A response naming your brand runs 12% shorter on two engines. The other two don't move at all.
An earlier post on this blog split citation count by whether a response named the target brand, and found a real gap on Claude and Gemini but nothing on ChatGPT and Perplexity. That left an open question: does response length move the same way, on the same engines, or is the citation-count effect independent of how much the engine writes? We split the same 8,664 real responses by word count instead of citation count and got a different answer than expected: a real length gap shows up, but on a different pair of engines entirely.
The length gap and the citation gap don't share engines
We pulled the response text from every real, non-error call across our 270 live audit panels (project management, customer support, and HR software; ChatGPT with web search, Gemini with grounding, Perplexity Sonar, and Claude with web search), counted words by splitting on whitespace, and split the average by whether that response's analysis flagged the target brand as mentioned. Sample sizes match the citation-count-vs-target-mention post exactly, from 327 up to 2,095 responses per cell.
| Engine | Brand named: avg words (n) | Brand absent: avg words (n) | Difference |
|---|---|---|---|
| ChatGPT | 628.0 (327) | 713.7 (973) | -12.0% |
| Gemini | 597.4 (508) | 675.4 (1,818) | -11.5% |
| Perplexity | 249.3 (491) | 252.0 (1,867) | -1.1% |
| Claude | 463.6 (585) | 466.3 (2,095) | -0.6% |
ChatGPT and Gemini write meaningfully shorter answers when the response names the target brand, 11 to 12 percent shorter than when it does not. Claude and Perplexity barely move, under 1.1 percent either way. That is close to the exact opposite lineup from the citation-count post: there, Claude and Gemini showed a real gap (citations up 13-15% when the brand was named) and ChatGPT and Perplexity showed none. Gemini is the only engine that shows a real effect on both measures, and even there the two effects point in opposite directions: naming the brand comes with 14.7% more citations and 11.5% fewer words at the same time.
What a shorter mention-side answer might mean
On ChatGPT, this is consistent with a pattern already documented on this blog: it writes the longest average answer of any engine (692 words) and cites the fewest sources per word. A long ChatGPT answer with no brand names in it reads like a broader category survey, more explaining, more hedging, more room for competitors. A shorter one that does land on the target brand reads more like a direct, decided answer. Gemini's case is different, since Gemini's shorter mention-side answers also carry more citations, not fewer, ruling out "shorter because it read less." Whatever is shortening those Gemini answers is not a smaller source pull.
The medians tell a similar story to the averages on three of four engines. Gemini's median drops from 609 words on a no-mention response to 588.5 on a mention response, and ChatGPT's from 712 to 642. Perplexity holds close, 220 to 225, and Claude actually ticks up slightly at the median (478 to 485) even though its average moves a fraction of a point the other way, a sign that a handful of long outlier responses are pulling Claude's average around, not a real directional effect.
Consistent with the known baselines
The blended, unsplit average word count per engine here (Claude 465.7, Gemini 658.3, ChatGPT 692.1, Perplexity 251.4, all across 8,664 responses) matches the response-length-by-engine post's published figures (466, 658, 692, 251) exactly, and the per-engine, per-side response counts (585/2,095, 508/1,818, 327/973, 491/1,867) match the citation-count-vs-target-mention post's counts exactly, confirming this is the same dataset and the same split, read a second way.
Find out whether the AI engines your buyers use write you a short, decided answer or leave you out of a long, hedged one.
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-scored every real (non-error) response via engine/scorer.ts this session, both branded and organic prompts, 8,664 responses total, and split word count (response text split on whitespace, no other normalization) by whether that response's analysis flagged the target brand as mentioned. No development-rail or fixture data is included; all responses came from live engine calls. Data collected June-August 2026.