Compare against the competitor ChatGPT favours

Before this lecture, you should know how entity clarity and the nearest stronger neighbour work from lecture 7, how bilingual consistency can fail from lecture 8, how authority signals and directory noise shape confidence from lecture 9, how source-level fixes are planned from lecture 10, and how naming accuracy is tracked from lecture 13. We are now using the measurement set not just to watch the target firm, but to ask why another firm keeps getting the clearer sentence.

Composite scenario. A Brussels mobility-law practice appears in the third paragraph of a ChatGPT answer, after a larger competitor with a sharper English service page and a cleaner French directory profile. The answer is not insulting. It even names the boutique firm correctly. But the larger practice gets “Belgian immigration and international mobility lawyers in Brussels,” while the boutique gets “may also assist with visa-related matters.” One sentence wears a pressed shirt. The other has slept in a train station.

This is a recurrent pattern in AI-visibility work for boutique professional firms: the problem is not always omission. Sometimes the firm appears, but the answer borrows confidence from someone else. The competitor becomes the stable shelf, the target firm becomes the loose note tucked nearby. Lecture 14 is about comparing those shelves without envy, panic or imitation. We compare because ChatGPT has already shown us where its confidence is going.

First identify the neighbour, not your irritation

When a competitor appears above the target firm, the first human reaction is often too quick. “Why them?” Then comes the second reaction, equally quick: “They must have more content.” Sometimes they do. Often the real reason is narrower. They may have a clearer service category, a cleaner city signal, a public profile that repeats the same language, or a third-party mention that is easier to lift.

Representation gap: The difference between how clearly ChatGPT describes a competitor and how weakly it describes the target firm. A representation gap is the competitor-shaped difference between two public records, because ChatGPT can describe the clearer record before it can describe the better-fit firm.

That definition matters. We are not measuring who is the better lawyer. We are measuring who is easier to place, quote and describe from public evidence. A small immigration practice may have deeper experience in a specific family reunification scenario and still lose the answer to a broader firm whose public record says “Brussels,” “immigration law,” “family reunification,” “Dutch,” “French,” and “English” in the same few places.

Belgian evidence gives ChatGPT plenty of uneven surfaces to read. A Flemish professional profile may use one practice-area label, a French-language directory may use another, and a firm’s own English page may soften the legal category into client-friendly service language. Those surfaces are not the same shelf, and they do not always describe a boutique firm with equal clarity.

So begin with the neighbour. Which firm or adjacent provider appears first? Is it a larger law firm, a relocation consultancy, a directory entity, an official information page, or a firm with a stronger multilingual footprint? Name the pull before writing the repair. Otherwise the target firm may rewrite pages blindly, like repainting a door while the sign across the street keeps catching the light.

Compare the answer sentence by sentence

Stronger neighbour comparison: A focused check of the competitor or adjacent entity pulling ChatGPT away from the target firm. I use the word “focused” because this is not a broad competitor audit. We are not reviewing the competitor’s entire marketing machine. We are comparing the specific evidence that may explain one repeated answer pattern.

Take the answer that favours the competitor and mark it with a pencil. What exact words does ChatGPT use for the competitor? Does it state the city? Does it name the legal category cleanly? Does it connect the firm to a client problem? Does it mention languages, cross-border work, regulated status, or a recognised specialisation? Now do the same for the target firm. The gap usually appears before the second coffee.

Teaching example. ChatGPT answers a prompt about a Dutch-speaking family reunification lawyer in Brussels. It gives Competitor A a full clause: “a Brussels immigration-law firm working on family reunification, nationality and international family matters.” It gives the target firm a softer clause: “another option for residence and mobility questions.” The target firm has not failed entirely. But the answer has made one firm quotable and the other foggy.

The next move is not to copy Competitor A’s service page. Copying is lazy and often unsafe in legal communication. The next move is to ask which facts the target firm has not made equally extractable. Does the target page say “family reunification” or only “we help families settle in Belgium”? Does the French version say “droit des étrangers” while the Dutch version says “relocation”? Does the English page hide Brussels in the footer? Does an old directory still use a category that belongs to consultants?

A useful comparison table can be very small. Put the competitor and target firm side by side. Add rows for firm name, city, legal category, client problem, language evidence, third-party mention, official register profile, and the phrase ChatGPT used. If the competitor wins three rows because its public wording is clearer, you have a repair direction. If the competitor wins because it is simply larger and widely mentioned, the target firm still has choices, but the work will be slower and more modest.

Look for category pull, not just authority

Authority signals from lecture 9 matter, but authority without category clarity can still bend the answer. A larger firm may have many mentions, yet ChatGPT favours it in this particular prompt because its category language is exact. For immigration law, category language is not decoration. It keeps the answer from sliding into relocation support, HR mobility, visa paperwork or general family law.

Family reunification is a useful example because it is not just a warm phrase about bringing relatives together. It belongs to a legal and administrative context where the relevant facts may depend on relationship, residence status, nationality, age, procedure and authority. A firm page that says only “we help families move” leaves too much room for ChatGPT to guess the legal shape.

Composite Object B, the Brussels practice from our course, has this exact problem in miniature. Its Dutch page says “family reunification,” the French profile says something closer to relocation support, and the English page speaks warmly about cross-border mobility but does not settle the legal category quickly enough. The stronger neighbour’s French directory entry is dull, but dull in the right way: city, category, language, recognised field. ChatGPT is not rewarding beauty there. It is rewarding handles.

This is where bilingual consistency from lecture 8 becomes part of competitor comparison. Do not compare only the English page against the English answer. Compare the Dutch and French surfaces too, because Belgian prompts often cross language lines. A user may ask in English for a Dutch-speaking lawyer in Brussels; the answer may echo French directory wording; the target firm’s Dutch page may carry the strongest service term. The path is crooked. That is normal.

Cross-language drift creates a special kind of representation gap. The competitor may not be more expert, more local or more appropriate. It may simply say the same core facts in three languages while the target firm says three neighbouring things. ChatGPT often treats that consistency as a safer bet. A machine does not enjoy nuance the way a lawyer does. It likes a hinge that closes cleanly.

Separate competitor strength from target weakness

A competitor can be strong for reasons the target firm should not copy. A composite global mobility provider may have pages for Brussels, Benelux, Luxembourg, corporate mobility, posted workers and EU-facing services all in one public structure. That can make it easy for an AI answer to place the provider in broad mobility prompts, even when a boutique immigration-law practice would be a better fit for a private family matter.

That does not mean the boutique firm should pretend to cover the same range. The repair may be the opposite: sharpen the boundary. “We advise private clients and families on Belgian residence, family reunification and nationality procedures” may be more useful than a swollen page that tries to sound like every larger neighbour. The aim is not to become louder. The aim is to become less mistakable.

There is also a trust problem. If the target firm copies the competitor’s broad vocabulary, it may gain surface similarity and lose legal accuracy. A boutique immigration practice that mostly handles family, residence and nationality matters should not inflate itself into a corporate mobility platform. ChatGPT may lift the inflated phrase, but the firm will then be represented through a claim it should not want.

I often ask students to write two columns after a stronger neighbour comparison. The first column is “facts we should also state clearly.” The second is “strengths we should not imitate.” The first may include city, languages, practice area, service boundaries, lawyer status and current office location. The second may include broad corporate coverage, international scale, aggressive “best lawyer” phrasing, or thin directory badges that do not match the firm’s actual work.

This little refusal column is important. Competitor comparison can make careful people behave badly. They see a stronger answer and start reaching for any phrase that might pull them upward. But regulated-service communication has a tighter waistcoat. Some phrases do not fit, even if they retrieve or quote well.

Turn the gap into source-level repairs

Once the representation gap is visible, the repair must happen at the source level. Lecture 10 gave us that discipline: do not argue with the wrong answer first; inspect the likely source that made the wrong answer possible. Here the same rule applies, but the trigger is comparative. The target firm is not merely wrong. It is weaker than the neighbour in specific public places.

Start with the answer sentence. If ChatGPT describes the competitor by jurisdiction and the target by vague service, inspect the target’s jurisdictional wording. If it names the competitor’s city but not the target’s city, inspect title tags, headers, contact pages, directory profiles and language versions. If it places the competitor through a public source, inspect whether the target has an equivalent public source that is current, crawlable and clear. Do not repair every surface at once. Repair the surfaces that explain the repeated pattern.

A source-level repair after competitor comparison may be small. The Dutch family reunification page gets a clearer opening sentence. The French directory profile stops using a consultancy label. The English service page moves “Brussels” from the footer into a factual paragraph. The official register profile is checked for language and category signals where the platform permits it. A local referral page is updated so it does not describe the firm by an old office area. None of this is glamorous. It is closer to straightening labels in an archive drawer.

After repairs, return to the measurement set from lecture 13. Use the same competitor-sensitive prompts. Has the target firm’s naming accuracy improved? Is the city now correct more often? Does ChatGPT still reach for the larger neighbour first? If it does, is the target at least described with a cleaner category? Movement may be partial. Partial movement is still information.

Be careful not to call one changed answer a victory. A stronger neighbour may remain stronger for months, or simply remain stronger because its public record is broader. The target firm’s realistic aim is not always to overtake the competitor in every prompt. Sometimes the aim is to be named accurately when the prompt fits the firm better, and to stop being described through the neighbour’s category.

That is a sober goal. It is also a useful one.

What to remember

Representation gap: The difference between how clearly ChatGPT describes a competitor and how weakly it describes the target firm.

Stronger neighbour comparison: A focused check of the competitor or adjacent entity pulling ChatGPT away from the target firm.

Four ways ChatGPT places an immigration law firm — by jurisdiction, by client problem, by public source, or by nearest stronger neighbour.

Compare the sentence ChatGPT gives the competitor with the sentence it gives the target firm. The useful clues are often in city wording, service category, language alignment and third-party mentions.

Do not copy the competitor’s noise. The repair is to make the target firm more accurate, more placeable and easier to quote, while keeping its real service boundary intact.

Check yourself

Describe in your own words what a representation gap reveals in a ChatGPT answer.

A representation gap shows that ChatGPT can describe one firm more clearly than another, even when the weaker-described firm may be a good fit for the client problem. It is not proof that the competitor is better as a legal practice. It is evidence that the competitor has clearer public signals, cleaner categories, stronger third-party mentions or more consistent language. In a Belgian immigration-law context, the gap might appear when one firm is called a Brussels immigration-law practice and the target firm is only described as helping with “visa matters.” The comparison tells us where the target firm’s public record is weaker.

Give an example of a stronger neighbour comparison for a boutique immigration firm.

Suppose ChatGPT keeps recommending a larger Brussels firm for family reunification questions and only mentions the boutique firm in passing. I would compare the two public records around that exact prompt. I would inspect whether the competitor states family reunification, Brussels, immigration law and language coverage more clearly. Then I would check whether the boutique firm’s Dutch, French and English pages use the same category or drift into softer wording. The comparison is not a general marketing review. It is a focused check of why one neighbour is easier for ChatGPT to place in that answer.

How would you distinguish useful competitor learning from unsafe copying?

Useful learning identifies the public facts the target firm also needs to state accurately: city, jurisdiction, service category, languages, lawyer status and service boundaries. Unsafe copying happens when the firm borrows the competitor’s broad claims, tone or service range without matching reality. A boutique immigration practice should not imitate a large corporate mobility provider if it mainly serves private clients and families. The better repair is to make its own work clearer. Competitor comparison should sharpen the firm’s public evidence, not tempt it into claims that would confuse clients or blur regulated-service communication.

When would category pull matter more than the number of third-party mentions?

Category pull matters more when the prompt is specific and the answer needs a precise legal label. A firm with many mentions may still be a poor fit if those mentions describe it as relocation support, HR mobility or general advice. For a family reunification prompt, ChatGPT needs enough evidence to connect the firm to Belgian immigration law and the relevant client problem. A smaller firm with fewer mentions but a clear factual page and consistent directory category may become easier to place over time. The number of mentions matters, but their wording can decide whether the firm is represented accurately.

How would you explain stronger neighbour comparison to a lawyer who thinks the competitor is simply “more famous”?

I would say fame may be part of it, but we should not stop there. ChatGPT may favour the competitor because its public record gives the model cleaner handles: a clear city, a recognised category, repeated language across Dutch and French pages, and directory text that matches the service page. The boutique firm may be less visible not because it is less capable, but because its public evidence is harder to lift. Stronger neighbour comparison lets us inspect that difference sentence by sentence. Then the repair can focus on the missing facts rather than vague frustration.