the itinerary builders described in AI travel agents, OTAs and tour operators can quote you accurately.
Start with your own report. Pull the last 28 days in Search Console, sort by impressions, read the top 20 query strings, and mark which are machine-written. Then compare the pages earning those impressions against the enquiries and bookings the same pages produced in Analytics. If the gap is wide, the work is on the booking measurement rather than on the content.
About the author
Andrew Compton is the founder of Basecamp Advertising, which works exclusively with adventure tourism and experience operators. He spent 8+ years in adventure tourism across New Zealand and the United States, guiding at Rotorua Canopy Tours and on the French Broad River in Asheville, and serving as assistant manager at a high-volume zipline operation in Texas. He has lived and worked across New Zealand, Australia, and the UK, and now works with operators across North America, Europe, Latin America, and Australasia on growing direct bookings, reducing OTA dependence, and getting a clear picture of what their marketing actually produces.
















