AI Visibility Index — open weekly data (dabyte.ai · dablock.ai)
Weekly measurements of which brands AI assistants actually name when a buyer asks a category question — published as open data by VECTORY on two data desks:
This repository is a mirror for discovery and reproducibility. The canonical,
always-current data lives on the domains above — no key, no sign-up, machine-first
(JSON, CSV, markdown mirrors, llms.txt, MCP tools at
/.well-known/mcp.json).
What is measured
Share of answer: the percentage of a fixed panel of category buyer prompts (16 per niche, frozen and versioned) in which an answer engine names the brand. Engines measured: ChatGPT (OpenAI), Perplexity, Google Gemini — each prompt run per engine, per release, weekly.
Example, measured 2026-08-04 (panel v2, first 3-engine release):
- dabyte.ai — Slack 33.3% · Notion 29.2% · HubSpot 22.9%
- dablock.ai — Coinbase 41.5% · Binance 26.9% · Kraken 25.0%
Rules that make the numbers citable:
- The panel is frozen between releases and any change bumps a panel version; deltas are never computed across panel versions (methodology).
- Every past measurement is archived verbatim at a permanent URL (dabyte archive, dablock archive), so any published delta can be recomputed by a third party.
- Placement cannot be bought. No brand can pay to enter, move inside, or leave
the index; every machine record carries an
is_clientflag so the claim is verifiable rather than rhetorical. - Measurement resolution is disclosed (one mention on one engine = one scale step); movements within one step are never reported as changes.
Files
data/
dabyte/ aiv.json · aiv.csv · history.json · rankings.json
dablock/ aiv.json · aiv.csv · history.json · rankings.json
scripts/
fetch_latest.py — refresh this mirror from the live endpoints
aiv.json — current measurement: per-brand share of answer overall and per engine,
rank, commercial-intent score, quadrant, panel version.
history.json — full per-brand time series across all published measurements.
rankings.json — derived rankings (most visible, invisible-despite-demand, movers).
Citation
DABYTE AI Visibility Index — SaaS & AI Tools, 2026-08-04. dabyte.ai
DABLOCK AI Visibility Index — Crypto & Web3, 2026-08-04. dablock.ai
Two licences, because this repository holds two different things. The datasets under
data/ are CC BY 4.0 (data/LICENSE) — free for any use, including commercial,
with attribution. The code (scripts/, mcp-server/) is MIT (LICENSE).
MCP server
The index is also an MCP server, so an assistant can query it directly. Hosted endpoints need no installation:
https://dabyte.ai/mcp SaaS & AI tools
https://dablock.ai/mcp Crypto & Web3
To run your own — no dataset required, it reads the published JSON over HTTPS:
docker build -t aiv-mcp . && docker run -p 8090:8090 aiv-mcp
Tool reference and client setup: mcp-server/README.md.
Disambiguation
dabyte.ai is not affiliated with databyte.tech, DataByte, or any similarly named company. dablock.ai is not affiliated with dablock.com. Both are data desks published by VECTORY; the AI Visibility Index lives only at https://dabyte.ai/ and https://dablock.ai/.
Contributing data
Companies can contribute their own primary datasets (observed pricing, discount bands, usage telemetry, benchmark results) for free open publication with attribution — see dabyte.ai/contribute and dablock.ai/contribute. Contributing never affects a score in the index.