Your KOL contract buys podium time, advisory hours, and a name on a paper. It does not buy citations, and citations are now what your buyer encounters first.
That sentence describes a gap that opened in about three years and that no standard medical affairs contract has yet been rewritten to address.
The sequence changed, and the contract did not
The surgeon who would have met your key opinion leader at a congress in 2019 now asks an AI engine about the procedure before the congress, before the sales call, and before any material you produced enters the picture at all.
Whether your KOL’s published work is what surfaces in that answer determines whether the relationship is compounding or merely occurring. And that outcome is decided by citation architecture, which is not a function anyone in a medical affairs organization currently owns.
Medical affairs measures reach, engagements, and sentiment. All three are real, all three are properly constructed, and all three are blind to this. A dashboard can report green for four consecutive quarters while the underlying work becomes structurally invisible at the layer the buyer now consults first.
Why this is not a measurement problem you can solve with a new metric
The temptation is to add a citation metric to the existing dashboard and continue. That will tell you the size of the gap and it will not close it, because the causes sit outside medical affairs entirely: in how affiliations are entered on papers, in how conclusions are structured, and in whether independent sources echo the claim.
The metric is worth having, but only as the opening of a conversation with commercial and with the KOLs themselves.
Cost per citation
Take your total KOL spend for the last twelve months. If you do not have a clean internal figure, CMS Open Payments publishes reported payments from medical device and drug companies to physicians and teaching hospitals, so a defensible external number is available for you and for every competitor in your category.
Now count the number of times an AI engine surfaced work by those individuals in response to your three core buyer queries: the category question, the entity question, and the procedure question. Run each on at least two engines, twice, on separate days.
Divide the first by the second. In most MedTech organisations the resulting figure is large enough to change the tone of a renewal conversation, and in a meaningful number it is undefined because the denominator is zero.
This is not an argument for reducing KOL investment. Your key opinion leaders are producing work of real clinical quality, and the relationships carry value that predates any engine and will outlast the current generation of them. The argument is that you are already spending the money and capturing a fraction of the available return, and that closing the gap is an architecture problem rather than a relationship problem.
Why some KOLs are cited and others are not
The difference is rarely the quality of the science. Across the audits I have run it is usually one of three structural factors, and all three are fixable.
1. Affiliation consistency
Author affiliations on PubMed-indexed papers are entered by hand, differently, by different co-authors, across years. Where a KOL’s affiliation to your company or to their institution varies across their publication record, the citation signal fragments in exactly the way a company name does so.
This is the same failure described in the entity consistency article, operating one level down. Engines’ weight author affiliations vary heavily in clinical contexts, and this is the source over which most companies exert no control at all, because nobody has ever asked a KOL to standardize how they are described.
2. Structural extractability
A review paper with a clear structure, an explicit answer to a common clinical question, and a conclusion stated in one self-contained block will be cited far more often than an original research paper of higher scientific value whose conclusion is distributed across a discussion section.
Engines cite what they can lift cleanly. This is uncomfortable for anyone who believes citation should track scientific merit, and it is nonetheless how retrieval works. The practical consequence is that a well-structured review commissioned deliberately can outperform three years of primary research in this specific channel.
3. Independent echo
A claim that appears only in work your company funded reads as sponsored. The same claim appearing in a society statement, a registry analysis, and an independent review reads as established.
Third-party echo is the slowest of the three to build and by a distance the most durable once built. It is also the factor most likely to be already claimed in your category, and it is worth checking before you fund anything else.
The check that tells you which factor is yours
When one KOL on a roster produces citations and the others do not, the difference is almost always one of these three. Find the one who does, work out which factor explains it, and the fix is usually replicable across the rest of the roster for very little money.
The society problem, and when contribution beats competition
One important qualification. In categories where a professional society has published guidance, the citation position at the top of the category is frequently already held by the society itself, and no company outranks a guideline.
In those categories the strategy is contribution rather than competition. Registry data. Guideline working group participation. Society-published review work. It is slower, it is less controllable, and it produces a position that is considerably harder for a competitor to displace.
Check whether a society has published in your category before funding anything. Ten minutes. It should be the first question in any KOL or visibility proposal you are shown, and it almost never is.
What to add to the next contract
Two clauses. Neither is unusual, neither imposes meaningful additional burden on the KOL, and neither currently appears in standard MedTech agreements.
A publication commitment with structural requirements
Not simply a paper, but a paper with a defined structure: an explicit question stated in the abstract, a conclusion given as a self-contained statement, and a stated relationship to the existing evidence base. This costs the author nothing additional in effort and changes the citation outcome substantially.
Attribution consistency
The affiliation as it will appear, specified in the agreement, matching your canonical entity description exactly. One line. It repairs the single most common source of fragmentation in the category and it is the clause with the highest ratio of effect to effort available to you.
The first company in a device category to write citation terms into a KOL agreement will be copied within a year. There is a durable advantage in being that company, and the window is open right now precisely because nobody is doing it.
The diagnostic to run before the renewal
Ask an engine your lead procedure question. Read the answer carefully. Three outcomes, each with a different action.
Your KOL’s work is present
Find out precisely why. It will be one paper or one institutional page doing the work. That structure is replicable and you have just found your template. Document it before the person who wrote it moves institution.
A competitor’s KOL is present and yours is not
This is a citation position being built against you while your medical affairs reporting shows healthy engagement. It is the most actionable finding available in this exercise and the most commonly missed, because nothing in your existing reporting surfaces it.
Nobody’s work is present
The category position is unclaimed. That window closes, usually within eighteen months, and usually to whoever moves first rather than to whoever has the strongest science. This is the only one of the three outcomes with a deadline attached.
Q4 and Q1 are when most KOL agreements renew. The diagnostic takes an afternoon, and it is the difference between renewing on last year’s terms and renewing on terms that account for how your buyer now behaves.
Frequently Asked Questions
How do you measure KOL ROI in MedTech?
Traditional measures are reach, engagements, and sentiment. These do not capture whether the KOL’s published work surfaces when a clinician consults an AI engine, which is increasingly the first step in evaluation. Cost per citation, calculated as total KOL spend divided by engine citations across core buyer queries, makes that gap visible.
Why are some KOLs cited by AI engines and others are not?
Three structural factors usually explain it: consistency of author affiliation across the publication record, whether conclusions are stated in an extractable self-contained form, and whether the claim is echoed by independent sources such as societies or registries. Scientific quality is rarely the differentiator.
What should be in a modern KOL contract?
Alongside standard terms, two additions are worth considering: a publication commitment with structural requirements covering the question, the conclusion format, and the relationship to existing evidence; and an attribution consistency clause specifying exactly how the affiliation will appear.
Does medical affairs own AI visibility?
In most organisations nobody does. Medical affairs owns the relationships and commercial owns the digital properties, and citation architecture for published work falls between them. Naming an owner is usually the first substantive step.
What is cost per citation?
Total key opinion leader spend over a period divided by the number of times AI engines surfaced work by those individuals in response to the organisation’s core buyer queries. It is a diagnostic figure rather than a target, intended to make an invisible gap visible before a contract renews.
Where can I find what companies pay physicians?
The CMS Open Payments database publishes payments reported by drug and medical device companies to physicians and teaching hospitals in the United States. It provides a defensible external benchmark for your own spend and for competitors in your category.
What if a professional society already owns the citation position in my category?
Then the strategy is contribution rather than competition. No company outranks a guideline. Registry data, guideline working group participation, and society-published review work build a position that is slower to establish and considerably harder for a competitor to displace.
How do author affiliations affect AI citation?
Affiliations are the mechanism connecting a company to peer-reviewed literature, and engines weight that literature heavily in clinical contexts. Because affiliations are entered manually and vary across co-authors and years, they are one of the most common sources of entity fragmentation and one of the least monitored.
EXTERNAL CITATIONS
• PubMed
• MedTech