Veliu Research Lab
Four AI surfaces.
No single winner for running shoes.
Luce asked ChatGPT, ChatGPT API, Gemini and Perplexity the same 30 unbranded shopper questions about running shoes, and they do not agree on who to recommend: ChatGPT leads with ASICS, ChatGPT API ties ASICS and HOKA, Gemini ties Brooks and Saucony, Perplexity leads with ASICS. This is the United States board. Luce writes its prompts in English here, and every page on this site stays in English whichever market it reports on.
API calls · en-US180
Answers captured · en-US78
AI surfaces43 APIs + web app
Markets measured3
Brands15
estimand: probe-design mention rate over Luce grammar luce-live-v0, rank-range leagues only; not demand-weighted · 30 prompts per cell
governs every board on this pageAll cells shown. Gold marks the top rate on that engine, and every brand tied with it. Grey marks a brand measured too imprecisely to rank this wave. The bar under each rate is its 95% interval.An outline and a tag mark each cell where ChatGPT web app vs ChatGPT API diverge beyond the sum of their interval half-widths.Dashed cells carry a low precision tag: measured, with intervals too wide to rank this wave. Every cell stays readable.Luce writes the questions (synthetic); the engines' answers are real (elicited).
Running shoes
VELIU·LUCEPILOT · W30
| Brand | ChatGPT web appmention rate · 95% CI | ChatGPT APImention rate · 95% CI | Geminimention rate · 95% CI | Perplexitymention rate · 95% CI | ChatGPT API trendvs last wave |
|---|---|---|---|---|---|
| Gap: no cell diverges beyond its intervals this wave | |||||
| ASICS | 83%70-97 | 83%67-97 | 83%70-97 | 73%57-87 | |
| HOKA | 80%67-93 | 83%70-97 | 87%73-97 | 70%53-87 | |
| Brooks | 77%60-90 | 67%47-83low precision | 90%77-100 | 70%53-83 | |
| Nike | 73%57-87 | 53%37-70low precision | 73%57-90 | 63%47-80low precision | |
| Saucony | 53%37-70low precision | 63%47-80low precision | 90%80-100 | 53%33-73low precision | |
| New Balance | 43%27-60low precision | 57%40-73low precision | 87%73-97 | 60%43-77low precision | |
| Adidas | 23%10-40low precision | 23%10-40low precision | 67%50-83low precision | 47%30-63low precision | |
| On Running | 7%0-17low precision | 7%0-17low precision | 23%10-40low precision | 17%3-30low precision | |
Behind these numbers
Wireless earbuds
VELIU·LUCEPILOT · W30
| Brand | ChatGPT web appmention rate · 95% CI | ChatGPT APImention rate · 95% CI | Geminimention rate · 95% CI | Perplexitymention rate · 95% CI | ChatGPT API trendvs last wave |
|---|---|---|---|---|---|
| Soundcore | 77%60-90 | 67%50-83low precision | 77%63-90 | 53%37-70low precision | |
| Sony | 60%43-77low precision | 70%53-87 | 80%67-93 | 67%50-83low precision | |
| AirPods | 50%33-67low precision | 57%40-77low precision | 57%40-73low precision | 37%20-53low precision | |
| Bose | 37%20-57low precision | 50%33-70low precision | 53%33-70low precision | 27%13-43low precision | |
| Sennheiser | 33%17-50low precision | 10%0-23low precision | 20%7-37low precision | 13%3-27low precision | |
| Galaxy Buds | 13%3-27low precision | 20%7-37low precision | 27%13-43low precision | 17%3-30low precision | |
| Jabra | 3%0-10low precisiondiverges | 27%13-43low precisiondiverges | 23%10-40low precision | 10%0-23low precision |
Behind these numbers
Visibility matrix · Running shoes on ChatGPT
Crawl presence on one axis. AI recommendation on the other. The crawl side never changes a Luce number.inputs: Common Crawl web graph, release cc-main-2026-jun-jul-aug, ingested 2026-09-04, top 10,000,000 of 119,722,885 domains by harmonic centrality · outputs: Luce wave luce-live-w30, grammar luce-live-v0
mention rate clears zerorate reaches zero
in the crawlabsent from the crawl
in crawl · recommended7
- ASICS83%95% CI 70-97HC rank 8,960
- HOKA80%95% CI 67-93HC rank 19,378
- Brooks77%95% CI 60-90HC rank 20,813
- Nike73%95% CI 57-87HC rank 262
- Saucony53%95% CI 37-70low precisionHC rank 57,502
- New Balance43%95% CI 27-60low precisionHC rank 423,948
- Adidas23%95% CI 10-40low precisionHC rank 1,415
in crawl · unrecommended1
- On Running7%95% CI 0-17low precisionHC rank 18,663
off crawl · recommendednot assessed
not assessed: a top 10,000,000 slice of the release cannot show a brand as absent from it
off crawl · unrecommendednot assessed
not assessed: a top 10,000,000 slice of the release cannot show a brand as absent from it
2 board categories publish the tables above and 3 more are measuring. 3 markets measured this wave. 15 markets and 10 sectors are queued, and nothing publishes until it is measured. How the board is built.