GEO for medical devices: the short version
In 60 words: Consumer medical devices are a category where AI search routes through regulatory status and intended use before the brand. To get recommended by ChatGPT, Perplexity and Google AI, a Shopify device store needs its FDA class or CE marking, the exact cleared intended use, contraindications and instructions for use, measurable specs, and claims that never exceed the clearance, all as crawlable text and structured data. This guide covers each lever with sources.
On 14 July 2025, the FDA sent a warning letter to WHOOP stating that its Blood Pressure Insights feature was being marketed in the United States without clearance, because providing a blood pressure estimate is an intended medical use and “not a low-risk function” (FDA, July 2025). WHOOP disagreed publicly the next day, and by 2026 the same warning letter was anchoring a consumer class action against the company (ArentFox Schiff, 2026). This is the device category’s defining GEO constraint in one episode: the moment a product page claims a use beyond what is cleared, it becomes both a regulatory liability and a claim an AI assistant will refuse to repeat.
The discovery channel is real and arriving fast. OpenAI launched ChatGPT Health in January 2026 and reported that more than 230 million people ask health- and wellness-related questions on ChatGPT every week, with agentic shopping features rolling into the same surface (Fortune, January 2026). A KFF tracking poll in March 2026 found about a third of US adults have turned to AI for health information (KFF, March 2026). When a shopper asks “best pulse oximeter for home use” or “FDA-cleared TENS unit for back pain”, the structure of your product data decides whether your device is in the answer.
This is Generative Engine Optimization (GEO) applied to consumer medical devices: oximeters, blood pressure monitors, TENS and EMS units, braces and supports, OTC hearing aids, thermometers and similar products sold direct to consumers. If you have not mapped how an assistant reads a cleared indication yet, start with what GEO is, the difference between AEO, GEO and SEO, and the 9 factors of a GEO readiness score. This guide goes deep on what is specific to medical devices on Shopify, and it is not medical or regulatory advice: confirm your own device’s status and claims with your regulatory function or counsel.
Why medical devices are a category AI treats differently
In most categories a shopper asks for “the best X” and the model returns a brand. For a medical device, the answer routes through regulatory status and intended use before brand. A device that is not cleared or CE-marked for the use a shopper describes, or whose page claims more than its clearance covers, is the device an AI is most likely to drop, because recommending it would mean repeating an unauthorized health claim.
Three forces make this category distinct.
The regulatory layer is the trust layer. A medical device in the US sits in one of three FDA classes by risk: Class I (general controls, most are 510(k)-exempt), Class II (special controls, most reach market through a 510(k) clearance showing substantial equivalence to a predicate, or a De Novo when no predicate exists), and Class III (the highest risk, requiring premarket approval) (FDA, Classify Your Medical Device; FDA, De Novo). In the EU, a device carries CE marking under Regulation (EU) 2017/745 (the MDR), is assigned to class I, IIa, IIb or III, and for everything above class I a notified body is involved in the conformity assessment (EUR-Lex, MDR 2017/745). This status is exactly the kind of verifiable, third-party signal an AI weighs heavily, and it is the first thing your page should make machine-readable.
The claims line is sharper than in any adjacent category. There is no “structure-function” safe harbor here the way there is for supplements. For a device, the governing concept is intended use: all promotional material must stay consistent with the cleared or approved indication, and a claim that goes beyond it creates a new intended use that would itself require review (21 CFR 801.6). The FDA acted on this aggressively in 2025: its Office of Product Evaluation and Quality issued 32 enforcement letters targeting device marketing, including claims outside cleared intended use, AI-enabled features promoted without authorization, and missing Unique Device Identifiers (Covington, March 2026). The WHOOP case is the consumer-facing example of the same rule.
The category is moving from accessory to mainstream. The OTC hearing aid category alone, created when the FDA’s final rule took effect in October 2022 and let adults with perceived mild to moderate hearing loss buy hearing aids without a prescription or exam, is now a market measured in the billions and drawing consumer-electronics entrants (Federal Register, August 2022). As more clinical-grade products reach direct-to-consumer shelves, the brands whose pages make status, indication and specs legible to a model are the ones AI will name.
Now the honest counterweight. AI shopping for devices is still early in absolute terms; search, Amazon and retailer marketplaces still drive far more device discovery than ChatGPT today, and regulators have themselves flagged accuracy gaps in consumer health AI, so do not treat AI recommendation as a substitute for the channels that already convert. The case for GEO here is the trajectory plus the fact that very few device stores have made their regulatory data machine-readable, which leaves the window wide open.
How AI actually recommends a medical device
The query shapes are specific, and they almost always pair a clinical goal with a status or spec constraint. A shopper rarely asks for a brand of oximeter; they ask “best FDA-cleared pulse oximeter for home use”, “TENS unit for sciatica”, or “OTC hearing aid for mild hearing loss”. The retriever matches that question shape to your page before the model writes a word.
The mechanics are consistent:
- Most device prompts pair a use with a constraint: “FDA-cleared X for [condition]”, “OTC X for [population]”, “X with [measurable spec]”. Retrievers match the question to your status, intended use and spec data, and filter out products whose status or indication is not machine-readable, or whose claim exceeds what is cleared.
- Regulatory status is the highest-value fact, and it is usually missing. Most device PDPs state a marketing benefit but never put the FDA class, the 510(k) or De Novo number, or the CE mark and MDR class in crawlable text. A model that cannot confirm the status has less reason to recommend the device for a clinical use.
- Specs and the instructions for use carry the rest. Accuracy ranges, measurement units, output limits, fit and sizing, battery and connectivity, and the contraindications in the IFU are what let a model match a device to a constrained query and answer the follow-up questions safely.
One rule is important enough to design around: engines disagree, so treat them as separate surfaces. An oximeter or brace that ChatGPT names can be refused by another assistant when a claim brushes the edge of the cleared indication, because each model brings its own caution to a health claim. Run your category queries across at least ChatGPT, Perplexity, Gemini and Claude instead of tuning for one.
Here is the mapping AI assistants most often draw between a shopper goal and the device facts they look for:
| Shopper goal | Device facts AI looks for |
|---|---|
| Home vitals monitoring | Pulse oximeter / blood pressure monitor, FDA clearance, accuracy range, cuff size or finger range, validated protocol |
| Pain relief at home | TENS / EMS unit, cleared intended use (for example “symptomatic relief of chronic pain”), output specs, electrode placement, contraindications (pacemaker, pregnancy) |
| Hearing support | OTC hearing aid, FDA OTC category compliance, intended for perceived mild to moderate loss in adults 18+, output limit, fit |
| Joint or back support | Brace / orthosis, intended use (support vs treatment), sizing, condition fit, whether it is a device or a general wellness product |
| Temperature / illness | Thermometer, clinical accuracy, measurement site, clearance, age range |
| Mobility / daily living | Aids cleared as devices vs general-use products, weight capacity, dimensions, safety standards |
If your product is cleared or CE-marked for one of these uses, say so explicitly and stay inside the indication: “FDA-cleared pulse oximeter, 510(k) cleared, accuracy +/- 2% SpO2, for adults” is the sentence a model can match and trust. “Hospital-grade health monitor” with no status and no spec is not.
The 7 on-page levers for medical devices
These are the edits you control on your own Shopify product pages, sequenced so the cleared-status fix that decides the most recommendations sits first. They are content and structured-data work, not theme rewrites.
1. State the regulatory status and the exact cleared intended use as text
This is the highest-leverage fix in the entire category, and it is the one almost no device store does well. The single fact an AI most wants, and a shopper most needs, is whether the device is authorized for the use being asked about.
Put it in the HTML as plain text, per market:
- US: the FDA class (I, II, III), the clearance pathway and number where applicable (510(k) number, or De Novo), and the exact cleared intended use statement. Link to the public FDA listing where one exists.
- EU: the CE mark, the device class under MDR 2017/745 (I, IIa, IIb, III), and the intended purpose. Name the notified body where relevant.
Then mirror it in structured data. Schema.org has a dedicated MedicalDevice type, part of the core vocabulary, with contraindication, adverseOutcome, seriousAdverseOutcome, procedure, preOp and postOp, plus the inherited legalStatus and recognizingAuthority from MedicalEntity (Schema.org, MedicalDevice). Carry the regulatory facts in additionalProperty so they are unambiguous:
{
"@type": ["Product", "MedicalDevice"],
"name": "Fingertip Pulse Oximeter Model PO-200",
"legalStatus": "FDA 510(k) cleared (K-number), Class II",
"recognizingAuthority": "U.S. Food and Drug Administration",
"contraindication": "Not for continuous monitoring. Readings may be affected by nail polish or poor circulation.",
"additionalProperty": [
{ "@type": "PropertyValue", "name": "Intended use", "value": "Spot-check measurement of SpO2 and pulse rate in adults" },
{ "@type": "PropertyValue", "name": "FDA class", "value": "Class II" },
{ "@type": "PropertyValue", "name": "Clearance", "value": "510(k) cleared" },
{ "@type": "PropertyValue", "name": "SpO2 accuracy", "value": "+/- 2% (70-100%)" }
]
}This is the single highest-leverage device fix and the biggest lever for AI visibility on Shopify, and it is exactly what Verity Score checks for the medical device vertical: whether your regulatory status and intended-use data are present and exposed as text and structured data, not buried in a spec-sheet image or implied by an unverifiable badge.
2. Publish measurable specs, not adjectives
Specs are the facts a model reasons over to match a device to a constrained query. “Accurate” is an adjective; ”+/- 2% SpO2 accuracy across 70 to 100%” is a fact. “Powerful” is an adjective; “output up to 80 mA into 500 ohms, 20 levels” is a fact. For each device put the clinically relevant numbers in HTML text and in additionalProperty: accuracy and measurement range, units, output limits, fit and sizing (finger range, cuff circumference, ear-tip sizes), battery life, connectivity, materials, and the safety or measurement standards it meets. For OTC hearing aids, the FDA’s own output limits (a general ceiling of 111 dB SPL, up to 117 dB SPL with input-controlled compression) are exactly the kind of spec that signals category compliance (Federal Register, August 2022). These tokens separate a device a model can recommend for a specific need from one it cannot place.
3. Make contraindications and the instructions for use accessible
For a device, the warnings are not fine print; they are content a careful model wants before it recommends. State the contraindications in plain text on the page (a TENS unit’s pacemaker and pregnancy warnings, an oximeter’s limits with nail polish or poor perfusion, a brace’s “not a substitute for treatment”), and make the instructions for use reachable as crawlable HTML or a linked document, not locked inside a single image. The EU now lets manufacturers provide electronic instructions for use for most devices under Commission Implementing Regulation (EU) 2021/2226, as amended in 2025, on the manufacturer’s website in the required language, which is a clean way to make the IFU both compliant and machine-readable (EUR-Lex, Reg 2021/2226). Carry the key warnings in the schema contraindication and adverseOutcome fields. A model that can see the safety information is more comfortable naming the product than one that cannot.
4. Keep every claim inside the cleared intended use
This is the lever generic GEO guides skip, and it is the strongest and most dangerous one for devices. The convergence to internalize: what the regulator forbids, the AI refuses; what the cleared indication permits, the AI can repeat.
In the US there is no structure-function exception for devices. Promotional labeling must match the cleared or approved intended use, and a claim beyond it is misbranding because it establishes a new intended use (21 CFR 801.6). The FTC separately requires that objective health claims be backed by competent and reliable scientific evidence, regardless of FDA status (FTC, December 2022). In the EU, claims must stay within the intended purpose covered by the CE certificate under the MDR. The enforcement is current and consumer-facing: the FDA’s 2025 device letters repeatedly cited claims outside cleared intended use, and the WHOOP blood-pressure case shows a wellness framing does not save a feature that reads as a diagnostic claim (Covington, March 2026).
Holding each claim to the cleared indication keeps a device both lawful and AI-recommendable:
| Defensible (inside the cleared / CE intended use) | Risky / forbidden (beyond the clearance) |
|---|---|
| “FDA-cleared pulse oximeter for spot-check SpO2 measurement in adults" | "detects COVID, pneumonia and heart disease early" |
| "OTC hearing aid for perceived mild to moderate hearing loss, adults 18+" | "restores normal hearing”, “cures hearing loss" |
| "TENS unit cleared for symptomatic relief of chronic intractable pain" | "cures sciatica”, “heals nerve damage”, “replaces physical therapy" |
| "Back support brace for lumbar support during activity" | "corrects scoliosis”, “treats herniated discs" |
| "Blood pressure monitor, clinically validated, for home measurement" | "diagnoses hypertension”, “prevents stroke” |
Verity flags benefit claims that cross beyond a device’s stated intended use or have no backing an AI could verify, which is the same gap a regulator would catch. See our claims and proof guide for the verification loop.
5. Distinguish a regulated device from a general wellness product, and tag the population
A large share of confusion (and enforcement) sits at the line between a medical device and a general wellness product. If your product is a cleared device, say so and stay inside the indication. If it is a general-use comfort or wellness product, do not borrow medical-device language (“clinical”, “diagnostic”, “treats”) that would reclassify it, which is precisely the trap WHOOP’s “medical-grade” framing fell into. Then tag the target population and use conditions on every SKU: who it is for (adults, age range), who should not use it (the contraindications), and the conditions of use. Put it in prose and in the targetPopulation-style additionalProperty and the schema contraindication field. Build use-led landing pages (“FDA-cleared devices for home vitals”, “OTC hearing aids for mild hearing loss”) that mirror how shoppers phrase queries and link to the matching SKUs.
6. Use an answer-first title and description formula
Lead with the answer, then layer detail. A workable title formula: brand + device type + regulatory status + key spec + intended use. For descriptions, layer an identity block (what it is, its cleared use, who it is for, in 50 to 75 words), then full specs (status, accuracy, output, fit, standards), then contraindications and “who should not use it”, then setup and what is in the box.
Weak: “Advanced Health Monitor. Hospital-grade accuracy for total peace of mind. Suitable for the whole family.”
Strong: “Fingertip Pulse Oximeter PO-200, FDA 510(k) cleared, Class II, +/- 2% SpO2 accuracy, for adult home spot-checks. A clip-on oximeter that measures blood oxygen saturation and pulse rate in adults, cleared for spot-check use, not continuous monitoring. For: home wellness tracking, recovery, altitude. Not for: infants, continuous monitoring, or as a substitute for medical diagnosis. Readings can be affected by nail polish or poor circulation. Includes device, lanyard, two AAA batteries.”
7. Answer the real questions in FAQPage schema, and handle reviews carefully
Add six to eight Q&As per PDP, wrapped in FAQPage structured data, answering what device shoppers actually ask AI: “Is this FDA-cleared / CE-marked?”, “What is it cleared to do?”, “How accurate is it?”, “Can I use it with a pacemaker / during pregnancy?”, “Is it for adults only?”, “Does it need calibration?”, “What is the warranty?”. Each answer should carry a specific fact and stay inside the intended use. On reviews, take extra care: testimonials that claim the device diagnosed or cured a condition are both a citation risk and an FTC problem, since the FTC treats health claims made through endorsements as the seller’s own claims (FTC, December 2022). Surface genuine reviews about usability, fit, accuracy and value, and do not feature outcome testimonials that imply an unapproved use. This is the same pattern described in our conversational content guide.
One prerequisite underpins all seven levers: your reviews and structured data have to live in the server-rendered HTML. Most AI crawlers never execute JavaScript, so an accuracy rating or star count that a review widget injects only after load is simply absent for them, and that rating belongs on the Product as AggregateRating, not on the Organization (Google treats site-wide self-ratings as self-serving and ineligible for rich results). Verity detects JavaScript-only reviews and checks AggregateRating against Google’s policy. See reviews and AI.
The technical layer: feed, crawlers, schema
The content levers above do most of the lifting for a device store. These technical points are where stale advice keeps getting passed around, so here is what the official documentation actually says in 2026.
Identifiers and the product feed. On a medical device the GTIN that AI shopping surfaces and Google read is the very same number that, once an FDA-accredited issuing agency like GS1 assigns it, doubles as the Device Identifier half of the FDA Unique Device Identifier (GS1, UDI; FDA, UDI Basics). Keep that GTIN identical across the product data and the schema gtin field. On Shopify, your product data already flows into ChatGPT through Shopify’s catalog with no separate feed to build, per OpenAI’s merchant documentation; should you upload a feed instead, OpenAI accepts Parquet, JSONL, CSV and TSV (not XML) and treats GTIN as optional, though for a device it sharpens the match. Hold the feed and the live page to the same values, since the model reconciles one against the other. See our walkthrough on selling on ChatGPT for Shopify.
Perplexity Merchant Program. Joining carries no fee, draws on the Shopify integration for US-shipping stores, and lists device cards without sponsorship. More on Perplexity Shopping.
robots.txt. Allow OAI-SearchBot, ChatGPT-User, PerplexityBot and Googlebot at minimum. Plenty of device sellers still believe blocking GPTBot pulls them out of ChatGPT, yet GPTBot and Google-Extended govern training, not search visibility, which runs through OAI-SearchBot. Verity probes each AI crawler tier (search, user, training) against your robots.txt. See robots.txt for AI crawlers.
Schema.org. Use the MedicalDevice type (core schema, a subtype of MedicalEntity) for the clinical fields, combined with Product for the commerce fields: gtin, brand, offers, aggregateRating, hasMerchantReturnPolicy and shippingDetails. The clinical properties that matter most are contraindication, adverseOutcome, seriousAdverseOutcome, legalStatus and recognizingAuthority, with the regulatory status and intended use carried in additionalProperty. Full detail in our schema.org for Shopify guide.
Off-site: where device AI authority is really built
Since an assistant checks a device’s cleared status and accuracy claims against records you do not control, off-site presence is part of GEO, not a side track from it.
Verifiable public records are the strongest off-site signal. A clearance an AI can confirm against the FDA’s public 510(k) database, or a CE certificate traceable to a named notified body, does more than any on-site badge, because it is independently checkable. Make sure your device is findable in those records and that your page references the same number.
Clinical and editorial coverage compounds it. Reviews from audiology, medical-device or consumer-health publications, validation studies, and category roundups that name your device for its cleared use are exactly the corroborating sources models lean on for health topics. Pursue them deliberately, and only for the indication you actually hold.
Marketplace and retailer reviews feed matched recommendations, with a caution. Reviews that mention the real-world use, fit and accuracy are the ones AI extracts to match a query. But the same FTC scrutiny that produced the device enforcement of 2025 applies to incentivized and outcome-claiming reviews, so encourage honest, specific feedback and avoid anything that reads as a paid or unapproved-use testimonial.
Your 30/60/90 plan
- Days 1 to 30, status and specs. On your hero SKUs, state the regulatory status (FDA class and clearance number, or CE mark and MDR class) and the exact cleared intended use as HTML text, and add
MedicalDeviceschema withlegalStatus,contraindicationand the regulatoryadditionalPropertyfields. Publish measurable specs. Confirm reviews are server-rendered and AggregateRating is on the Product. Check robots.txt allows OAI-SearchBot, ChatGPT-User, PerplexityBot and Googlebot. - Days 31 to 60, claims and IFU. Audit every claim on every device page against the cleared intended use and remove or rewrite anything that exceeds it, including any wellness-vs-device line that reads as a diagnostic claim. Make contraindications and the instructions for use accessible as crawlable text or a linked document. Add six to eight FAQs per hero PDP, and prune any review or testimonial that implies an unapproved use. Build two or three use-led landing pages.
- Days 61 to 90, authority and measurement. Confirm your devices are findable in the relevant public registry and that your pages cite the same clearance. Pursue two or three clinical or editorial placements for the cleared use. Test your category queries monthly across ChatGPT, Perplexity, Gemini and Claude, and track whether you appear, in what position, and whether the status, indication and specs are reported accurately. Google Search Console’s generative AI performance report gives a free first-party view in the markets where it is active.
How Verity Score fits in
Verity Score reads a Shopify store the way an AI assistant checking a device against its cleared use would, and the medical device vertical is built in. It checks whether your regulatory status and intended-use data are present and structured rather than trapped in an image, flags benefit claims that go beyond a device’s cleared indication or have no backing data, detects reviews that load only via JavaScript, validates AggregateRating against Google’s self-serving rule, probes which AI crawlers your robots.txt allows, and scores the completeness of your product record. Each finding comes with the fix.
Medical devices are a category where the same data discipline serves two masters at once: the regulator who decides whether your claims are lawful, and the model that decides whether your device gets named. The brands structuring their status, intended use, specs and contraindications as clean, machine-readable data now, and never claiming an inch beyond what they are cleared to claim, are the ones AI will recommend when a shopper asks for an FDA-cleared oximeter or an OTC hearing aid for mild hearing loss.
Want to know whether AI can confirm your devices against their cleared indication? Run a free GEO audit in 60 seconds.