The shortlist changes with every word the buyer types

When a process engineer asks an AI engine who builds CIP, SIP, or buffer-prep skids, the engine names a handful of firms — and that handful becomes the shortlist. This study measures exactly who gets named, on which engine, for 27 distinct buyer queries.

Do any fabricators show strength on all four engines at once?

No fabricator in this study shows strength on all four engines at once, and the matrix below is the fastest way to see it. Each row is a real firm leading its lane; each column an AI engine. A firm with a coherent cross-engine position would show four dark cells. None do. Sani-Matic and CSI own three engines and collapse on ChatGPT; IPEC and the ChatGPT specialists invert the pattern.

Read the table by row rather than by column. A row of four dark cells would mean a firm that a process engineer finds whichever assistant they happen to open, which is the position every fabricator assumes it already occupies. What the data shows instead is a set of firms that are dominant in one place and invisible in another, with no obvious relationship between the two. The pattern is not a ranking of quality. It is a map of which pages each engine could read.

Heat scales with share of voice: 0 → 25 → 50 → 75 → 100. Darker is stronger.

Does AI reward documented capability or brand reputation?

AI rewards documented capability, not reputation. Across all three lanes in this study — and consistent with our first report — between 70 and 79% of the citations behind AI answers point to fabricators' own capability pages, not directories, trade media, or third-party rankings.

The first report established this across ten categories at one query each. The deeper query set confirms it: when a process engineer asks an AI engine who builds CIP skids, the engine assembles its answer largely from what fabricators have published about themselves. Being named is not a matter of decades of standing a firm cannot change. It comes down to whether a firm's pages state, in clear and specific language, exactly what it builds and to what standards.

The firms that get named are the ones whose pages document their work. The firms that are absent usually describe it in general terms — "custom stainless fabrication," "quality process systems." A credentialed fabricator can be invisible to AI not because the engines judged it inferior, but because they could not read what it does. The gap is in the pages, and the pages are within the firm's control.

How often does a single firm hold the shortlist on every engine?

A single firm holds the shortlist on every engine in only seven of the 27 buyer queries in this study, and most of those seven belong to a large supplier rather than a fabricator. In the other 20, no fabricator holds a strong position on all four engines at once. And in 76% of shortlist placements, even the firms that are named score at or below 20% on at least one engine.

The matrix above is the study in one image, and the numbers behind it are blunt. Of the 27 queries, only seven produced a firm commanding all four engines — and most of those are dominated by a single large supplier, not a fabricator. Everywhere else, visibility is split: a firm leads two or three engines and collapses on the rest.

This is not noise. The four engines draw on different indexes and weight sources differently — ChatGPT leans heavily on the Bing index, Perplexity and Google move on their own crawls, and Claude favors deeper corroboration. A firm optimized for one engine is not automatically legible to the others. The practical consequence: a fabricator can be the most-cited name in its lane on three engines and still be absent from a third of the AI answers buyers actually see.

How does the buyer's phrasing change which fabricators get named?

The buyer's phrasing decides the shortlist, and it does so more forcefully than anything the fabricator controls on the day. Queries phrased in specific, technical language — "cGMP," "single-tank," "inline dilution," "sterilize-in-place" — surface fabricators with displaceable positions. Queries phrased generically — "system suppliers," "manufacturers USA," "single-use" — surface the large equipment and consumables brands and consolidate around them.

This is the finding a fabricator can act on directly, because it says where to compete. In the CIP lane, the broadest query — "best CIP skid fabricators for pharmaceutical manufacturing" — is the most open of all fourteen: no firm holds more than 55% across this study's runs, with IPEC narrowly ahead and soft positions beneath. This is the same query our first report measured, and the rosters agree. But reword the same intent as "leading automated CIP system vendors" and the answer consolidates instantly around GEA, which holds 100% on every engine.

The reason is structural. "System supplier" and "single-use" are the vocabulary of branded platforms and consumables — the language pulls the engines toward Cytiva, Sartorius, GEA, and their peers. "Fabricator," "builder," and a named standard are the vocabulary of custom shops — and that is where a credentialed fabricator's own capability page can win the citation. The lane is not locked or open. Each query is, and the buyer decides which by how they ask.

Why are Sani-Matic and CSI missing from ChatGPT in the CIP lane?

Sani-Matic and CSI are missing from ChatGPT across almost the entire CIP lane while leading the other three engines, a pattern that points at the index rather than at the firms. We measured the absence, not its cause; the likeliest explanation, set out below, is that ChatGPT's web results lean on the Bing index. The most common visibility failure is not weak performance everywhere. It is total absence on one engine while leading the others. In CIP, the two most-cited fabricators lead Google, Perplexity, and Claude across nearly every query — and disappear on ChatGPT across almost all of them.

Sani-Matic is named in the top three on Google, Perplexity, or Claude in twelve of the fourteen CIP queries, and is the outright lane leader by combined share of voice. Yet it scores zero on ChatGPT in eleven of those fourteen queries. Central States Industrial shows the same fracture even more starkly: present across the lane on three engines, and zero on ChatGPT in thirteen of fourteen. These are not marginal firms — they are the lane's leaders, missing from one entire engine across their whole product line.

A failure that uniform almost always has a single technical cause. ChatGPT's web results lean on the Bing index; a firm that is thin in Bing, or that blocks the relevant crawler, goes dark on ChatGPT specifically while ranking normally everywhere else. It is the most fixable kind of invisibility — one cause, one engine, an entire lane of queries recovered.

Which of the three skid lanes is easiest for a fabricator to win?

SIP is the easiest of the three lanes for a fabricator to win, because it is the least consolidated: no firm exceeds 29% combined share of voice and every SIP query in the study contained a displaceable position. CIP is harder but still open at the specialist end. Buffer preparation is the most locked, anchored by the large single-use platforms. The leaderboards below show each contest separately, because a firm's standing in one says almost nothing about its standing in the others.

Scored by how many shortlist positions are soft and displaceable — rather than by who happens to sit at the top — the three lanes sit at very different points. Each leaderboard shows combined share of voice across that lane's queries, by engine.

CIP — Clean-in-Place

SIP — Sterilize-in-Place

The least consolidated lane in the study. Every SIP query contained displaceable positions, and outside the generic head term no large supplier owns the field. For a credentialed fabricator with structured pages, this is where movement is fastest, because there is barely an incumbent to displace.

Why is buffer preparation the hardest lane for a fabricator?

Buffer preparation is the hardest lane for a fabricator because the answers are dominated by the large single-use and consumables platforms rather than by custom shops. The leaderboard below shows the contest, and the pattern is different in kind from CIP and SIP rather than different in degree.

Buffer Preparation

The most consolidated of the three. Buffer-prep answers are dominated by the large platform suppliers — Cytiva, Sartorius, Thermo Fisher, Merck — whose single-use and consumables businesses pull the lane toward branded products. A specialist fabricator can still hold a defensible position on specific queries like inline dilution, where Exentec leads, but the generic and single-use queries belong to scale.

What happens when one firm trades under two names?

When one firm trades under two names, the engines treat it as two firms and split its share of voice between them. IPEC and Garrett Industrial Systems are the same company, and the engines treat them as two, splitting the firm's CIP share of voice across both identities.

In this study's CIP leaderboard, IPEC appears at 34% combined and Garrett Industrial at 20%, separately. They are one fabricator. Consolidated, the firm would rank among the top three in CIP; split, it sits fourth and tenth, diluting its authority on every engine and confusing the entity signal the engines rely on to decide whom to name. Entity fragmentation is one of the few visibility problems a firm can fix without producing a single new page — by aligning name, domain, and capability descriptions into one coherent identity across every source AI reads.

The split is easy to create without noticing. A firm acquires another and keeps both websites alive because customers know both names. A legal entity differs from the trading name that appears on the trade stand. A regional arm runs its own domain with its own capability pages. Each of those decisions is sensible on its own terms, and each hands the engines a reason to score two partial entities instead of one complete one.

What makes it worth fixing ahead of almost anything else is the arithmetic. IPEC's two identities do not average out to a middling position. They rank fourth and tenth, which in a shortlist of three means the firm is absent from answers it would otherwise appear in. Consolidation does not require better pages or new content. It requires deciding which name the firm is, and then making every source the engines read agree.

How can an industrial manufacturer improve visibility when engineers ask AI assistants about suppliers?

An industrial manufacturer improves visibility with engineers asking AI assistants about suppliers by making its own capability pages readable, specific and consistently named, in that order. Closing the gap is structured work, not a campaign. The SIGNALS framework scores a manufacturer's pages across the dimensions the engines actually reward, then rebuilds the capability content to match. The shape of the work:

What should you do if you fabricate process skids?

If you fabricate process skids, do three things, and the first one before the other two.

Your visibility is not one number — it is four, and at least one is probably zero. A firm can lead CIP on Google, Perplexity, and Claude and still be missing from ChatGPT entirely, as Sani-Matic and CSI are. The only way to know where you stand is to measure all four engines, query by query.

The gap is on your pages, and the pages are yours to change. Because the engines in this study cite fabricators' own capability pages 70 to 79% of the time, the lever is how clearly your pages document your systems, controls, validation posture, and materials — and whether the AI crawlers and the Bing index can read them at all. Validation posture is the half most capability pages leave as an adjective, and the macro report sets out what documentation a pharma skid fabricator actually hands over, stage by stage, which is the level of detail the cited pages state in readable text.

Compete where the queries are open. The specialist and application queries — cGMP, single-tank, inline dilution, sterilize-in-place — name fabricators directly. The generic "system supplier" queries belong to the global brands. Knowing the difference is where the effort should go.

AI visibility across CIP, SIP, and buffer-prep skids

How was this study built, and who produced it?

SIGNALS built this study from 27 buyer queries run across four AI engines. It is Part II of the SIGNALS pharmaceutical process skid study, produced by SIGNALS, a BlackSig Systems company. It goes deeper into the three lanes our first report identified as winnable for an independent fabricator — CIP, SIP, and buffer preparation.

If you fabricate process skids and want to see exactly where your firm appears — and doesn't — across ChatGPT, Google, Perplexity, and Claude, we can produce a firm-specific AI Visibility Snapshot mapping your share of voice against your named competitors, query by query.

How this study was built

If you fabricate process skids, here's where to start →