Strategic Cites: 3 Must-Know Factors That Drive AI Engines

A week of arguments about the MedTech AI Visibility Index has clarified something the ranking alone did not show. Engines are not applying a single standard when they decide who to cite. They are applying three, and different categories run predominantly on different ones. This matters commercially because most organizations optimizing for AI visibility are spreading effort evenly across all three. This means they are underfunding the one factor that actually decides their category and overfunding two that will not move their position for eighteen months.  

Factor one: entity clarity

Can the engine state what your company is without contradicting itself across sources. This is the floor. Below it, nothing else works, because attribution requires an entity to attribute to. An engine that cannot resolve your company with confidence will attribute the claim to a competitor it can describe cleanly, regardless of the quality of your content. The mechanics are covered in detail in the piece on entity consistency. The short version is that MedTech companies generate identity signals across an unusually wide set of sources, each with its own naming convention, and five or six variants of one company is normal in the category.

How to tell if this is your constraint

Ask two engines who your company is and compare the answers. If they differ materially, entity clarity is your constraint and nothing else on this list is worth funding until it is fixed.

What it costs to fix

Less than anything else in this article. Reconciling names across ten sources is administrative work of roughly ninety minutes plus follow-up. Adding schema is a short technical task. There is no content budget, no agency, and no eighteen-month horizon involved.

Factor two: claim structure

Is your answer extractable as a self-contained block, or is it dissolved into prose written for a human who scrolls. Engines cite what they can lift cleanly. A page that states a question plainly and answers it in one block that stands alone will be cited more often than a page of higher quality whose conclusion is distributed across four paragraphs and dependent on the surrounding context. This is uncomfortable for anyone who tracks editorial quality to be cited, and it is nonetheless how retrieval works. The practical implication is that a well-structured piece of moderate depth can outperform a superior piece written as a continuous argument.

What extractable actually means

•  The question is stated in the words a buyer would use, not in category jargon •  The answer is complete without the paragraph before it or after it •  Any qualification sits inside the answer block rather than in a later caveat •  The claim is stated once in a form that could be quoted, rather than built up across a section This is also why explicitly labelled question and answer blocks retain value even though Google deprecated FAQ rich results in May 2026. The search feature is gone. The format is close to ideal as an input for a system assembling an answer, and the FAQPage schema type remains valid and parseable.

How to tell if this is your constraint

You are entity-clean, you are being named in entity queries, and you are still absent from category and procedure answers. That pattern means the engine knows who you are and cannot find a liftable claim on your pages.

Factor three: third-party echo

Does anyone independent state the claim in language close to yours. A claim that appears only in material your company produced or funded reads as marketing. The same claim appearing in a society statement, a registry analysis, and an independent review reads as established fact, and engines weight it accordingly. This is the slowest of the three factors to build and by a wide margin the most durable once built. It is also the one that cannot be bought directly, which is why it is systematically under-invested in by organizations that plan in annual budget cycles.

Where echo comes from in healthcare

Professional society guidance. Registry analyses. Guideline working groups. Editorial boards. Independent review articles indexed in PubMed. These are relationship and contribution channels rather than marketing channels, and the organizations that hold positions in them are usually the ones who have been contributing for years without treating it as a visibility activity.

How to tell if this is your constraint

The companies cited in your category are neither notably entity-clean nor structurally sophisticated, but are consistently referenced by societies, registries or trade press. If that is the pattern, content and schema work will not move your position and you should stop funding them as if they will.

Finding out which factor your category runs on

The diagnostic takes an afternoon and it is the single most useful piece of analysis available before a 2027 visibility budget is set.

Step one: identify the three most cited companies in your category

The most cited, not the largest. Those are frequently different, and the difference is itself informative.

Step two: check each of them against all three factors

Entity clarity: ask two engines who they are and compare. Claim structure: look at whether their key claims are stated in extractable blocks. Third-party echo: search for their central claim in sources they do not control.

Step three: read the pattern

All three entity-clean but structurally ordinary means your category runs on entity clarity. This is the cheapest outcome and the most common in newer categories. Structurally excellent means your category runs on claim structure. The fix is a content architecture project of roughly a quarter, and it is largely within your control. Neither, but heavily echoed by societies and registries, means your category runs on third-party echo. This is eighteen months of contribution work, and the correct response is to start it now rather than to fund a content programme that will not move the position.

Why the answer differs by category

The pattern that appears to explain most of the variation is the depth of the literature underneath a category. Where a procedure category is well established and the published evidence base is deep, third-party echo dominates, because there is a great deal of independent material for an engine to draw on and company-produced content competes against it at a disadvantage. Where a category is newer and the literature is thin, entity clarity and claim structure dominate, because the engine has less independent material available and will draw on whatever it can resolve and extract. This has a direct planning consequence. A company in a new category running the playbook designed for an established one will spend a year building echo it cannot yet build, while the entity position that was actually available goes to a faster competitor.

What to do with this before the 2027 budget

One. Run the three-factor diagnostic on your category. An afternoon. Two. Fund the factor that decides your category first, disproportionately, rather than splitting the budget evenly across three. Three. If the answer is third-party echo, say so in the plan and set the expectation on an eighteen-month horizon. A contribution strategy funded on a two-quarter expectation will be cancelled at the point it starts working. Four. Whatever the answer, fix entity clarity anyway. It is the floor, it is cheap, and no other factor produces a return while it is unresolved.  

Frequently Asked Questions

How do AI engines decide who to cite?

Three factors dominate. Entity clarity, meaning whether the engine can state what a company is without contradiction across sources. Claim structure, meaning whether an answer is extractable as a self-contained block. And third-party echo, meaning whether independent sources state the claim in similar language. Different categories run predominantly on different factors.

Which factor matters most?

Entity clarity is the floor, because attribution requires a resolvable entity. Beyond that floor, the dominant factor varies by category, and it appears to track the depth of the published literature underneath the category.

How do I find out which factor my category runs on?

Identify the three most cited companies in your category, not the three largest. Check each against all three factors. If they are entity-clean but structurally ordinary, the category runs on entity clarity. If they are structurally excellent, it runs on claim structure. If neither but heavily referenced by societies and registries, it runs on third-party echo.

What makes a claim extractable?

The question is stated in the words a buyer would use, the answer is complete without the surrounding paragraphs, any qualification sits inside the answer block, and the claim is stated once in a quotable form rather than built up across a section.

Does FAQ schema still help after the May 2026 deprecation?

Yes, for different reasons. Google stopped showing FAQ rich results in search on 7 May 2026, but the FAQPage type remains valid and parseable, and an explicitly labelled question and answer block is close to an ideal input for a system assembling an answer.

What is third-party echo?

Independent sources stating a claim in language close to the company’s own. In healthcare this comes from professional society guidance, registry analyses, guideline working groups, editorial boards and independent review articles. It is the slowest factor to build and the most durable once built.

Why does the dominant factor differ between categories?

It appears to track the depth of the literature underneath the category. Established categories with deep evidence bases are dominated by third-party echo. Newer categories with thin literature are dominated by entity clarity and claim structure.

What happens if you optimise for the wrong factor?

A company in a new category running the established-category playbook spends a year attempting to build echo it cannot yet build, while the entity position that was actually available goes to a faster competitor.   EXTERNAL CITATIONS •  Google Search Central, FAQ structured data and the May 2026 deprecation •  PubMed