[Brand name changed. Based on a composite of patterns observed across multiple children's supplement brands in our corpus.]
We are calling this brand Bloom Pediatrics. It built a following on clean labels, bright packaging, and copy that reads like a conversation with a pediatrician who also happens to run a wellness Instagram. The brand's children's multivitamin line, which includes liquid drops, gummies, and elderberry formulations, regularly appears in Amazon's top-ten selling positions. The product pages are polished. The claims are confident. And across four major AI engines, those claims are being systematically rewritten before they ever reach a shopper.
We pulled the six most-trafficked Bloom Pediatrics product pages from SimilarWeb data in April 2026 and ran each page's primary marketing copy through ChatGPT, Perplexity, Gemini, and Claude. The question was simple: "What does this product do?" The answers were not.
What the pages actually say
Bloom Pediatrics' Liquid Morning Multivitamin for Kids lists, among other phrases, "supports immune function," "promotes healthy growth and development," and "helps maintain strong bones." These are textbook structure-function claims, the kind DSHEA permits as long as the FDA disclaimer rides alongside them.
The Elderberry Gummies page goes further. It describes the product as providing "immune defense," references "antioxidant protection," and at one point uses the phrase "helps your child's body fight back." That last phrase sits in an interesting gray zone. "Fight back" against what? The page never names a disease. But the implication is hard to miss if you have read a few hundred FDA Warning Letters. And the engines have read all of them.
The Vitamin D3+K2 Liquid Drops page includes "supports calcium absorption for strong bones and teeth" and "promotes cardiovascular health." The Toddler Multivitamin references "brain development" and "cognitive function support." None of these phrases, standing alone, would trigger a Warning Letter. Taken together, they paint a picture of a product line making broad health claims that sit right at the DSHEA boundary.
What the engines say instead
Here is where it gets interesting. When we asked ChatGPT to summarize what Bloom Pediatrics' Elderberry Gummies do, the response included a disclaimer the brand never wrote: "These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease."[1] The brand's own product page does carry a version of this disclaimer, buried in small type. ChatGPT promoted it to the second sentence of its summary.
Perplexity dropped "helps your child's body fight back" entirely. It replaced the phrase with "may support immune system function," a softer construction that strips the combative language the brand chose. Claude's summary was the most conservative of the four, stripping the copy down to ingredient lists and RDA percentages while omitting every structure-function claim.
Gemini kept the most original language but added a caveat at the end: "As with any supplement, consult your child's pediatrician before use." Bloom Pediatrics' page does not include that recommendation. Gemini wrote it in.
The engines are not censoring these brands. They are doing something more subtle: rewriting marketing copy through the filter of enforcement history the brand's own team has never read. EvidenceSignal Research
The consistent gap
Across all four engines and all six product pages, we found a consistent pattern. Every engine softened or removed language that implied a product could intervene in a health outcome. "Supports" survived. "Promotes" survived with caveats. But "fight back," "defense," and "protection" were either dropped, rewritten, or footnoted with disclaimers.
The gap is not random. It maps almost exactly onto the distinction between structure-function claims and disease-proximate language that the FDA has enforced in Warning Letters since 2017. In our corpus of supplement-related letters, the word "defense" appears in 14, nearly always in a disease-claim context. "Protection" appears in 22. The engines appear to have absorbed this pattern.
What makes this worth tracking is that nobody on Bloom Pediatrics' marketing team appears to have noticed. The brand's Amazon listings, Shopify product pages, and social media copy all use the same language. The disconnect only becomes visible when you compare what the brand says with what a shopper actually reads after an AI engine processes the query. The shopper never sees the original copy. They see the engine's version.
Why this matters for the brand
Bloom Pediatrics is not in regulatory trouble. The brand has not received a Warning Letter. Its claims are, by most readings, within DSHEA boundaries. The problem is not compliance. The problem is that the AI-mediated version of the brand is quieter, more hedged, and less persuasive than the version the marketing team wrote.
For a children's supplement brand, the stakes are specific. Parents searching "best kids multivitamin" or "is elderberry safe for toddlers" increasingly get their answers from AI summaries, not from clicking through to product pages. If the engine's summary strips the strongest selling language and adds disclaimers the brand did not write, the brand is losing control of its own pitch at the exact moment the shopper is making a decision.[2]
We estimate Bloom Pediatrics' AI citation share across the four engines at roughly 8.2% for children's multivitamin queries, down from 11.4% in January 2026.[3] The decline correlates with increased engine caution around immune-related claims following a batch of NAD decisions in late 2025 that targeted elderberry and zinc marketing. Bloom Pediatrics was not named in those decisions. But the language patterns the decisions flagged overlap with the language the brand uses.
What to do about it
Two concrete steps for any brand in this position:
- Run your top product pages through all four engines monthly. Compare the output to your original copy. The delta is your AI-visibility gap.
- Identify which specific phrases the engines are softening or dropping. If a phrase triggers caution across three or more engines, it is likely sitting near an enforcement boundary your team has not mapped.
Bloom Pediatrics has built a brand that parents trust. The question now is whether the engines trust it, too, or whether the enforcement data has taught them to keep their distance.[4]
Footnotes
- ChatGPT response captured April 12, 2026, using the GPT-4o model via standard web interface. Prompt: "What does Bloom Pediatrics' Elderberry Gummies for Kids do?" The FDA disclaimer appeared as the second sentence of the response, unprompted.
- Gartner estimates that by 2026, 25% of product research queries will be answered by AI-generated summaries without a click-through to the source page. For supplement queries specifically, EvidenceSignal's internal tracking suggests the figure is closer to 35%.
- AI citation share measured using EvidenceSignal's weekly index of 40 standardized supplement purchase-intent phrases across four engines. Methodology at /methodology.
- Bloom Pediatrics is a fictional composite. This analysis was conducted using publicly available product pages and AI engine responses from multiple children's supplement brands.