A branded prompt pulls 15 percent more citations on two engines. That is not why it is slower everywhere.
An earlier post found that a branded, head-to-head prompt ("X vs Y") runs slower than an organic one on three of four engines, and traced the likely cause to the engine needing to reconcile two named entities instead of pulling one ranked list. That left an open question: does the same branded-comparison mechanic also pull in more citations, more sources needed to compare two named things instead of list several? We joined the same prompt-type field to citation count across the same dataset to check.
Claude and Gemini cite noticeably more on a branded prompt. ChatGPT and Perplexity barely move.
Across 270 live audit panels and 8,664 real non-error responses, we split every response by whether its prompt was branded (names a specific brand, including head-to-head comparisons) or organic (category-level, no brand named) and averaged citation count per engine.
| Engine | Branded avg citations (n) | Organic avg citations (n) | Difference |
|---|---|---|---|
| Gemini | 15.31 (547) | 13.29 (1,779) | +15.2% |
| Claude | 6.68 (631) | 5.82 (2,049) | +14.8% |
| Perplexity | 8.85 (543) | 8.79 (1,815) | +0.7% |
| ChatGPT | 6.40 (327) | 6.50 (973) | -1.6% |
Gemini and Claude both cite meaningfully more when the prompt names a brand, 15.2 and 14.8 percent respectively. Perplexity and ChatGPT are flat, within a couple of percentage points either way, well inside noise. That two-and-two split matches the earlier latency post's engine grouping almost exactly, but not quite, which is the useful part.
The overlap with the latency finding is partial, not complete
The latency post found branded prompts running slower on Claude (+10.5% median), Gemini (+15.8%), and Perplexity (+6.7%), with ChatGPT the lone exception at -4.7% (branded slightly faster). If more citations were the whole explanation for that slowdown, all three slower engines should show a citation bump. Only two do.
- Claude and Gemini: slower on branded prompts and citing meaningfully more. The extra sourcing work is a plausible explanation for at least part of the slowdown on these two.
- Perplexity: slower on branded prompts (+6.7%) but citation count does not move (+0.7%, noise-level). Whatever costs Perplexity the extra time on a branded prompt, it is not pulling in more sources. That fits a pattern established previously: Perplexity's citation count sits inside a fixed 5-to-10 window regardless of response speed, so a branded prompt cannot push it higher even when the model works harder.
- ChatGPT: the one engine that answers branded prompts faster, and its citation count is flat too (-1.6%). Consistent, since there is no extra sourcing cost to explain a slowdown that never happens.
The direction holds across categories on Claude and Gemini, not on the other two
Splitting the same branded/organic comparison by category (PM, HR, CS software) checks whether one vertical is driving the engine-level numbers above.
| Engine | PM | HR | CS |
|---|---|---|---|
| Claude | +13.9% | +23.9% | +12.5% |
| Gemini | +12.5% | +25.8% | +24.0% |
| Perplexity | -0.8% | +4.0% | +5.5% |
| ChatGPT | +4.6% | -9.6% | -2.7% |
Claude and Gemini post a positive branded-vs-organic citation gap in all three categories, all six cells above 12 percent, confirming the engine-level averages are not one category's artifact. Perplexity and ChatGPT stay inside a roughly plus-or-minus 10 percent band in every category, no consistent direction, consistent with the flat blended averages rather than a category effect canceling itself out.
Find out which of the four engines your buyers actually use, and whether a branded or organic question changes how much sourcing work it does on your category.
Get an AI Visibility Audit, $490Methodology
Data drawn from 270 live, search-grounded audit panels (project management, customer support, and HR software brands), each run across four AI engines: ChatGPT with web search, Gemini with grounding, Perplexity Sonar, and Claude with web search. We re-read every real (non-error) response from results.jsonl fresh this session, joined each response's promptId to its panel.json prompt record to read the promptType field (branded or organic), and used the raw length of the response's citations array as the citation count. Two panels (teaser-gladly-live, teaser-kustomer-live; 43 of 8,664 responses) predate the promptType field; for those we applied the same legacy fallback the scoring engine itself uses, classifying a prompt as branded if its text contains the panel's brand name. Total response counts by engine (Gemini 2,326, Perplexity 2,358, Claude 2,680, ChatGPT 1,300) match every prior post's baseline exactly. No development-rail or fixture data is included; all responses came from live engine calls. Data collected June-September 2026.