Compare Public Sources for Agreement

Prerequisites: Before this lecture, you should be able to save an AI answer record from Lecture 2 and read the answer through name, place, service, and source from Lecture 4. You should also be able to check name variants from Lecture 5, place signals from Lecture 6, and service category from Lecture 7.

A clinic manager opens five tabs before lunch: the clinic website, the map profile, a booking platform, a medical tourism directory, and a review site. The website says the clinic is a general dental practice with cosmetic and restorative services. The map profile calls it a dentist. The booking platform lists whitening, veneers, implants, dentures, and braces in one long menu. The old directory says “international cosmetic dental center,” which sounds impressive until somebody notices it still uses a branch description the clinic stopped using.

Then an assistant answer appears. It does not quote any one tab exactly. It says the clinic is a “cosmetic dental clinic in Phuket with implant services for international patients.” The sentence feels assembled from small pieces lying around the public web. Nobody intended that exact sentence. Still, it is now what a patient may read before deciding whether to call.

The answer is only the visible edge

By Lecture 8, the course changes direction. Earlier lessons trained you to mark what the answer says: the name used, the place assigned, the service inferred, and the source borrowed. Now we widen the desk. The answer remains the starting point, but the work becomes comparison across the public places where clinic evidence appears.

Public evidence is often untidy. A Thai dental clinic may have a careful Thai contact page, a simplified English treatment page, a map profile edited by different staff, a booking platform with template categories, social posts written quickly, patient reviews, and directory entries copied from older material. Each surface may be partly true. The problem starts when they point in slightly different directions.

Source alignment is agreement across public surfaces on name, place, category, treatments, branch, and current status. That does not mean every surface must use identical wording. A Thai legal-name block and an English patient-facing page naturally do different work. Alignment means that a patient, and also an assistant, can move across those surfaces and still meet the same clinic identity.

Source conflict is public sources disagree about name, place, category, treatments, branch, or current status. Conflict can be obvious, like two different districts. It can also be quiet: one profile calls the clinic cosmetic-first, another presents it as general, and a booking platform lists procedures the current site treats more carefully. Quiet conflicts often travel into AI answers because they look harmless one at a time.

Build the table before interpreting

Do not begin by deciding which source is right. That pulls you into argument too early. Start by making a small source table for one answer record. Across the top, write the four patient-answer readings: name used, place assigned, service inferred, and source borrowed. Down the side, write the surfaces you can inspect: clinic website, Thai page, English page, map listing, directory profile, booking platform, review site, social profile.

Then fill the table with phrases, not opinions. If the map uses a shortened English name, copy it. If the directory says “cosmetic dental center,” copy it. If the English site says “dental clinic in Phuket,” copy that. If the Thai page gives a district or branch wording, copy the visible wording rather than replacing it with a cleaner internal version. The table should feel clerical. That is useful.

Object B is the right composite scenario for this step. It is a Phuket dental clinic serving local patients, tourists, and returning foreign patients. Its website gives current service wording, the booking platform keeps a broad treatment menu, reviews emphasize cosmetic results, and an older medical tourism directory still presents the clinic in travel-friendly language. The assistant answer does not need to invent much. It can pick one phrase from each surface and produce a description that sounds coherent but overstates the clinic’s current public role.

That is why the table matters. Without it, the clinic may say, “The assistant is wrong.” With it, the clinic can say something more useful: “The assistant’s category probably comes from the directory and reviews; the current website does not support that emphasis clearly enough.” The second sentence gives the repair somewhere to begin.

Agreement is a hierarchy, not a word match

When public surfaces agree, they should agree at the level of hierarchy, not only vocabulary. A clinic can use “dentist,” “dental clinic,” “general dental care,” and “family dentistry” across different places without conflict if the role remains stable. Another clinic can use different Thai and English phrasing and still be aligned if the name, place, service scope, and branch facts point to the same reality.

Hierarchy is where many dental clinics lose control. The clinic’s own website may present general dental care first, then cosmetic services, then restorative treatments. The booking platform may flatten everything into a treatment catalogue. Reviews may make whitening louder because patients write about visible results. A directory may use “cosmetic dentistry” because it fits medical tourism search habits. None of those surfaces has to be malicious. Together, they can move the service inferred away from the clinic’s intended service category.

Ask three hierarchy questions. What is the clinic mainly presented as? Which services are available but not defining? Which claims are conditional, such as doctor availability, consultation, branch, language support, or referral? If those levels disappear on outside surfaces, the assistant may turn a menu item into an identity.

The same applies to place. A clinic may be in Phuket province, in a specific local area, with one branch and a named access point. A tourist-facing directory may reduce this to “Phuket.” A review may mention a hotel area. A map may show the pin correctly but use a shortened place label. If one surface wrongly says “Phuket Town” while another points elsewhere, the assistant has a conflict, and patients receive the blur.

Outside sources have different shapes

A clinic-owned page, a map profile, a booking platform, a directory, a review source, and a social profile do not carry the same kind of evidence. The course is not asking you to rank them with a score. It is asking you to notice what each surface is good at and where it tends to distort.

The clinic website should usually carry the strongest current explanation of name, location, category, services, doctors, and limits. That does not mean the assistant will always prefer it. It means the clinic has the most control there. If the website is thin, outside sources become louder than they deserve.

A map listing is strong for practical identity: name, address, category, opening hours, phone, route, and sometimes reviews. It is weaker for service nuance. A map category may say “dentist” or something similarly broad. That can be fine. But if the listing name, address, or branch wording differs from the site, the conflict affects the patient before any treatment explanation begins.

Directories and booking platforms are useful but template-shaped. They often want clean categories, long treatment menus, tourist-friendly language, and comparison-ready profiles. That structure can help discovery, especially for foreign patients. It can also pull the clinic toward the platform’s categories rather than the clinic’s own care role.

Reviews are evidence of patient experience, but they are not clinic-owned explanation. A review may mention whitening, pain relief, hotel convenience, a friendly doctor, or a price surprise. These details can influence an assistant because they are vivid. Yet vivid does not mean representative. For this lecture, mark when review wording appears to be stronger than the clinic’s own service evidence.

Name the conflict before writing the correction

A source conflict should be named before it is repaired. Otherwise the clinic may make pleasant edits that do not address the answer risk. I usually mark conflicts in plain language: name conflict, place conflict, category conflict, treatment conflict, branch conflict, current-status conflict. The labels are simple because the situation is already complicated enough.

A name conflict might be a Thai legal name on one page, an English trade name on another, and an old transliteration in a directory with no published bridge between them. A place conflict might be a map pin in one area, a directory saying a broader tourist zone, and an assistant answer assigning a more specific district that the clinic never uses. A category conflict might be the website saying general dentistry while a booking platform and reviews push cosmetic dentistry to the front.

Treatment conflict is especially common. The clinic may once have listed implants, braces, or a surgical service because a directory profile expected a full menu. Later the clinic’s own site becomes more careful, perhaps because doctor availability changed or the clinic wants consultation language. The old surface remains. An assistant sees both. The answer may choose the stronger-sounding claim.

The correction should follow the risk. If the conflict affects name or location, it usually deserves fast attention because patients may identify the wrong clinic or travel toward the wrong area. If the conflict affects category or treatment scope, it may create wrong inquiries or trust problems. If the conflict sits in a low-attention old profile, it may still matter when that profile supplies the exact phrase the assistant repeats. Convenience is a poor priority system. Answer risk is better.

A useful alignment note can be written in five lines: answer claim, strongest supporting surface, conflicting surface, likely risk, first correction. For example: “Answer says cosmetic dental clinic. Supporting surface: reviews and old directory. Conflicting surface: current website says general and restorative care. Risk: foreign patients expect cosmetic-first clinic. First correction: revise English service overview and request directory update.” That note is small enough to use again in a monthly review.

What to remember

• Source alignment is agreement across public surfaces on name, place, category, treatments, branch, and current status. In this lecture, alignment means that the clinic’s site, map listing, booking profiles, directories, reviews, and social pages do not teach competing versions of the clinic.

• Source conflict is public sources disagree about name, place, category, treatments, branch, or current status. A conflict may be loud, like a wrong district, or quiet, like an old treatment label that still appears in a directory.

• Compare public surfaces before writing corrections. A clinic can only fix the visible evidence path once it knows which surface supplied, amplified, or contradicted the AI answer claim.

• Outside sources are useful but shaped by their format. Map listings, booking platforms, directories, reviews, and social pages each simplify the clinic in different ways.

• The safest repair is not to make every source identical. It is to make their hierarchy agree: same clinic identity, same practical place, same care role, same current limits.

• The four patient-answer readings are: name used, place assigned, service inferred, and source borrowed, because a clinic becomes trustworthy to AI only when those four claims point to the same public evidence.