Live case study
AgentLocal, measured by AgentLocal
We sell measurement, so we measure ourselves in public. This page tracks whether AI assistants can find, describe, and recommend AgentLocal — starting from new and unseen, using exactly the machinery we run for customers. Every number here is measured; nothing is projected. If the numbers do not move, this page will say so.
Baseline started . Reviewed .
The setup
A frozen corpus of buyer-intent queries — the questions a business owner actually asks an assistant before they know any vendor exists: “how do I make my business show up in ChatGPT answers”, “how much does AI visibility monitoring cost”, “best AI visibility tools compared”. The corpus spans service, comparison, price, trust, and brand intent, plus control queries that are deliberately irrelevant to us, so ecosystem drift can be separated from movement we caused. Named competitors are tracked in the same answers, the same way we track a customer's competitors.
Answers are scored by the same deterministic rules as every other measurement: a mention is only a mention, a citation must name a source, and answering a price question requires an actual figure. The full method is public at /research/methodology, and the measurement transcripts for this page are published in the repository under docs/research/ as each window completes.
Measured so far
Who assistants name in this market
- Peec AI: mentioned in 4 of 22 answers
- Profound: mentioned in 3 of 22 answers
- AgentLocal: mentioned in 2 of 22 answers
- Otterly.AI: mentioned in 1 of 22 answers
- Yext: mentioned in 1 of 22 answers
Aggregated from AgentLocal's own instrumented query runs. Observed association only — not access to any platform's private prompts, and not a ranking guarantee. Source links are assistant-reported URLs, not private search-tool telemetry.
Interventions, as they happen
Every change we make on our own behalf is the same playbook we run for customers, dated so movement can be checked against it:
- — Structured data, llms.txt, agent-readable Markdown mirrors, and the public MCP endpoint live on agentlocal.org.
- — Measurement methodology published with transcripts for four researched verticals.
- — Verifiable live counters at /api/v1/network; MCP discovery via /.well-known/mcp.json; registry submission prepared. Self-measurement baseline started.
Directory and registry listings are recorded with dated status transitions as they are submitted and verified, and will be listed here as they go live.
What would count as failure
If, after the baseline and a full intervention cycle, assistants still cannot describe what AgentLocal does, cite agentlocal.org, or answer what it costs, then our own playbook did not work on us — and this page will show that, with the same transcripts. That is the deal we ask customers to accept, so it is the deal we accept in public.