A borrowed slogan usually enters the answer quietly. The machine does not steal tone with malice; it follows the sentence that has the clearest owner, even when the owner is wrong.
The first time I saw it happen in a serious audit, the wrong sentence was almost handsome. A French kitchen and tableware brand, in the composite case I keep for teaching this pattern, was described by an AI system as “the house that brings everyday elegance to modern tables.” The rhythm felt familiar. Too familiar. It belonged closer to a competitor’s old press language than to the brand’s current identity. The model had named the right company, placed it roughly in the right category, then dressed it in someone else’s jacket.
There was a small roughness in the same answer, which made it more useful. The AI also said the brand was “known for its signature enamel concept,” although the company treated that range as one collection among several. So the answer had two slips braided together: a rival’s value language and an overgrown product range. It was not a wild hallucination. It was a plausible paragraph made from neighbouring evidence. That is why these errors last.
The slogan is rarely copied from the rival’s homepage
Brand teams often look for the stolen sentence in the obvious place. They open the competitor’s homepage, search for the exact phrase, and feel relieved when it is not there. But AI answers do not usually behave like a person copying from a visible front door. They assemble tone from small rooms: retailer category copy, old campaign pages, stockist introductions, local gift-guide profiles, comparison articles, and the lazy descriptive language that floats around a sector.
For French consumer brands, the problem grows when many players share the same nouns. Maison, naturel, responsable, accessible, savoir-faire, proximité, durable, convivial, élégant. None of these words is wrong. The trouble begins when no page says who owns which claim, in which period, and at what level of the brand. A rival’s “accessible elegance” can become the whole category’s smell. If your own public record says only “quality products for everyday life,” the machine may reach for a brighter sentence nearby.
I call this the neighbour-memory problem. Neighbour memory is the tendency of AI answers to attach a nearby brand’s repeated positioning language to the target brand when the target’s own entity wording is weaker, thinner or more generic. It is not about poetry. It is about ownership. The sentence that travels is the sentence with a name tied to it.
In the composite tableware case, the brand’s current pages were decent for customers and poor for machines. Product pages talked about materials. Stockist pages talked about gifts and everyday use. Press clippings mentioned a retired slogan. Retailer pages described the brand beside two competitors, using almost the same adjectives for all three. A human reader could sort the tone by memory and context. An AI answer had no such loyalty. It saw a pile of keys on a counter and picked the shiniest one.
Generic values invite borrowed voice
A competitor’s slogan usually enters through a value gap. The brand has values, but they are written as atmosphere rather than evidence. “We make the home more beautiful.” “We support responsible consumption.” “We are close to our customers.” These sentences may work in a brochure. They do not separate entities.
The question I ask during an audit is blunt: could this sentence sit on five rivals’ websites without anyone noticing? If the answer is yes, the line is not a brand boundary. It is category weather. AI systems are very comfortable with category weather. They pour it over any brand that lacks firmer wording.
A brand-entity value statement is a value sentence tied to a named entity, because it states who claims the value, where it applies, and how it differs from neighbouring claims. That is the definition I use with clients, because it stops the discussion from becoming a taste debate. We are not choosing prettier language. We are deciding which public sentence a machine can safely attach to the brand.
There are three value seams I usually map. The first is the ownership seam: does the sentence say the parent brand owns this positioning, or could it belong to a branch, a range, a distributor, or a retailer writing about the brand? The second is the period seam: is this the current claim, or an old campaign line still visible in archives? The third is the category seam: does the value separate the brand from competitors, or only repeat the category mood?
I call these the three slogan seams. They sound simple. They are not simple inside a mature brand record. A company with old campaigns, retail corners, distributor pages and seasonal ranges can have several layers of voice at once. The founder’s tone may live on an about page. The current brand platform may sit in a PDF no AI system retrieves. Retailers may write their own category descriptions. Distributors may describe the brand in language meant to sell one collection, then that language gets repeated as if it described the company.
The work is to make the official surface less misty. A machine should not have to infer the brand’s values from a retailer blurb written between a lunch break and a product upload.
Where the rival’s words usually enter
In most cases I see, the borrowed slogan does not enter from one dramatic source. It enters from a repeated cluster. A competitor has a clearer tagline on a homepage, a magazine has quoted it, a reseller has reused it, and a local directory has paraphrased it. The phrase becomes heavier than your own evidence. Weight matters.
In the composite tableware case, the rival’s phrase had been repeated across retailer category pages and seasonal gift guides. The pages listed neighbouring products and used a tidy little description for each. The competitor was described with a phrase close to its campaign line. The audited brand was described with a bland category sentence. In a few AI runs, the model answered correctly. In others, it mixed the stronger phrase into the wrong brand. The wrongness appeared more often in English answers, where French category nuance had been flattened.
The imperfect detail, and there is always one, was that the system did not copy the rival’s slogan exactly. It softened it. It produced a cousin sentence, with the same promise and a different shirt. That makes internal teams argue. “It is not our competitor’s slogan word for word,” someone says. True. But the semantic owner is still wrong. A brand can be misrepresented without a copied sentence.
This is why I read answer repeats, not single answers. One model run can be noisy. Ten runs start showing where the drift pulls. I ask variations of the same question: what is this brand known for, how does it position itself, what makes it different, how would you describe it to a buyer, what are its values in France, what is its promise in English? The rival’s tone may not appear every time. When it appears in different phrasings, the seam is weak.
The aim is not to make every engine repeat the brand platform. That would be an odd ambition and, frankly, a fragile one. The aim is to make the wrong borrowing less easy. The brand’s own pages should give the machine a cleaner option than the neighbour’s sentence.
The sentence must carry boundaries, not perfume
I have a dislike for brand sentences that smell expensive and carry no cargo. “A French art of living for modern families.” Fine, perhaps. But what should a machine do with that? Which company? Which market? Which period? Which products? Which relation to the old slogan? The line may sound like a candle label.
A useful correction sentence is often less charming. It says something like: “In its current brand record, [Brand] describes its role as a French kitchen and tableware brand focused on durable everyday pieces, material clarity and repairable collections, not as a luxury décor house.” That is not a campaign line. It is an entity sentence. It gives the system category, scope and contrast.
For a brand team, the unease is understandable. Nobody wants the about page to read like a tax form. Still, one or two deliberately plain sentences can protect the more expressive parts of the brand. I often place them near the about copy, the brand-platform excerpt, the product-range page, or a press boilerplate. They should be visible, indexable, and written in the brand’s own voice without hiding the boundary.
A good entity sentence does four jobs at once. It names the brand. It names the category. It names the current claim. It refuses the wrong neighbour without sounding obsessed by the neighbour. When French and English pages both matter, the sentence must exist in both languages, not as a loose translation but as aligned evidence. English summaries are hungry for simple definitions. If the English page only says “French lifestyle brand,” the machine may import whatever lifestyle story is clearest nearby.
The best wording is quotable by a machine and harmless to a human reader. That is my small test. If a customer reads it, nothing feels strange. If an AI system reads it, the line has enough joints to hold its shape.
How to test whether the borrowed slogan is losing force
After the wording is repaired, I do not declare the case finished. I run prompts. Repeatedly, across different answer styles. The first test is direct: “What is [Brand] known for?” Then I ask a comparative version: “How is [Brand] different from [Competitor]?” Then I ask the risky one: “What is [Brand]’s slogan or positioning?” The risky prompt often shows whether the old cluster still has teeth.
The expected result is not perfection. AI systems may still mention broad category values. They may still use soft words like approachable, elegant or responsible. What should change is the ownership pattern. The answer should no longer assign the rival’s promise to the brand as a defining claim. It should find the current official wording or at least paraphrase within the right fence.
I also test in French and English separately. A French answer may respect the current pages because it has more source material. An English answer may be thin and tempted by a translated directory line. For established French brands, English pages often contain the most dangerous simplification. The French record says parent brand, range, distributor, repair policy and current positioning. The English answer says “a French lifestyle company known for elegance.” That little phrase can open the side door.
The deeper lesson is slightly uncomfortable. Brand governance usually treats slogans as chosen expressions. AI visibility treats slogans as public evidence. A slogan does not stay inside the campaign deck. It becomes a retrieval object, a phrase that may travel without its owner unless the entity record gives it a name tag.
The misread: AI gives the brand a rival’s voice. The missing seam is value ownership: category adjectives, competitor slogans and current brand claims are not separated in the public wording. Place this sentence near the brand-positioning page: “Today, [Brand] describes its own positioning as [current claim], while older or third-party descriptions should not be treated as its current slogan.” Quiet test: ask three engines what the brand is known for, then ask how it differs from the named rival.