What AI can already tell about the visitor browsing your hotel's website

Lighthouse Direct has been using predictive AI to read hotel website visitors and act on what it learns since 2018. For years now, this technology has been scoring real visitors on a live network of hotels, learning from hundreds of millions of sessions, and quietly deciding which guest needed an incentive to book and which one just needed the right nudge at the right moment. Underneath that scoring sit five algorithms, each answering a different question about who's actually on the site.

Here's what each one does, and what it means for your direct channel.

The conversion problem this was built to solve

On average, 98% of hotel website visitors leave without booking. That's not a traffic problem, it's a conversion problem, and most hotel websites still treat every visitor the same way: same homepage, same offer, same generic urgency banner, regardless of whether that visitor is a repeat guest three clicks from booking or a first-time browser three months out from a trip they haven't committed to.

Blanket discounting doesn't fix this. It just as easily hands a promo code to someone who was already going to book at full rate, quietly eroding average daily rate (ADR) for no reason. Predictive AI targeting exists to make that distinction, visitor by visitor, in real time.

What predictive AI targeting actually does

Beyond broad segments like "returning visitor" or "mobile user," Lighthouse Direct's predictive AI also evaluates more than 400 behavioral and contextual signals per visitor, including where they came from, how they're browsing and what they've searched, and compares all of it to hundreds of millions of past sessions across a network of 80,000-plus hotels.

Five algorithms sit under this technology. Each one answers a different commercial question.

Intent: is this visitor close to booking?

The intent algorithm predicts how likely a visitor is to complete a booking on this visit. It's the foundation the other four build on, because everything else, whether to offer a discount, an upsell or nothing at all, depends first on knowing how close someone already is to converting.

Why it matters: a visitor with high booking intent doesn't need an incentive. Showing them one anyway is money left on the table. A visitor with low intent might need a reason to stay on the page at all. Getting this distinction right protects rate integrity while still giving hesitant visitors a reason to convert.

Spend: what is your website visitor's budget?

The spend algorithm predicts how likely a visitor is to opt for a higher or lower rate on the site, in other words, which room category they're actually likely to book.

Why it matters: this is where average booking value (ABV) gets protected instead of guessed at. A visitor already leaning toward a superior room doesn't need a discount, just reassurance and a reason to book now, whether that's a price widget or a highlight of what makes the room worth it. A visitor leaning toward a standard room is the one worth nudging with a package or an upgrade offer, since that's where the incremental ABV actually comes from.

Flexibility in destination: how open is this visitor to booking elsewhere

The destination flexibility algorithm predicts how likely a visitor is to book a different destination than the one they're currently searching.

Why it matters: the right move depends on what kind of property is asking. A single independent hotel wants to reassure a destination-flexible visitor and keep them from wandering, promoting the location itself, local activities, and what makes the destination worth choosing. A hotel group or brand with properties in multiple destinations can instead redirect that same visitor to a sister property elsewhere in the portfolio, turning flexibility into a cross-sell instead of a loss.

Flexibility in dates: how movable is this trip

The dates algorithm predicts a visitor's willingness to shift their travel dates. Paired with occupancy data, this becomes a tool for filling specific low-demand periods rather than discounting broadly across the calendar.

Why it matters: instead of running a sitewide sale to fill a soft week, marketing and revenue teams can surface a date-shift incentive only to the visitors who were already flexible enough to consider it, which is a much cheaper way to move the same room nights.

Length of stay: will this guest extend their trip

The length of stay algorithm predicts whether a visitor is likely to extend their stay beyond what they originally searched for.

For visitors who are less likely to extend their stay dates beyond their search, this is the moment for an incentive: book three nights, pay for two, or a complimentary extra night with breakfast included. For visitors already predicted to extend on their own, the smarter move is a premium upsell instead of a discount they didn't need to convert.

Why it matters: length of stay has always been one of the biggest levers on RevPAR (revenue per available room), and until now it's been managed with blunt instruments, like minimum-stay restrictions or broad multi-night promotions. A predictive signal lets hotels apply the right nudge to the right visitor instead of the same offer to everyone.

What this looks like once it's running

The five algorithms aren't theoretical. THE THIEF, a hotel in Oslo, ran an A/B test using the intent algorithm to show a targeted secret-sale offer only to visitors with a 0 to 40% probability of booking. The result: a 25% conversion uplift among those low-intent visitors, €26,000 saved in promotional spend by no longer discounting visitors who didn't need it, and €30,545 in revenue specifically from low-intent visitors who converted on the offer.

That didn't come from showing every visitor the same offer. It came from knowing, visitor by visitor, who actually needed one.

Five questions, not one blanket answer

Intent tells you if someone's close. Spend tells you what they're worth. Flexibility in destination and dates tells you how movable their plans are. Length of stay tells you whether to incentivize a longer trip or upsell a stay that's already extending on its own. Which signal matters most depends on the situation: a soft midweek period calls for the dates signal, a rate-integrity concern calls for intent, a portfolio of premium rooms calls for spend. Each one lets a hotel's website move past one-size-fits-all messaging and respond to whatever question is actually in play.

That kind of targeted, signal-by-signal response is what makes predictive AI inside Lighthouse Direct less of a feature and more of the foundation the rest of the direct channel strategy sits on.