Browse sections
Brands covered
Each cell is how many of the 10 questions mentioned that brand for that model. Brands are alphabetical, not ranked.
| Brand | Qwen | Kimi | DeepSeek | GLM | Total /40 |
|---|---|---|---|---|---|
| Ahrefs | 5/10 | 8/10 | 7/10 | 1/10 | 21/40 |
| Ahrefs Brand Radar | 2/10 | 2/10 | 7/10 | 0/10 | 11/40 |
| AthenaHQ | 2/10 | 8/10 | 6/10 | 0/10 | 16/40 |
| BrandRank | 0/10 | 1/10 | 1/10 | 0/10 | 2/40 |
| HubSpot | 1/10 | 8/10 | 1/10 | 1/10 | 11/40 |
| LLMention | 0/10 | 0/10 | 0/10 | 0/10 | 0/40 |
| Otterly AI | 0/10 | 0/10 | 0/10 | 0/10 | 0/40 |
| Peec AI | 7/10 | 10/10 | 7/10 | 1/10 | 25/40 |
| Profound | 7/10 | 10/10 | 7/10 | 1/10 | 25/40 |
| Rankscale | 0/10 | 4/10 | 5/10 | 0/10 | 9/40 |
| Scrunch AI | 2/10 | 8/10 | 7/10 | 0/10 | 17/40 |
| Semrush | 6/10 | 8/10 | 7/10 | 1/10 | 22/40 |
What stood out in this run
- Peec AI was named in 25 of 40 answers across the four models.
- Profound was named in 25 of 40 answers across the four models.
- Semrush was named in 22 of 40 answers across the four models.
- Ahrefs Brand Radar, AthenaHQ, BrandRank appeared in the Chinese-hosted models but not in GLM’s answers; a single-model check would have missed them.
- 2 of 12 tracked brands were named in none of the 40 answers.
Websites the answers cited
This run recorded no citation URLs in the answer text. The paid report asks for the sources explicitly and records the URLs the answers name; when a model returns none, the report says so rather than reporting an empty list as "no citations".
What these numbers can show, and what they cannot
- Can show: which brands these models named, on this date, for these questions; how much the models disagree with each other.
- Cannot show: whether a brand is good or worth buying; that you would get the same answer yourself; how the answers will change next month.
Answers vary by person, time, region, language, and model version. Ten questions is a small sample, and models can be wrong. That is why this page states its date and conditions and why a paid report re-asks the same questions over time instead of trusting one snapshot.
How we asked
| Model | Model id | Web search | Question language |
|---|---|---|---|
| Qwen | qwen3.8-max | on | English |
| Kimi | kimi-k3 | on | English |
| DeepSeek | deepseek-v4.1-flash | on | English |
| GLM | glm-5.3 | off (model does not support) | English |
Update log
This run: 2026-11-13. The next run is planned for December 2026. Earlier runs stay on record with their dates so changes can be compared; the same questions are re-used.
Corrections
This page records what the listed models answered on 2026-11-13 under the conditions above. It is not a review, recommendation or ranking of any brand. If you believe something here is wrong, or you want a page about your brand removed, write to services@llmention-geo.com and we will check and reply.
Want the full picture for your brand?
A paid report asks 30 questions, three runs each, against one or four engines, keeps every answer verbatim, and counts mentions as “n of N” rather than percentages. See pricing or run the free readiness scan first.