By Tessar Napitupulu, Founder & CEO, PT Arfadia Digital Indonesia, GEO pioneer in Indonesia since 2023. More from Tessar.
A tour package page can lose half its organic traffic to an AI Overview while the business behind it books more trips than ever. A dashboard built entirely around sessions and rankings will show that as a failure. It isn't one, and the reason it isn't one is the same reason travel measurement needs a different framework now than it did five years ago: the traveler increasingly gets their answer before they ever click anything, and the metrics built for a click-based internet were never designed to see that.
Why Doesn't Traditional Traffic Measurement Work for Travel Anymore?
Two forces are compounding at once. First, tour bookings are considered purchases, decisions made over days or weeks, often across several sessions and devices, which means last-click attribution was already an incomplete picture of travel demand before AI search existed. Second, AI Overviews and AI chat platforms are now answering a meaningful share of planning and comparison queries directly, which means a growing share of the value an operator's content provides never generates a session at all. A destination guide that gets summarized inside an AI answer, with the traveler's question fully resolved without a click, has still done real marketing work, building awareness and shaping the eventual choice, even though Google Analytics will record exactly nothing for it.
What Should Replace Raw Traffic as the Primary Metric?
Package inquiry submission rate, tracked as the number of qualified inquiries per thousand sessions on itinerary and package pages specifically, not the site as a whole. This single change in what gets reported first fixes the most common measurement failure in travel marketing: treating a ten-thousand-visit inspiration article and a three-hundred-visit package page as comparable performance, when the inspiration article may generate a handful of inquiries and the package page may generate dozens. Inquiry rate, not visit count, is the metric that actually tracks toward revenue.
What to Report First, Second, and Last
The order that keeps a travel marketing report honest about what actually drives revenue.
Package inquiry rate
Per thousand sessions, on itinerary and package pages specifically.
Inquiry-to-booking close rate
Tracked separately, since a rising inquiry rate with a falling close rate signals a sales-process problem, not a marketing one.
AI citation frequency
Against a fixed prompt panel, re-tested weekly given documented citation drift.
Assisted conversions & branded search
Leading indicators for AI-influenced demand that last-click attribution misses.
Raw traffic and keyword rankings
Still worth tracking, but reported last, as context rather than the headline.
Created by Arfadia • arfadia.com/blog
Should Every Page Type Be Measured Against the Same KPI?
No, and using one KPI across every page type is a large part of why travel measurement breaks down. Each of the four stages a traveler moves through, inspiration, itinerary research, comparison, and booking, produces content with a genuinely different job, and each job has a different honest success metric.
| Funnel stage | Content type | Honest success metric |
|---|---|---|
| Inspiration | Destination guides | Assisted conversions, internal click-through into the funnel |
| Itinerary research | Itinerary and planning articles | AI citation frequency, scroll depth, time on page |
| Comparison | Private vs. open trip, operator vs. DIY content | Click-through to package pages, time to decision |
| Booking | Package pages | Inquiry submission rate, inquiry-to-booking close rate |
Applying "inquiry rate" to a destination-inspiration article, or applying "raw traffic" to a package page, both produce numbers that look meaningful but measure the wrong thing for that page's actual job. Segmenting the report by funnel stage, rather than reporting one blended average across the whole site, is what keeps each metric honest.
What Are the Practical Tooling Limitations Worth Knowing About?
Google Search Console does not currently isolate AI Overview appearances or clicks as a distinct, filterable row the way it does for standard organic results, which means a page's AI-citation performance has to be tracked outside of Search Console entirely, through a dedicated prompt-panel process or a third-party AI-citation monitoring tool. GA4's default event tracking does not distinguish an AI-referred visitor from any other direct or referral visitor unless UTM parameters or a self-reported form field are deliberately added, so that instrumentation has to be built rather than assumed to already exist. Neither limitation is a reason to skip AI-visibility measurement. Both are reasons to expect that measuring it will always look more manual and proxy-based than measuring classic organic search did, at least until platform-level reporting catches up with how much query volume now resolves without a click.
How Do You Measure Something That Never Generates a Click?
With proxies, deliberately chosen and consistently tracked, rather than a single clean number, because a single clean number for AI-mediated influence does not currently exist with standard analytics tools. Three proxies do the practical work. AI citation frequency, tracked against a fixed panel of 15 to 25 representative prompts run across ChatGPT, Perplexity, Gemini, and Google AI Overviews, shows whether the operator is even in the running for the queries that matter, and needs re-testing weekly rather than monthly given documented citation drift of roughly 40% for Perplexity and 59% for Google AI Overviews month to month. Branded search volume, tracked over time, tends to rise when AI-mediated awareness is working even when direct traffic to the site does not, since a traveler who first encountered a brand inside an AI answer often searches the brand name directly afterward rather than clicking through immediately. And a simple "how did you hear about us" field on every inquiry form, with an explicit AI-assistant option rather than just "search engine" or "social media," captures self-reported attribution a pixel-based system cannot.
Why Mentions and Citations Need to Be Counted Separately
A mention names a brand inside an AI-generated answer. A citation attaches a clickable, attributed source link to that mention. The two are not interchangeable for measurement purposes: a brand can be mentioned frequently without ever being cited, which builds awareness but drives no direct traffic at all, or cited rarely but with high-value placement on exactly the query that matters most for conversion. Reporting a single combined "AI visibility" number without separating these two collapses a meaningful strategic distinction into a vanity metric. A report that tracks them separately can tell an operator whether a content gap is an awareness problem or a technical citation-eligibility problem, which call for entirely different fixes.
| Signal | What it tells you | What it can't tell you |
|---|---|---|
| AI mention (unlinked) | Brand awareness is building inside AI answers | Whether that awareness converts to a visit or a booking |
| AI citation (linked) | The operator is a source the AI trusts enough to attribute | Whether the traveler actually clicks through |
| Branded search lift | Awareness is translating into intent to seek the brand directly | Which specific content or platform drove it |
| Inquiry-form "how did you hear" field | Direct, self-reported attribution, including AI sources | Full-funnel accuracy; travelers misremember or skip the field |
How Does the RoGEO Framework Fit Into This?
RoGEO, Return on Generative Engine Optimization, structures exactly this set of proxies into one reporting framework rather than leaving each metric to be tracked ad hoc: citation frequency (how often and how prominently an operator's content is cited across the platforms that matter), reference depth (whether that citation is a passing mention or a primary recommendation the AI is building its answer around), and revenue attribution (connecting citation presence, branded search movement, and self-reported inquiry attribution back to actual bookings, imperfectly but consistently, over time). None of the three components alone tells a complete story. Reported together, on a fixed cadence, they give a travel operator the closest available approximation to "is our AI visibility actually working" without pretending a precision that current analytics tools cannot deliver.
Three Components, Reported Together
No single component tells a complete story on its own. Reported as a set, they approximate GEO ROI without overstating precision.
Citation Frequency
How often and how prominently the operator is cited across the platforms that matter.
Reference Depth
Whether the citation is a passing mention or a primary recommendation.
Revenue Attribution
Citation, branded search, and self-reported inquiries connected back to actual bookings.
Created by Arfadia • arfadia.com/blog
What Does a Realistic Monthly Reporting Structure Actually Look Like?
A workable monthly report for a travel operator leads with package inquiry rate and inquiry-to-booking close rate, broken down by destination cluster, since aggregate site-wide numbers hide which specific clusters are actually working. It follows with AI citation frequency against the fixed prompt panel, broken down by platform, since Perplexity, Gemini, and Google AI Overviews behave differently enough that an aggregate score obscures which platform needs attention. It includes branded search trend and self-reported AI-source inquiries as supporting context. And it reports raw traffic and keyword rankings last, framed explicitly as context rather than the headline metric, which is the single structural change that prevents a shrinking organic-click environment from reading as a failing business when the business itself may be doing better than ever.
This measurement discipline pairs directly with the long-tail content strategy in our piece on why OTAs own generic search but not the booking, and the citation-building work described in our GEO for travel and tour operators service.
Frequently Asked Questions
Should we stop tracking traffic and keyword rankings altogether?
No, keep tracking them, just stop leading with them. They remain useful context for diagnosing why inquiry rate or citation frequency moved, they should no longer be the first or only metric a report opens with.
How often should AI citation frequency actually be measured?
Weekly, given documented monthly citation drift in the 40 to 59% range depending on the platform. A monthly test will regularly mistake normal drift for a real change in performance.
Our CEO wants a single ROI number for our GEO investment. What do we tell them?
That a single, fully attributed ROI figure is not currently achievable with standard analytics for AI-mediated awareness, and that the honest alternative is a small set of proxies, citation frequency, branded search movement, and self-reported inquiry attribution, reported consistently over time rather than compressed into one number that would overstate the precision actually available.
Is inquiry rate a fair metric for every type of content, including inspiration articles?
No. Apply it to itinerary and package pages, where a direct inquiry is the natural next action. Measure inspiration content through assisted conversions and internal click-through into the funnel instead, since expecting a direct inquiry from a top-of-funnel article sets the wrong bar for what that content is supposed to do.
What's the single easiest change to make this quarter?
Add an explicit AI-assistant option to the "how did you hear about us" field on every inquiry form. It costs nothing to implement and immediately starts capturing a signal most operators are currently missing entirely.
Do we need special software to track AI citation frequency, or can we do it manually?
A manual prompt panel, run consistently by a real person against a fixed query list, works and is how most of the evidence cited in this article was originally gathered. Dedicated monitoring tools save time at scale but are not a prerequisite for starting; consistency of the panel and the testing cadence matters more than the tooling.
How do we explain a drop in Google Search Console traffic to a client or stakeholder who only looks at that one dashboard?
Show the inquiry rate and close rate trend alongside it, on the same page, for the same period. A traffic drop next to a stable or rising inquiry rate tells a very different story than a traffic drop next to a falling inquiry rate, and most stakeholders update their read of the situation once they see both numbers together rather than traffic in isolation.
Sources & References:
- BrightEdge, distinction between AI mentions and AI citations, and average citation count per AI Overview answer.
- Search.Agency, "Travel SEO and AI Overviews in 2026, a 105-Query Study," 3 to 4 July 2026, for citation-drift and measurement-methodology context.
- HyperMind, citation-drift measurement across Perplexity and Google AI Overviews, referenced for the 40% and 59% monthly drift figures.
- CausalFunnel, travel industry lead conversion rate benchmarks, for inquiry-to-booking close rate context.
- Arfadia, AI Citation Rate Report 2026, arfadia.com/resources, for the RoGEO framework definition (citation frequency, reference depth, revenue attribution).