Your visibility problem might be a single-engine problem

In customer support software loss records, 171 of 174 competitor mentions (98.3%) come from a single engine: ChatGPT. Gemini and Perplexity barely surface a competitor in that category at all, and Claude surfaced none in the sample. In project management software, the same kind of loss record spreads across all four engines, none dominating. Same measurement, two completely different shapes, and the fix that works for one does nothing for the other.

The setup

We ran 270 live AI visibility panels across three B2B software categories, each testing a different brand against buyer-intent prompts on four search-grounded engines. In every response where the tested brand did not appear, we logged which competitor the engine named instead and which engine produced that response. Summed by category and engine, that produces a loss record: how much of the competitor-mention volume each engine is responsible for.

The question this answers is not "how much competitor mention volume exists" but "is that volume coming from one engine or from all of them." The answer changes what an audit's headline number is actually telling you.

Loss records by engine, two categories

In customer support software, 58 live panels produced 174 loss mentions, and 171 of them came from ChatGPT alone. Claude did not name a single competitor in a customer-support response where the tested brand was absent. Gemini and Perplexity combined for 3 mentions out of 174. Whatever is driving competitor visibility in this category, it is happening almost entirely inside one engine's retrieval path.

In project management software, 173 live panels produced 519 loss mentions, spread across all four engines with no engine above 34% or below 12%. A brand losing visibility in project management is losing it everywhere at once. A brand losing visibility in customer support is, in practical terms, losing it on ChatGPT.

Why the concentration happens in one category and not the other

Each engine grounds its answers in a different slice of the web, weighted by its own search index and freshness window. When a category has a small set of pages that most engines converge on, retrieval looks similar across engines and the loss record spreads out, which is the project management pattern. When one engine's retrieval path pulls from sources the other three do not weight the same way, that engine ends up responsible for most of the competitor-naming, which is the customer support pattern. HR software loss records (117 mentions) sit closer to the spread pattern too: ChatGPT 38.5%, Claude 25.6%, Perplexity 21.4%, Gemini 14.5%, no single engine over half. Customer support is the outlier, not the norm.

The practical cause is specific to the engine and category combination and is visible in the citation source list for that engine's responses, not guessable from the blended score alone.

What this means for a visibility strategy

Before spending on content or source work to fix a competitor gap, check whether the loss record for that category is concentrated in one engine or spread across all four. If it is concentrated, the fix target is that one engine's retrieval sources, and time spent on the other three will not move the number. If it is spread, as in project management, the fix has to work across multiple retrieval paths at once, because no single engine explains most of the gap.

Skipping this check means a brand can do real content work, see no movement in its blended score, and conclude the work did not help, when the real issue was that the work targeted engines that were never the source of the competitor volume in the first place.

An AI visibility audit reports loss records broken out by engine for exactly this reason: so the concentration, or lack of it, is visible before any content budget gets spent.

Methodology

270 live teaser panels across three B2B software categories (58 customer support, 173 project management, 39 HR), each 10 buyer-intent prompts run across four search-grounded AI engines (ChatGPT with web search, Gemini with grounding, Perplexity Sonar, Claude with web search). In every response where the tested brand did not appear, the competitor named in that response was logged against the engine that produced it. Loss mention counts are the sum of those competitor appearances per engine and category, not a per-brand count. Data from CitedWell's audit engine, collected through July 2026, live-engine runs only.