Google Search knowledge graph

When Google shows a knowledge panel, `knowledgeGraph` returns its title, type, description, official website, profiles and attributes, keyed by `kgmid`, for brand SERP audits and entity monitoring.

Returned in result

Request

curl -X POST "https://api.answerline.dev/v1/monitor/google" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"project management software","country":"US","include":{"aioverview":{"markdown":true}}}'

What the knowledge graph panel is

A knowledge panel is the box Google shows beside or above results when a query clearly refers to one entity: a company, person, place, product, film or organization. It summarizes who or what the entity is, where the description comes from, official links and social profiles, key facts, and entity-specific blocks such as ratings or tickets.

result.knowledgeGraph returns that panel as structured fields. The panel is often the first thing a searcher reads about a brand, so it gets audited: whether a panel exists, whether the description, official website and linked profiles are right, and whether they change after a rebrand, acquisition or leadership change. kgmid shows which entity Google picks for an ambiguous name, and the entity-specific blocks carry ratings, contact links, streaming options or admission prices for businesses, titles and venues.

Field reference

Core fields of result.knowledgeGraph

Entity-specific fields

Prices, ratings and counts in these blocks are display strings, not numbers.

Annotated excerpt

Values are illustrative; names and types follow the POST /v1/monitor/google result schema.

{
  "result": {
    // Omitted when the results page has no knowledge panel.
    "knowledgeGraph": {
      "title": "Example Robotics",
      "type": "Robotics company",               // free-form label
      "kgmid": "/g/11example00",                // stable entity key
      "description": "Example Robotics is a manufacturer of warehouse robots...",
      "source": { "name": "Wikipedia", "link": "https://en.wikipedia.org/wiki/Example_Robotics" },
      "website": "https://www.example-robotics.com/",
      "attributes": [
        { "key": "Founded", "value": "2014" },
        { "key": "Headquarters", "value": "Pittsburgh, PA" }
      ],
      "profiles": [
        { "name": "LinkedIn", "link": "https://www.linkedin.com/company/example-robotics" },
        { "name": "YouTube", "link": "https://www.youtube.com/@examplerobotics" }
      ],
      "peopleAlsoSearchFor": [
        { "name": "Sample Automation", "link": "https://www.google.com/search?q=Sample+Automation" }
      ]
      // Entity-specific blocks (webRatings, statsCard, listenOn, ...) appear only when relevant.
    }
  }
}

Cost

No include flag is needed. The panel comes with every Google Search request at the base price: 18.9 credits for a one-page async task, +25.9 for a synchronous call. The panel comes from the first results page, so keep pages at 1.

Example: brand panel audit

The script checks a list of brand queries against what each brand expects: the official website domain and required profiles. It reports panel presence, the kgmid to pin, mismatches, and a fingerprint of the description and attributes so you can detect changes between runs.

import hashlib
import json
import os
import time
from datetime import date
from urllib.parse import urlparse

import requests

API = "https://api.answerline.dev"
HEADERS = {"Authorization": f"Bearer {os.environ['API_KEY']}"}
DAY = date.today().isoformat()
EXPECTED = {
    "example robotics": {"domain": "example-robotics.com", "profiles": {"LinkedIn", "YouTube", "X"}},
    "sample automation": {"domain": "sampleautomation.example", "profiles": {"LinkedIn"}},
}


def run_tasks(tasks):
    resp = requests.post(f"{API}/v1/async/task/batch", headers=HEADERS, json=tasks, timeout=60)
    resp.raise_for_status()
    pending = {item["task"]["id"]: tasks[item["index"]] for item in resp.json()["results"] if item["success"]}
    done = []
    while pending:
        time.sleep(10)
        for task_id in list(pending):
            body = requests.get(f"{API}/v1/async/task/{task_id}", headers=HEADERS, timeout=30).json()
            if body["task"]["status"] in ("COMPLETED", "FAILED"):
                task = pending.pop(task_id)
                if body["task"]["status"] == "COMPLETED":
                    done.append((task["payload"]["query"], body["response"]["result"]))
    return done


def host(url):
    return (urlparse(url or "").hostname or "").removeprefix("www.")


tasks = [
    {"taskType": "GOOGLE", "payload": {"query": q, "country": "US"}, "idempotencyKey": f"kg-{i}-{DAY}"}
    for i, q in enumerate(EXPECTED)
]

for query, result in run_tasks(tasks):
    panel = result.get("knowledgeGraph")
    expected = EXPECTED[query]
    if not panel:
        print(f"{query}: NO PANEL")
        continue
    issues = []
    if host(panel.get("website")) != expected["domain"]:
        issues.append(f"website is {panel.get('website')!r}")
    linked = {p.get("name") for p in panel.get("profiles", [])}
    missing = expected["profiles"] - linked
    if missing:
        issues.append(f"missing profiles {sorted(missing)}")
    facts = {"description": panel.get("description"), "attributes": panel.get("attributes", [])}
    fingerprint = hashlib.sha256(json.dumps(facts, sort_keys=True).encode()).hexdigest()[:12]
    source = (panel.get("source") or {}).get("name", "none")
    print(f"{query}: kgmid={panel.get('kgmid')} type={panel.get('type')!r} source={source} facts={fingerprint}")
    for issue in issues:
        print(f"    - {issue}")

Store the fingerprint per query and day. When it changes, diff the stored description and attributes to see what Google changed, and check whether kgmid changed too: a new ID means Google now resolves the query to a different entity.

Pitfalls

knowledgeGraph is omitted, not null, when there is no panel. Use result.get("knowledgeGraph") and record “no panel” as an observation.

profiles[].name is display text, so match on a set of names you have seen, or on the host of link, rather than exact spellings.

attributes[].key values follow the result language, which follows the country. Keep country fixed per series, or change detection will flag translations as changes.

Titles can change or collide between entities with the same name. Track by kgmid.

Most entity-specific fields appear only for some entity types. Code every one of them as optional.

Related

• Brand protection and competitor analysis.
• Structured data and AI visibility: entity signals across search and AI answers.
• Tracking SERP feature changes for turning snapshots into change events.
• ChatGPT sources, the pages an AI assistant cites about an entity.
• Google Search engine guide for targeting, batches and billing.

More Google Search data

All Google Search fields · API reference