A healthcare company appearing in an AI-generated answer is an observation. Whether the answer identifies the right organization, describes the right service and makes medically supportable claims requires a separate assessment. A visibility score cannot do all of that work.
Table of Contents
SV’s analysis of 73 public RankLens Health Care profiles found that 50—68.5% of the roster—display discovery scores of 50% or less. None displays 100% at the profile level. The more instructive findings, however, emerge inside individual reports: Charles River Laboratories International ranges from 50% to 100% discovery across its five recorded service contexts, despite four samples being reported for every row. [1] [2]
This is not evidence that the company’s services are better or worse in one area, nor that patients are receiving incorrect advice. It is evidence of variation within a narrow visibility test. For healthcare communications teams, the appropriate response is to investigate representation by service and audience—not turn the aggregate score into a proxy for trust.
Scope and disclosure: RankLens is an SV product, and SV has a commercial interest in AI visibility measurement and marketing services. This article is a descriptive review of public records, not an independent clinical evaluation. October 1, 2026 is the date of review, not a verified date for the underlying model runs. No patients, treatment outcomes or medical decisions were studied.
The healthcare snapshot is broader than a pharmaceutical ranking
The unweighted mean displayed discovery score is 43.2%, and the median is 45%. The roster includes pharmaceutical, device, laboratory, insurance and care-related businesses under the source’s Health Care classification. It should not be described as a survey of hospitals, a pharmaceutical-only study or a ranking of clinical quality. [1]
| Displayed discovery band | Profile count | Share of roster |
|---|---|---|
| 0% | 4 | 5.5% |
| Above 0%, up to 25% | 11 | 15.1% |
| Above 25%, up to 50% | 35 | 47.9% |
| Above 50%, up to 75% | 17 | 23.3% |
| Above 75%, up to 100% | 6 | 8.2% |
| Total | 73 | 100% |
There is no defensible leap from this distribution to “healthcare brands are missing from most patient answers.” The tested category phrases are not a documented sample of patient questions, and profile averages do not supply the number of valid answers behind every score. Some detailed reports show 0/0 rows, which cannot establish absence from completed responses. [5]
The useful implication is narrower. A broad sector average is an entry point for investigation, not a sufficient measure of whether an organization is represented correctly for the audiences and services that matter to it.
Charles River: one company, different service-level results
Charles River’s report contains five category rows, each with four recorded samples. Four rows show appearances in all four samples. The biotechnology-support row shows two appearances. Together, those observations produce 18 appearances across 20 sample slots and a displayed profile discovery score of 90%. [2]
| Category phrase in report | Recorded appearances | Displayed discovery |
|---|---|---|
| Preclinical and clinical research | 4/4 | 100% |
| Drug development services | 4/4 | 100% |
| Biotechnology support solutions | 2/4 | 50% |
| Laboratory animal management | 4/4 | 100% |
| Contract research organization | 4/4 | 100% |
The difference is worth investigating because it occurs within one profile with complete displayed denominators. But it is not a controlled test of wording alone: the categories can represent different tasks and competitive sets. The report does not isolate why one context produced fewer recorded appearances.
A healthcare communications team could use this pattern to prioritize a more detailed audit. Does the lower-scoring context correspond to an important service? Is the company’s role clearly distinguished from that of suppliers, research partners or unrelated organizations? Are the questions relevant to the people the organization actually serves? These are proposed follow-up questions, not defects established by the score.
The operational benefit of service-level analysis is precision. Instead of launching a generic campaign to make a corporate name more visible, the team can first determine whether a specific information need is underrepresented, inaccurately described or simply poorly measured.
Dentsply Sirona and DexCom: high coverage is still not uniform coverage
Dentsply Sirona displays 90% profile discovery, with 18 recorded appearances in 20 sample slots. DexCom displays 85%, with 17 in 20. Neither result describes every possible question about those businesses. Both reports show variation across the five tested contexts, despite relatively high overall scores. [3] [4]
| Profile | Recorded appearances across five rows | Profile discovery | Entity-row discovery range |
|---|---|---|---|
| Dentsply Sirona | 18/20 | 90% | 75%–100% |
| Charles River Laboratories International | 18/20 | 90% | 50%–100% |
| DexCom | 17/20 | 85% | 75%–100% |
DexCom’s report illustrates how even closely related contexts can produce different observations. Its real-time blood-sugar-tracking row records 4/4 appearances, while the continuous-glucose-monitoring row records 3/4. That one-response difference is 25 percentage points at the entity level. It is a useful reminder of how coarse a four-sample estimate is. [4]
Dentsply’s populated rows likewise range from 3/4 to 4/4. The correct conclusion is that the report contains high recorded coverage across the selected contexts, not that the company is universally preferred or clinically superior. A communications metric becomes less credible, not more, when it is asked to support an outcome it never measured. [3]
Johnson & Johnson shows why a low score needs a denominator
Johnson & Johnson’s profile displays 10% discovery. Its detailed table, however, shows four 0/0 entity rows and one populated row with two appearances out of four samples. The displayed 10% is consistent with averaging the five displayed row scores, but it is not evidence that the company was absent from 90% of completed healthcare answers. [5]
That difference should travel with the finding wherever it is cited. An unresolved denominator cannot tell a reader whether a category was untested, unavailable or subject to another reporting convention. Treating it as a known failure would create a stronger negative claim than the observations support.
The same discipline applies to organizational identity. A future study should establish whether it is testing a corporate parent, a service, a device or a product brand, and how each is matched. Without that policy, a reference to the intended product might be overlooked—or a reference to a related but different entity might be counted as success.
Visibility, confidence and medical accuracy are different questions
RankLens reports measures including discovery, LLM confidence and brand matching. Its definitions describe model-reported confidence and matching to the intended brand or domain. Neither is a clinical validation score. A confident answer may still require factual review, while correctly identifying a company says nothing by itself about the accuracy of a medical claim in the answer. [3]
The World Health Organization has warned that large language model outputs can sound authoritative while containing serious errors, including in health-related responses. That warning supports keeping information quality separate from presentation and confidence. It does not establish that any answer behind this RankLens snapshot contained a medical error; the raw answers were not clinically audited for this article. [6]
For healthcare organizations, the distinction suggests a two-part measurement program. The first part records whether the relevant organization or offering appears in a defined context. The second evaluates whether the answer represents it accurately and, where medical claims occur, whether those claims withstand qualified review. The first can identify material for the second, but cannot substitute for it.
A responsible report should also separate a simple corporate description from advice that could affect a health decision. Those are different levels of consequence. Treating every mention as the same kind of success would hide the very distinctions that make healthcare communications different from a conventional brand-awareness exercise.
Build the next audit around audiences and evidence
The next-stage research design should specify the audience before choosing the questions. A person researching a provider, a laboratory evaluating a service and a professional reviewing a product may need different information. This snapshot does not establish how any of those groups actually uses AI; a follow-up should not assume their questions are interchangeable.
Then define the evidence required for an accurate answer. A proposed review checklist would distinguish identity, service description, source citation and any consequential factual claims. Reviewers should be able to record “not applicable” as well as “correct,” “incorrect” and “insufficient evidence.” A brand mention without a medical claim should not be scored as though a treatment recommendation had been evaluated.
Preserve the underlying responses, the exact questions and the test conditions. Publish how aliases, renamed entities and parent–product relationships are handled. Separate completed non-mentions from failed requests and unresolved records. These steps would allow later reviewers to understand not only the result, but the decisions that produced it.
Finally, route findings to the appropriate owner. An identity problem may belong with corporate communications; an unsupported medical claim requires qualified review; an incomplete record belongs with the measurement team. Increasing the visibility score is not necessarily the remedy for all three.
The useful finding is specificity, not a healthcare popularity contest
The 73-profile snapshot provides a starting point for understanding how healthcare businesses appear across selected AI test contexts. Its strongest examples show that a high corporate score can conceal a lower service-level result, and that a low score can conceal unresolved data.
For healthcare communicators, the goal should therefore be accurate representation in relevant contexts, supported by enough evidence to distinguish a repeatable issue from a small-sample result. Visibility can help locate the question. It cannot, on its own, establish trust, medical accuracy or the quality of care.
Methodology and limitations
SV transcribed selected displayed fields for all 73 records in the supplied Health Care table and calculated an unweighted mean, median, threshold counts and score bands. Profile percentages were not pooled into a response-level rate. Source sector assignments and names were retained rather than independently reclassified or audited for current corporate status.
Charles River, Dentsply Sirona and DexCom were selected to illustrate within-profile results with complete displayed denominators. Johnson & Johnson illustrates unresolved 0/0 reporting. These purposive cases do not form a representative sample of healthcare communications. No observation was invented for a row without a positive sample denominator, and no official score was silently corrected.
Public labels identify ChatGPT responses, but the complete prompts, raw answers, exact model versions, fieldwork dates and sampling independence were not independently verified. Retrieved month labels were inconsistent. October 1, 2026 is consequently a review date only. Four samples per entity provide a narrow descriptive basis; no statistical significance or causal effects are asserted. The data do not measure medical accuracy, product efficacy, patient safety, care quality, professional preferences, traffic or revenue. This article is communications research, not medical advice.
Sources
- RankLens: Health Care company-profile table. Selected displayed fields reviewed October 1, 2026.
- RankLens: Charles River Laboratories International brand-visibility report. Company and entity-level data.
- RankLens: Dentsply Sirona brand-visibility report. Company data, entity table and metric definitions.
- RankLens: DexCom brand-visibility report. Company and entity-level data.
- RankLens: Johnson & Johnson brand-visibility report. Includes four unresolved 0/0 rows.
- World Health Organization: WHO calls for safe and ethical AI for health. Published May 16, 2023.
