← All AI Visibility Index leaderboards
Free original research · AI Visibility Index · DevOps & dev tools
Which vector databases do AI engines actually recommend?
As of July 1, 2026, Weaviate leads with a 19.1% share-of-model — it appears in 19.1% of all vector databases recommendations across 5 major AI engines, measured over 500 AI answers (10+ runs per engine, 95% CI).
See your product's share-of-model — free How we measured thisThe leaderboard — Vector databases by share-of-model
Which vector databases do AI engines recommend when developers building RAG and AI features ask what to use to store and search embeddings? We ran 10 buyer prompts × 10 runs across 5 engines (Perplexity, Google Gemini, ChatGPT (OpenAI), Claude (Anthropic), Grok (xAI)) — 500 total AI answers. Last updated July 1, 2026.
Weaviate leads our AI Visibility Index for Vector databases at 19% share-of-model across 5 AI engines.
Share of model = % of all recommendations (across every prompt and engine) that named the product. Per-engine columns show how often each engine recommends the product. CI = Wilson 95% confidence interval on share-of-model.
Download this dataset — free & open
The full leaderboard as data, yours to use. Licensed CC BY 4.0 — free to reuse with attribution.
Cite this: Clear Cited AI Visibility Index — Vector databases. Retrieved July 1, 2026. CC BY 4.0. https://clearcited.com/ai-visibility-index/vector-databases/
What the leaders have in common
Patterns measured from this edition's data — not opinion. Every figure below is reproducible from the dataset at the foot of this page.
- The top three — Weaviate, Qdrant, Pinecone — together hold 57% of the category's share-of-model. AI answers here cluster around a short list.
- Weaviate doesn't win on one engine — it appears in 5 of the 5 engines measured. Breadth across engines, not a single-engine spike, is what carries a category lead.
- No single tool tops every engine: Pinecone, Weaviate each lead at least one. Which product a buyer hears depends on which assistant they ask — the reason we never collapse the engines into one number.
- The gap between #1 (Weaviate) and #2 (Qdrant) is 0.1% of share-of-model — close enough that a strong cycle could flip it.
Get the monthly movers for Vector databases
One email a month: who moved up or down in AI answers for vector databases, measured across all five engines. Double opt-in, one-click unsubscribe.
Google AI surfaces — measured separately
Google AI Overviews and AI Mode are a distinct search surface, not one of the five assistants ranked above. We measure how often a buyer prompt triggers an AI answer here and which vendors it cites — on their own track, with their own sample size, 95% CI and date, and never summed into the five-engine share-of-model. Why we keep them separate →
No measured Google AI-surface edition for this category yet — it appears here once measured (owner-armed via DataForSEO, spend-capped). We never show a fabricated number.
Methodology — reproducible, not vibes
Buyer "money prompts"
10 real buyer questions a person would ask an AI when choosing vector databases (e.g. "best vector databases for a Series A startup"). Each prompt is run 10+ times per engine.
5 engines, measured separately
We query Perplexity, Google Gemini, ChatGPT (OpenAI), Claude (Anthropic), Grok (xAI) independently, because the engines disagree — being strong in one says nothing about the others. Per-engine columns expose that spread.
Share-of-model + Wilson CI
For each product we report its share of all recommendations in the field, with a 95% Wilson confidence interval — the honest way to summarize a small, noisy sample.
Heuristic detection, disclosed
A product is "recommended" when its name or a known alias is named in the answer (boundary-aware); a citation is counted when its own domain appears in the answer's sources. A mention is not always a positive endorsement — we say so.
Pre-registered + no conflict of interest
The prompt set, brand universe and metrics are fixed before each run (no post-hoc cherry-picking), and Clear Cited excludes its own clients from this public ranked Index. Full methodology & policy →
Conflict-of-interest policy: Clear Cited excludes its own clients from this public ranked Index to avoid a conflict of interest. Excluded brands may still be measured, but are not ranked here; the policy is disclosed rather than the exclusion hidden.
API answers approximate, but do not exactly replicate, the consumer apps (different system prompts, tools, browsing defaults). We treat each edition as a point-in-time measurement and re-run on a cadence.
FAQ
Which vector databases do AI engines recommend most?
As of July 1, 2026, Weaviate leads the Clear Cited AI Visibility Index for Vector databases with a 19.1% share-of-model — meaning 19.1% of all product recommendations across 5 AI engines named it. See the full ranked table above.
What is share-of-model?
Share-of-model = the percentage of all product recommendations, across every buyer prompt and engine, that name a given product. It answers: when an AI recommends something in this category, how often is it this product?
How is the AI Visibility Index measured?
Each buyer prompt is run at least 10 times per engine across Perplexity, Google Gemini, ChatGPT (OpenAI), Claude (Anthropic), Grok (xAI). We detect which products each answer recommends, compute each product's share-of-model, and report a Wilson 95% confidence interval. The method is reproducible — the same one Clear Cited uses in its paid AI-visibility audits.
Why do different AI engines recommend different products?
AI answer engines draw on different training data, retrieval sources, and ranking — so the 'best tool' a buyer hears depends heavily on which assistant they ask. That is why the Index measures and reports each engine separately.
Is this ranking sponsored or pay-to-play?
No. The AI Visibility Index is free, independent original research. Products are not charged to appear and cannot pay to rank higher. It reflects what AI engines actually say, measured transparently.
What's the best vector database for a Series A startup building RAG in 2026?
When buyers ask AI a question like this, the vector databases our Index sees recommended most often — by share-of-model, across 5 engines — are Weaviate (19.1%), Qdrant (19.0%), and Pinecone (18.8%), as of July 1, 2026. That reflects the aggregate of our buyer-prompt measurement for the category; individual prompts and engines vary, so check the per-engine columns in the leaderboard above.
Best vector database for a small team adding semantic search?
When buyers ask AI a question like this, the vector databases our Index sees recommended most often — by share-of-model, across 5 engines — are Weaviate (19.1%), Qdrant (19.0%), and Pinecone (18.8%), as of July 1, 2026. That reflects the aggregate of our buyer-prompt measurement for the category; individual prompts and engines vary, so check the per-engine columns in the leaderboard above.
Want your product's real share-of-model?
We measure where ChatGPT, Perplexity, Gemini, Claude & Grok send buyers in your category — reproducibly — and map the fastest way in.
Get a free AI-visibility teardown