The recommendation knew too much

The hotel recommendation appeared useful: a quieter room, a shorter route, an amenity aligned with the trip. The guest had not been told which behavior produced it, how long the inference would remain, or who else might receive it.

Relevance had crossed into a second decision. The customer was no longer choosing only a hotel. The customer was being asked—sometimes invisibly—to accept the data relationship behind the choice.

Privacy began with purpose

NIST's Privacy Framework gave organizations a voluntary structure for identifying and managing privacy risk. The FTC's guidance on protecting personal information added an operational discipline: know what is collected, keep only what is needed, protect it, and dispose of it responsibly.

For hotel research, purpose came first. A guest could volunteer that an accessible route, adjoining room, late arrival, or quiet location mattered. The property could use that context to organize relevant evidence without treating every click as permission to build a lasting profile.

Choice needed to remain a choice

The FTC's 2022 dark-patterns report documented interfaces that can manipulate or obscure consumer choice. Personalization should not depend on a preselected consent, a confusing refusal path, or a penalty for returning to an unpersonalized experience.

It should be explainable and reversible. Inferences deserved stronger scrutiny than facts a customer supplied directly. Shared planning also complicated ownership: one traveler could not silently consent for every family member or colleague reviewing the trip.

Accessibility preferences could improve relevance, but they were not an invitation to profile disability. W3C personalization guidance supported adaptable content while the property still needed privacy restraint and an accessible default path.

The customer controlled the context

SiteSee can organize property evidence around preferences the customer chooses to share and preserve that context for an authorized human conversation. The operating system can govern access and retention; it should not transform ambiguous browsing into an unsupported conclusion about a person.

That boundary applies the earlier sales-evidence rule: behavior can improve service when it is interpreted cautiously and connected to a legitimate customer task.

Relevance became an act of trust

The quieter-room recommendation could still appear. This time the guest could see why, change the preference, remove it, or ask a person to help.

The result felt more personal because the customer retained agency. Trust was not the price paid for relevance. It was one of the outcomes good personalization had to produce.