The attribution problem
When a traveller reads a ChatGPT recommendation mentioning your hotel and then searches your name in Google, visits your website directly, or goes straight to your Booking.com page, none of those actions register as “referred by ChatGPT” in your analytics. The referrer is either Google (branded search), direct, or the booking platform. The AI that triggered the consideration is invisible.
This is a genuine measurement gap, and no analytics tool has fully solved it. Anyone selling you a tool that claims to “track AI-referred bookings” should be asked to explain exactly how it distinguishes AI-driven direct traffic from other direct traffic. The answer is usually that it can’t, and it’s making an attribution model assumption.
Method 1: Direct visibility testing (20 minutes per week, high signal)
Set up a weekly testing routine. Every Monday morning, open ChatGPT, Perplexity and Google AI Overview and ask:
- “[Your category: boutique hotel / family rental / B&B] in [your city or neighbourhood]”
- “[Category] near [your nearest prominent landmark]”
- “[City] hotels under [your approximate price point]”
- One query using your specific distinctive attributes: “hotels with [your key feature] in [your city]”
Record the results in a simple spreadsheet:
| Week | ChatGPT | Perplexity | Google AI Overview | Notes |
|---|---|---|---|---|
| Aug 22 | Not mentioned | #3 of 4 | Not mentioned | Just added llms.txt |
| Aug 29 | Not mentioned | #2 of 4 | Not mentioned |
This is the most honest visibility metric available. Track the trend, not individual snapshots.
Method 2: Direct traffic trend (Google Analytics, monthly)
In Google Analytics 4, look at your direct/none traffic over the past 12 months. Compare it to the months when you implemented AI visibility improvements (adding schema markup, adding llms.txt, improving your GBP).
If direct traffic shows a rising trend beginning 6-8 weeks after implementation, and the new direct sessions show high engagement (longer sessions, more pages visited, direct to contact or booking page), this is consistent with AI-driven consideration visits. It’s not proof. It’s a consistent signal.
Segment your direct traffic by session quality: visitors who come directly, view 3+ pages, and either book or go to your contact page are more likely to have been pre-sold by AI than casual wanderers.
Method 3: Branded search impressions (Google Search Console)
Open Google Search Console and look at the impressions and clicks for your hotel’s branded name terms over time. A rising trend in branded search impressions means more people are searching for your hotel by name.
If you are not running paid advertising with your hotel name and have not done any PR that would explain a branded search spike, rising branded search impressions correlate with increasing mentions of your hotel in AI and editorial contexts.
The honest expectation
Most hotels implementing AI visibility improvements for the first time will see their first measurable signal at 6-10 weeks: they begin appearing in Perplexity results, then gradually in Google AI Overview. ChatGPT changes are slower and less predictable.
The direct revenue signal is harder to isolate for at least 3-6 months of consistent implementation and monitoring. This is not a quick-win channel. It is a structural investment.
See also: Measure AI search visibility · full guide | How long does it take to appear in AI search