Research / Edition 2026-09
Named in most answers, recommended first in none: the gap between mention and recommendation
Key findings
- Weaviate and Milvus were each named in 90% of vector-database answers and first in none. Pinecone was named in 94% and first in 38%.
- Twenty vendors were named in 70% or more of their category's 50 answers and were first in 2% or fewer, including AWS (84%, hosting), Kit (88%, newsletters), Arize (94%, observability), Mighty Networks (88%, communities) and v0 (90%, app builders).
- Synthesia was named in 100% of avatar answers and first in 4%. HeyGen was named in 100% and first in 92%.
- Sixteen vendors were named first in 60% or more of their category. ElevenLabs (94%), HeyGen (92%) and Gusto (86%) hold the strongest positions in the Index.
- In nine categories the top two vendors have identical answer share and the leader is decided by who is named first.
Weaviate appears in 90% of AI answers about vector databases and is named first in 0%. So is Milvus. Pinecone appears in 94% and is first in 38%. The Memetik Index tracks both numbers because they measure different things: answer share is whether a model knows you belong in the category; named-first share is whether it would pick you.
Which vendors are mentioned everywhere but recommended nowhere?
Twenty vendors were named in at least 70% of their category’s 50 answers and were the first vendor named in 2% or fewer.
| Category | Vendor | Answer share | Named first | Leader (named first) |
|---|---|---|---|---|
| LLM observability | Arize | 94% | 2% | LangSmith (16%) |
| Vector databases | Weaviate | 90% | 0% | pgvector (32%) |
| Vector databases | Milvus | 90% | 0% | pgvector (32%) |
| AI app builders | v0 | 90% | 0% | Lovable (80%) |
| Newsletter platforms | Kit | 88% | 0% | Substack (64%) |
| Paid community platforms | Mighty Networks | 88% | 2% | Circle (60%) |
| AI code review | Qodo | 86% | 2% | CodeRabbit (74%) |
| Transactional email | SendGrid | 86% | 2% | Resend (56%) |
| App hosting | AWS | 84% | 0% | Vercel (62%) |
| AI outbound | Instantly | 82% | 2% | Apollo (64%) |
| Browser automation | Browserbase | 82% | 2% | Playwright (42%) |
| AI outbound | Smartlead | 78% | 2% | Apollo (64%) |
| Transactional email | Mailgun | 78% | 0% | Resend (56%) |
| AI website builders | Hostinger | 76% | 2% | Wix (74%) |
| AI voice | Murf | 74% | 0% | ElevenLabs (94%) |
| Authentication | Supabase Auth | 74% | 0% | Clerk (58%) |
| AI avatar video | Colossyan | 72% | 0% | HeyGen (92%) |
| AI code review | Greptile | 72% | 0% | CodeRabbit (74%) |
| Inference hosting | Baseten | 72% | 0% | Hugging Face (34%) |
| AI image generation | Ideogram | 70% | 0% | Adobe Firefly (36%) |
Each of these vendors is on the shortlist in most answers and opens almost none of them. AWS is the clearest case: named in 42 of 50 hosting answers, the recommendation in zero.
Where does named-first decide the leader?
Nine categories have a tie on answer share at the top. The Index ranks them by named-first share.
| Category | Vendor | Share | First | Vendor | Share | First |
|---|---|---|---|---|---|---|
| AI avatar video | HeyGen | 100% | 92% | Synthesia | 100% | 4% |
| Automation platforms | Zapier | 100% | 70% | n8n | 100% | 14% |
| LLM APIs | OpenAI | 100% | 66% | 100% | 4% | |
| CRM for startups | HubSpot | 100% | 64% | Pipedrive | 100% | 16% |
| Link-in-bio tools | Linktree | 100% | 48% | Beacons | 100% | 26% |
| Product analytics | Amplitude | 100% | 30% | Mixpanel | 100% | 22% |
| AI coding agents | Cursor | 98% | 48% | Claude Code | 98% | 38% |
| Vector databases | pgvector | 98% | 32% | Qdrant | 98% | 24% |
| AI image generation | Adobe Firefly | 96% | 36% | Midjourney | 96% | 34% |
HeyGen and Synthesia are mentioned in every answer. HeyGen opens 46 of them; Synthesia opens 2. Google’s models are in every LLM API answer and open two.
Who is both named and picked?
Sixteen vendors were named first in 60% or more of their category’s answers. The strongest: ElevenLabs (100% share, 94% first), HeyGen (100%, 92%), Gusto (100%, 86%), Lovable (100%, 80%), CodeRabbit (100%, 74%), Wix (96%, 74%), Opus Clip (96%, 72%), Stripe Billing (100%, 72%), Zapier (100%, 70%). These are the vendors a model reaches for before it lists the alternatives.
Why does the gap exist?
Answers open with one recommendation and then list alternatives “depending on your needs”. The opening name is the model’s default. The alternatives are its hedge. A vendor can be on the hedge list in every answer without ever being the default, and a buyer skimming the answer reads the default.
What this means for a vendor
Answer share says the model has you filed in the category. Named-first share says it would pick you, and that is the number a buyer acts on. Every model knows Weaviate; none defaults to it. Coverage across the sources the models read raises answer share. A consistent, specific reason to pick you first, in the sources they cite most, raises named-first share.
Limitations
Named first is measured as the first tracked vendor to appear in the answer text, which is usually but not always the recommendation. Fifty answers per category means one answer moves first share by two points. One edition.