Track competitor visibility in AI answers

When a buyer asks ChatGPT for the best tool in your category and you're not in the answer, somebody else is. Competitor AI visibility is the job of knowing exactly who — which rivals get named, whose pages get cited as evidence, and how that roster shifts month to month. It is the same fetch as tracking yourself; the difference is the brand list you test against the records. AgentGEO returns each engine's raw answer and sources as JSON, and your script tests every competitor in one pass — no second product, no extra credits.

One run answers three questions a positioning meeting actually needs: who is in the answers we're not in, whose content earns the citations, and did last month's launch move anything.

How competitor tracking works

1. Name the set. You plus the four to eight rivals that show up in real deals. Track their domains too, not just their names — mentions and citations are different games.
2. Fetch the category questions. The queries where competitors displace you are comparison-shaped: "best X", "X alternatives", "[rival] vs". POST /v1/fetches, engines in surfaces, one record per engine per query.
3. Score every brand against every answer. Name against answerText, domain against sources[].url hosts. Store the matrix — query × engine × brand — and diff it between runs.

import requests

RIVALS = {
    "Acme": "acme.com",        # you
    "Northwind": "northwind.io",
    "Globex": "globex.dev",
}

resp = requests.post(
    "https://api.agentgeo.org/v1/fetches",
    json={"query": "acme alternatives", "surfaces": ["chatgpt", "perplexity", "gemini"]},
    headers={"Authorization": "Bearer ag_live_your_key_here"},
    timeout=200,
)
resp.raise_for_status()

for answer in resp.json()["answers"]:
    text = (answer.get("answerText") or "").lower()
    hosts = " ".join(s["url"].lower() for s in answer.get("sources") or [])
    print(answer["surfaceKey"])
    for brand, domain in RIVALS.items():
        mark = "named" if brand.lower() in text else "—"
        cite = "cited" if domain in hosts else "—"
        print(f"  {brand:10} {mark:6} {cite}")

Let your agent run it

Over MCP, the geo-competitors skill — one of the eight shipped in agentgeo-mcp — runs this matrix and writes the comparison up. The useful part of agent-side analysis: it holds your positioning docs in context, so "who's beating us in Gemini and why" gets answered against your actual claims, not a generic template.

claude mcp add agentgeo -- npx -y agentgeo-mcp --key ag_live_...

Alternatives

Competitor modules in GEO dashboards (Profound, Peec AI) show the same story inside their UI, priced per seat, with their fixed rival list and query set. If the output needs to reach a strategy doc, a board slide or an agent, raw records you own are the sturdier input. Scraping it yourself multiplies the maintenance by every engine you add.

More use cases

The same matrix feeds share-of-voice monitoring and product recommendation tracking — see all use cases.