Hospital procurement teams, biomedical engineers and clinicians already use AI assistants to shortlist equipment. They ask what a device does, how two models compare, whether it is approved in India and who sells it. The answers draw on spec sheets, distributor pages, regulator listings, review sites and comparison articles. If your own pages are not among them, someone else is describing your device.
This article is for marketing and regulatory leads at medical device and medtech companies, and for agencies serving them. It explains how to audit which sources assistants use for device queries and how to publish specifications and regulatory status so they can be quoted correctly. It describes a method and reports no results. Claims about regulatory status are sensitive, so your regulatory affairs team should own every statement. For wider context, see our notes on medical device marketing.
Know how India regulates devices
The Central Drugs Standard Control Organisation's medical device and diagnostics page says devices are regulated under the Drugs & Cosmetics Act, 1940 and the Medical Devices Rules, 2017. It states that devices fall into four risk classes under the First Schedule: Class A (low risk), Class B (low-moderate), Class C (moderate-high) and Class D (high). It lists licensing, import and manufacturing applications through the Sugam portal and archived notices on when licensing started for each class, including 1 October 2023 for Class C and D devices.
Why this matters for AI answers: users ask "is it CDSCO approved?" and assistants often answer with a confident yes or no. The plain truth is more specific. The right answer includes the device's risk class, whether the licence is a manufacturing or import licence, and the licence number and holder. Without a clear source, assistants generalise, and sometimes they are wrong.
Build the device query set
Create about fifty queries from the buyer's side. Include the following groups, using your own product category.
| Group | Example pattern | What to check |
|---|---|---|
| Category discovery | "Best [device type] for hospitals in India" | Which brands are named, and from which sources |
| Specification | "[Model] specifications", "[Model] power requirements" | Whether figures match your current datasheet |
| Regulatory | "Is [Model] CDSCO approved?", "[Device] licence India" | Class, licence type and holder named correctly |
| Comparison | "[Model A] vs [Model B]" | Which comparison pages are cited |
| Procurement | "[Device] price in India", "[Device] distributor Delhi" | Which dealers and marketplaces surface |
| Clinical use | "What is [device] used for?" | Whether intended use is described within your label |
| Service | "[Model] service engineer", "[Model] spare parts" | Support information availability |
Run each query on the assistants your buyers use, three times each, and keep the full text and cited URLs. Record the date, surface and whether web search was on.
What to look for
Score each answer against a short list:
- Specification accuracy. Compare every number to your current datasheet. Note discrepancies by model and version, because old model revisions and variants confuse assistants.
- Intended use. Does the answer stay inside your approved intended use, or does it extend it? Extensions are a regulatory risk even when an assistant, not you, wrote them.
- Regulatory status. Does it state a class, licence type, licence number or holder? Is it right? Does it say "approved" without a source?
- Source type. For each cited URL, tag: manufacturer, distributor, regulator, hospital, medical society, review site, marketplace, news, forum, PDF.
- Comparison framing. Whose framework defines the comparison? If a distributor's table frames your device against a cheaper alternative on three criteria, that framing may dominate the answer.
Keep a separate list of every page that is cited for a competitor and not for you. Those pages are your gap map.
Which pages assistants tend to cite
Do not assume. The audit will show which types appear. Common candidates worth checking are manufacturer product pages, PDF datasheets, distributor catalogues, marketplace listings, hospital tender documents that are publicly posted, regulator pages and comparison articles on trade sites. The mix will differ by category. A diagnostic imaging system will draw on different sources from a consumable.
A hypothetical illustration: a fictional infusion pump brand finds that assistants mostly cite one marketplace listing for specifications, and that the listing shows a previous model's flow range. The manufacturer's own page is a scanned PDF with no text, and so is rarely cited. The fix has two parts: publish the specification in HTML with a version and date, and have the distributor correct the listing.
Publish specifications that can be quoted
A device page that assistants can quote accurately is plain and structured.
- Spec table in HTML. One row per parameter, with units. Include model number, variant, version date and what changed from the previous model.
- Intended use statement. The wording from your label, not marketing language. Link to the instructions for use where public.
- Regulatory status block. Risk class under the Medical Devices Rules, 2017; licence type; licence number; licence holder; date verified. Only include what your regulatory team can document, and update it when the licence changes.
- Claims discipline. Avoid "CDSCO approved" shorthand unless regulatory affairs confirms it is accurate for the specific device and the context. State the facts instead.
- Downloads. Provide the datasheet as a PDF that has a text layer, linked from an HTML page that repeats the key facts.
- Variants and accessories. Separate pages or clearly labelled sections for each, so a specification for one is not attributed to another.
- Contact and service. Service network, response expectations that are real and demonstrable, and spare parts availability.
On markup, schema.org's MedicalDevice type is narrow. Its own properties cover items such as contraindication, adverse outcome and procedure, and it inherits fields such as legalStatus and recognizingAuthority from MedicalEntity. For a commercial product page, many teams find Product markup with named properties more practical for specifications, with MedicalDevice where its fields apply. Whichever type you use, mark up only what is visible on the page. Our schema markup service page explains how we approach this.




