The GLP-1 Shelf: What Digital Shelf Leaders Do Across Food, Snacking, and Consumer Health.

Three unrelated portfolios — ambient rice and ready meals, chocolate and salty snacks, and over-the-counter digestive care — have almost nothing in common on the shelf. Different buyers, different occasions, different regulators. And yet, put the demand data for all three side by side and the same four findings fall out of every one of them. That convergence is the whole point of this piece: it is the evidence that GLP-1 is not a category fad but a structural change in how the digital shelf reads a product.

The change lands differently on the three shoppers the Genrise digital shelf framework uses to describe the category. The Human Shopper — still roughly 85% of traffic — types a short query like protein snack or high-protein rice into a retailer search bar and scans the results. The AI-Assisted Human — around 10–15% and rising — asks an assistant like Amazon's Rufus, now operating inside Alexa for Shopping, or Walmart's Sparky, a full natural-language question such as what's a filling, high-protein snack that keeps me full between meals? — and takes the shortlist it returns. The Autonomous Agent — under 1% today, emerging fast — selects and buys against structured attributes and a ratings floor with no human review at the point of decision. Each of the three reads the health-engaged signal through a different lens, and a brand optimized for only one of them is exposed in the other two.

This is a point of view, not a neutral evidence report. It carries a thesis and it prescribes. The strategic recommendations that sit outside the scope of our dated evidence reports are, by design, inside the scope of this one. What follows is the shared pattern, then what it means at each category desk, then the reason the answer is a system rather than a copy refresh.

1. The shift beneath the shift

Three things are true at once, and holding all three is the point.

First, the signal is real and it is large — but the medicated cohort is the visible edge, not the market. In the United States, current GLP-1 use for weight loss reached 11% of adults in 2026, up from 3% in 2024 (Gallup, 2026); on a broader measure roughly 12% of adults, and about 23% of households, now include a GLP-1 user (Circana, November 2025). That is a mass-market cohort. But it is dwarfed everywhere by the health-engaged shopper standing behind it. In research fielded for a major beverage manufacturer, 86% of Americans said they were actively adding protein to their diet — against the 12–16% on a medication. One widely reported weight-support frozen line disclosed that 77% of its sales came from shoppers not on a GLP-1 at all. The commercially useful audience is the far larger protein-first, portion-aware shopper. GLP-1 is the reason that shopper is suddenly legible in the data.

Second, it is a restructuring, not a collapse — and it partly reverses. The strongest peer-reviewed evidence, a Journal of Marketing Research study published in December 2025 built on a roughly 150,000-household panel, found GLP-1 households cut grocery spend 5.3% within six months of adoption, and more among higher-income households, with the sharpest declines concentrated in calorie-dense categories. But the same basket shows categories rising — yogurt, nutrition bars, fresh produce — and roughly a third of users in the study discontinued, reverting toward their pre-adoption baskets. This is demand moving, not demand disappearing.

Third, the shelf that receives this shift is increasingly an AI-mediated one. The health-engaged shopper is exactly the shopper most likely to ask an assistant "what's a filling, high-protein snack" rather than browse a planogram — and the assistant answers with a shortlist of five, not a shelf of fifty. That is where the demand shift and the discovery shift compound, and it is where the opportunity and the risk both concentrate.

The rest of this piece is built on that third point. The demand shift is not something a brand controls. The way the shelf reads a product in response to it is.

2. Who's actually shopping this way

The single most common strategic error here is treating a US headline as a global fact. It is not, and the gap is large enough to change the strategy market by market.

The United States is the lead indicator. Prevalence is high (about one in eight adults currently on a GLP-1, per KFF's late-2025 tracking poll), use skews female and middle-aged, and there is no equivalent of Europe's regulatory brake on the category. Where an effect appears in US basket data, GLP-1 and the broader health-engaged shift are the plausible drivers. Oral formulations and direct-to-consumer pricing are widening access, while high discontinuation keeps the cohort churning — a large, growing, high-turnover population of current, former, and returning users whose health-aware habits frequently outlast the prescription.

The United Kingdom is the lead European market, and it is smaller and confounded. The strongest UK evidence — a peer-reviewed population study published in January 2026 — put past-year GLP-1 use for weight loss at 2.9% of adults, roughly 1.6 million people, with access overwhelmingly private rather than through the NHS. That is around a quarter of the US rate. Critically, the UK also ran two major regulatory changes through the same window: the ban on volume promotions for less-healthy products from October 2025 and advertising restrictions enforceable from January 2026. When a UK shelf shift matches a US one despite four-to-six times lower medicated prevalence, the honest read is that regulation, not GLP-1, is the likelier driver.

Continental Europe is smaller still, and structurally capped. Germany excludes weight-loss GLP-1s from statutory insurance as "lifestyle" medicines, holding the cohort down; France only began reimbursing in mid-2026. An ING analysis estimated the near-term hit to total EU-plus-UK calorie demand at around 0.25%, rising to an outer bound of 2.5–3.5% by 2030 in its most aggressive scenario. Europe also carries a reputational wrinkle the US does not: German nutrition authorities and consumer groups are actively criticizing high-protein marketing as an unnecessary trend, which tempers how loudly a protein-led play can be pitched there.

The operating rule that follows: write the strategy for the broad health-engaged shopper who exists in volume everywhere, use the US as the place to test where prevalence is highest, and never port US-scale creative or claims into Europe on the assumption the demand is the same size. It is not.

3. A restructuring, not a collapse — and it's the same everywhere

Here is the finding that turns three separate category analyses into one point of view. Across staples, snacking, and consumer health — portfolios with nothing structurally in common — the demand reshapes into the same three directions.

What contracts: the calorie-dense, impulse, sugar-led occasion. Savory snacks fell about 10% in the peer-reviewed panel, the sharpest decline in the basket, with comparable drops in sweets, cookies, and sweet baked goods. Limited-service restaurant spend fell around 8%. The occasion contracts before the recipe changes — the reward snack and the checkout impulse go first, because they were built on frequency and easy reach.

What holds, and partly reverts: center-store staples and anchor meals. These show a mix-shift rather than an abandonment, and because roughly a third of users discontinue and drift back toward prior baskets, the downside is cushioned by reversion. Portion-controlled, high-protein, ready-to-eat formats are the best-positioned versions of these categories rather than the most exposed.

What grows: anything that reads as protein, fiber, functional, or symptom-supporting. Yogurt was the only category to rise with statistical significance in the panel, alongside nutrition bars and fresh produce.

There is also a real demand mechanism underneath the growth, not just a marketing narrative. Between a quarter and 40% of the weight lost on a GLP-1 can be lean muscle mass, and clinical guidance converges on 1.2–1.6 grams of protein per kilogram of body weight per day to preserve it — a target most users do not hit. The protein tilt is physiological before it is promotional.

4. The words shoppers search are the words brands can't say

This is the signature constraint of the entire opportunity, and it is identical in all three categories: the language the shopper searches is precisely the language the brand cannot legally use.

A shopper types "GLP-1 friendly snack," "Ozempic constipation," "food that keeps you full." Those are the high-intent, high-visibility phrases. And they are almost all unusable. "GLP-1 friendly," satiety, appetite, and weight-loss framings are treated as health or disease-related claims and are effectively off-limits on UK and EU product pages, restricted on Amazon across every marketplace, and — in the US — unregulated, contested, and increasingly seen as a brand-risk rather than a badge. Naming a prescription drug to sell a food or supplement is restricted or prohibited across all three regimes and a likely marketplace-policy violation.

What is available is the compositional and functional language a product genuinely qualifies for. Under EU and GB rules, "source of protein" requires at least 12% of energy from protein and "high protein" at least 20%; "source of fiber" and "high fiber" carry defined thresholds; "low fat" is defined; and a small set of authorized function claims — protein's contribution to the maintenance of muscle mass, sugar-free gum's dental benefits — can be used where the product qualifies. In the US, structure/function language ("supports muscle") is permissible for supplements under DSHEA in a way it is not in the EU or UK, which is a divergence that does not travel.

The strategic move, then, is not to find a cleverer way to say the forbidden thing. It is to bridge the gap with non-claim adjacencies: protein and fiber where the numbers support them, portion and format and occasion language that carries the positioning without making any claim, and symptom-to-attribute translation that meets the need without naming the drug. This is the same discipline our product claims and AI visibility analysis describes — the gap between what a shopper wants to hear and what a brand may lawfully say is not a copywriting problem to be finessed. It is a structural feature of the category, and closing it defensibly is the work.

5. Center-store staples and ambient meals

The desk selling ambient rice, grains, pasta, sauces, and ready meals.

The exposure. Carbohydrate-centric staples sit in the mix-shift-not-collapse middle. The panel data puts pasta, grains, side dishes, and pantry staples down in a 5–10% range, but the effect attenuates after six months and reverses on discontinuation. Ready meals are measurably negative on the frozen side, yet portion-controlled, high-protein formats within that same category are among the better-positioned formats in the store. This is the category where the average most badly misrepresents the archetype.

Executable now, without reformulation

• Lead with portion and format convenience as a no-claim asset. Single-serve microwave pouches and right-sized formats structurally match the smaller-portion demand created by appetite suppression — and communicating "single-serve" or "perfectly portioned" makes no regulated claim at all.
• Write product pages and enhanced content in natural, intent-rich language for assistants rather than keyword strings — "a high-protein rice bowl that's ready in two minutes and easy on the stomach" — and complete every structured attribute field. See the AI product descriptions approach for the mechanics.
• Build out Q&A sections that answer the real adjacent questions ("how much protein," "how many calories," "is this a small portion") without touching prohibited framing.
• Frame balanced-meal and recipe context — pairing with vegetables and lean protein — which is lifestyle context, not a health claim.

Gated on legal or R&D

• "Source of protein" or "high protein" on standard white rice or ambient pasta generally requires reformulation — plain rice derives well under the 12% energy-from-protein threshold. Wholegrain and pulse-blend variants are where protein and fiber claims become credible; verify per SKU.
• Fiber and wholegrain references need per-SKU verification against actual composition before use.
• Low-fat is a strong, immediately available angle for tomato-based sauces, which frequently qualify; sugar and salt are the reputational exposure there and any "reduced" claim needs a substantiated 30% reduction.

6. Snacking and confectionery

The desk selling chocolate, sugar confectionery, salty snacks, gum, mints, and bars.

The exposure. This is the category hit hardest and the one where a single number is most misleading. Savory snacks and sugar-led treats are the steepest measured declines. But within the same portfolio, premium chocolate, gum and mints, and protein bars are simultaneously growing. The portfolio answer is a hard split between losing and gaining archetypes.

Executable now, without reformulation

• Defend exposed indulgence through portion architecture, not reformulation. The evidence points to downsizing within a brand — mini, treat-size, single-serve — rather than abandonment; a small format reads as "a treat, not a meal." Sugar confectionery with no protein or functional story is the least defensible archetype, so portion is its main protection.
• Gum and mints are the cleanest legal-and-relevant fit in any portfolio, and usually an under-used one. Dry mouth is a documented GLP-1 symptom, and "reduction of oral dryness" is an already-authorized dental claim in the EU — the one place a real medication symptom aligns with a claim a brand may lawfully make, with no weight or GLP-1 framing required. The dental-benefit platform often already exists on-pack and is simply under-leveraged.
• On protein and functional formats, use quantified protein and fiber claims where the SKU qualifies, and lane them by occasion so your own bars, meal-adjacent protein, and indulgence-with-benefit lines do not cannibalize each other.

Gated on legal or R&D

• Protein-forward reformulations that need to carry "high protein" require hitting the threshold, and — for France specifically — a Nutri-Score check, because the 2023 algorithm caps the protein benefit for products high in sugar or salt, so a protein reformulation of a sweet snack may not move its grade.
• Any structure/function "supports muscle" language works only in the US and does not travel to the EU or UK.

7. Consumer health: OTC and supplements

The desk selling digestive care, antacids, fiber, and daily supplements.

The exposure. This is the category where the shift is demand-creating, not demand-shifting — but only for the right sub-set of the portfolio, and only where prevalence is high. The gastrointestinal side-effect cluster (nausea, bloating, constipation, reflux) affects 40–70% of users and maps directly onto digestive-care indications. Because users churn, the symptom-sufferer pool is a flow, not a stock — a constant stream of new starters at dose initiation, which favors continuous acquisition over lifetime loyalty.

Executable now

• Translate symptom to attribute to occasion language, never to a drug name. "For when your routine changes," "gentle, as-needed relief you can keep on hand," "everyday hydration support" — occasion and format framing carries the positioning without a claim.
• Where a product is a structure/function supplement rather than a monograph OTC drug, it can legally reference "GLP-1 users" and "low-calorie diets" with the appropriate disclaimers in the US — a claims advantage the OTC drugs in the same aisle do not have. Invest disproportionately in that position.
• Engineer companion-care storefronts for AI ingestion — structured attributes, deep Q&A, recent reviews — and close obvious basket gaps such as hydration and electrolytes.

The trap to avoid

• Do not fight for a searched symptom the clinical evidence steers away from your product. The clearest example: bulk-forming fiber is actively counter-recommended for GLP-1 constipation in favor of other agents, so competing head-to-head for "GLP-1 constipation" with a bulk-fiber product is a losing query. Repivot to a defensible adjacency — closing the fiber gap on a low-calorie diet, gentle daily regularity — rather than a symptom-rescue claim the evidence undercuts.

8. How AI-mediated discovery decides — and what you can shape

Every playbook above routes through the same chokepoint. The health-engaged shopper increasingly discovers products through an assistant, and the assistant's selection logic is not the retailer search bar's logic.

Amazon's Rufus, now operating inside Alexa for Shopping, has assisted more than 300 million customers, is on pace for over $12 billion in influenced sales with engaged users converting roughly 60% better, and grew usage 149% year over year. Walmart's Sparky serves the same persona, with around 50% adoption and materially higher average order values. These are covered in depth in the Amazon Rufus deep-dive, the Walmart algorithm analysis, and the cross-assistant field guide. The behavioral backdrop is stark: AI-source traffic to US retail surged 693% in the 2025 holiday window and converted 31% better (Adobe), and roughly 20% of holiday orders touched an AI-influenced journey (Salesforce).

What matters for this piece is how these assistants choose. Continuous tracking of Rufus recommendations found the assistant essentially does not surface products below 4.0 stars — the median recommended product sits around 4.5 stars with a median review count near 3,000. It narrows a field of dozens to about five named products, drawing on structured backend attributes, review text, and community Q&A as much as on front-end copy.

That splits cleanly into what a brand controls and what it can only influence — the distinction at the center of Genrise's AI Shelf Visibility work:

• Directly controllable: product-page titles, bullets, and descriptions written in natural language; complete structured attribute fields; Q&A coverage; review solicitation to clear the ratings and review-count floor; and schema on owned sites mirroring the retailer feed. The PDP content checklist is the operational anchor for most of this.
• Influence only: the third-party review, forum, and editorial pool that AI answers lean on heavily. A brand cannot control it, but it can seed genuinely useful content, encourage authentic reviews and Q&A, and keep its brand-entity information accurate.

The load-bearing conclusion, true in all three categories: for a health-framed query, review and attribute health is more often the gate than the copy. You can write a flawless, compliant, intent-rich description and still not surface, because the SKU sits at 3.8 stars with forty reviews. Fix the reviews and the attributes first. The copy is necessary and not sufficient.

9. Why this is a content-intelligence problem, not a copywriting one

Here is where the point of view resolves. The public research on this shift is directionally clear and specifically silent. It proves that calorie-dense categories fall and protein categories rise — and it goes dark exactly where the money is. There is no published basket panel outside the US. There is no robust proof that rice held up better than pasta, or that wholegrain behaved differently from refined, at the granularity a category lead actually needs. The studies aggregate "sweets" and "savory snacks" and do not decompose them.

That silence is not a gap to be filled with a confident guess. Manufacturing sub-category precision the data does not support is the fastest way to lose credibility with a sophisticated audience. The honest position is that the per-SKU opportunity cannot be read off a public report — it has to be found, product by product, in the signals that do exist at that granularity: the review text where shoppers describe their actual use case, the structured attributes that are complete or missing, the Q&A that is answered or unanswered, and the assistant surfaces where a product does or does not appear for a given natural-language prompt.

Doing that once, by hand, for a hero SKU is an afternoon. Doing it across a portfolio, in multiple markets, and keeping it current as the citation pool shifts week to week and as assistants change their logic, is not a copywriting task. It is a continuous, always-on content-intelligence task — mining reviews for the language of real need, auditing attributes against what the assistants weigh, and monitoring AI visibility as an ongoing measurement rather than a one-time audit. This is precisely the case set out in our answer engine optimization and generative AI in ecommerce analyses, and it is the reason the response to a structural demand shift has to be a system and not a campaign. The demand shift is permanent. The shelf's reading of it changes constantly. Only a standing capability keeps pace with both.

10. Through the three-persona lens

The same shift, read by the three shoppers, asks for three different things at once.

The Human Shopper still types a health modifier into a retailer search bar and scans a page of results — so head-term coverage and honest, scannable on-page compositional claims still matter, and abandoning classic search discipline to chase the assistant leaves the majority of traffic unserved. The AI-Assisted Human asks a conversational question and receives a shortlist — so natural-language copy, Q&A depth, and review sentiment decide whether a product is in the answer at all. The Autonomous Agent evaluates programmatically against a ratings threshold, review count, structured attributes, and price and inventory clarity — so machine-readable completeness and a clean ratings floor become the price of entry, independent of how good the copy reads to a human.

A brand built for only the first is invisible to the second and third. A brand that chases the third while neglecting the first strands the long tail of conversational queries. The health-engaged shift does not let a category pick one surface — it arrives on all three, which is the underlying argument for treating the digital shelf as one system with three readers.

11. What we don't know — and won't pretend to

The credibility of this point of view depends on being precise about its limits.

Evidence density is wildly uneven. The US produces quantified, peer-reviewed household-panel findings; Europe is largely thin, single-source, or inconclusive. These should never be leveled into one global claim, and US figures are not presented here as European fact — European prevalence runs several times lower and is confounded by regulation that has nothing to do with GLP-1.

Sub-category granularity mostly does not exist. The public studies aggregate to a level above the archetype, and no UK or German equivalent of the US basket panel has been published. No search-volume data ties health-aware terms to specific food categories beyond isolated retailer disclosures. Where a number would require inventing precision, this piece declines to invent it.

Discontinuation cuts both ways. Roughly half to two-thirds of users discontinue within a year, which softens the downside through reversion but also makes any bet defined narrowly by medication status risky over a long horizon — another reason the durable strategy is written for the broad health-engaged shopper, not the prescription.

And AI-mediated discovery is a moving, imperfect target. The assistants have documented accuracy and hallucination issues on discovery and health queries, their handling of health topics is still evolving, and their citation pools shift within weeks. Visibility is point-in-time discovery mechanics, not endorsement — which is exactly why it needs continuous measurement rather than a single audit.

None of this weakens the thesis. The convergence across three unrelated portfolios is robust precisely because it does not depend on the fragile sub-category numbers. What is happening is clear. What any single brand should do about it is specific to its portfolio, its markets, and its content infrastructure — and that is the conversation the playbook above is built to start.

The health-engaged shift is not a moment to message around. It is a change in how the digital shelf reads a product, arriving on three surfaces at once, and it rewards the brands that treat their content as a living system rather than a fixed asset. GLP-1 is simply the clearest signal that the change has already begun.

Genrise builds the always-on content intelligence that finds and closes these opportunities across the digital shelf — mining reviews, auditing attributes, and monitoring AI visibility continuously, per SKU, across markets.

References

This article synthesizes primary and commercial sources triangulated across three category analyses. Confidence grading, single-source flags, and the full data-gaps register sit in the underlying Genrise analysis. Client-identifying detail has been generalized to category archetypes and symptom-to-attribute patterns. All claims recommendations require legal sign-off before deployment.