Google knowledge panel data: which searches get one and what fields come back

When you search for a person, a place or a well-known thing, Google often puts a box on the side of the results with a name, a one-line label and a short description: the knowledge panel. If you do entity SEO, brand monitoring or data enrichment, it is a useful structured fact. But which searches get one, and how complete is it?

We ran 40 entity searches through the Google Search API in October 2026 and measured it. This guide has the numbers, code to check your own entities, and a clear note on the gaps. Results change, so yours will differ.

How we ran it

We used 40 searches in five groups, each in us:

• People (8): taylor swift, elon musk, lionel messi, barack obama, serena williams, albert einstein, beyonce, tim cook.
• Companies (8): apple, nike, openai, tesla, spotify, starbucks, microsoft, adidas.
• Places (8): eiffel tower, tokyo, grand canyon, mount everest, central park, colosseum, santorini, golden gate bridge.
• Concepts (8): photosynthesis, inflation, blockchain, black hole, machine learning, democracy, supply and demand, gravity.
• Media and products (8): the godfather, minecraft, iphone 16, python programming language, harry potter, stranger things, playstation 5, wikipedia.

It cost 40 credits. These are searches we picked, checked once, so the figures describe this set and not Google as a whole.

What the field looks like

knowledge_panel is null when there is no panel, and otherwise an object with three keys:

{
  "title": "Lionel Messi",
  "type": "Argentine soccer player",
  "description": "Lionel Andrés \"Leo\" Messi Cuccittini is an Argentine professional footballer ..."
}

Across the 25 panels we saw, title was present every time, type in 20 and description in 15. Descriptions ran from 142 to 417 characters, with a median of 221.

Which searches got a panel

25 of the 40 searches did, 62%, but it depended entirely on what was searched for:

People and places always got a panel and rarely an overview. Concepts were the opposite: only gravity had a panel, and seven of the eight concept searches had an AI Overview. Across all 40, 7 searches had both, 18 had a panel only, 14 an overview only and 1 had neither.

So for entity searches, the two features split the work: Google answers a “who” or “where” with a panel and a “what is” with an overview. If you track one of them, you are missing the other for half of your searches, which is why AI Overview tracking and panel tracking are worth running together.

How complete the panels were

Completeness differed by group:

Every people panel was complete. Places had a label for all 8 but a description for only 4: tokyo came back as Capital of Japan with no description, and santorini as Island in the Aegean Sea. And the four company panels, apple, nike, starbucks and adidas, returned only a title. If you want company facts, this field will not give them to you.

The type field is display text

type is the label Google shows, and it is not a clean category. A few examples from our run:

Lionel Messi          Argentine soccer player
Tim Cook              Chairman of the board of directors of Apple
Minecraft             Video game
The Godfather         R 1972
Stranger Things       2016
Central Park          4.71,755 Google reviews Park in Winter Park, Florida

The Godfather returned R 1972, a film rating and a year, and Stranger Things just 2016. And Central Park returned a string that appears to mix a rating (4.7), a review count (1,755 Google reviews) and a category, for a Central Park in Winter Park, Florida, not the one in New York. So before you rely on a panel, check that the title and the type describe the entity you meant. Two of the 25 titles also did not match the query text (Grand Canyon National Park for grand canyon, and Beyoncé for beyonce), which is fine, but shows that names are normalised.

The code

Put your key in an environment variable:

export CRAWLERDATA_API_KEY="YOUR_API_KEY"

import os
import time

import requests

API_URL = "https://api.crawlerdata.com/v1/jobs"
HEADERS = {"Authorization": f"Bearer {os.environ['CRAWLERDATA_API_KEY']}"}


def google_search(query, country="us", retries=4):
    body = {"endpoint": "google-search", "params": {"query": query, "country": country}, "queue": "live"}
    for attempt in range(retries):
        resp = requests.post(API_URL, json=body, headers=HEADERS, timeout=180)
        if resp.status_code == 429:
            time.sleep(int(resp.headers.get("retry-after", 10)))
            continue
        resp.raise_for_status()
        job = resp.json()["job"]
        if job["status"] == "completed":
            return resp.json()["result"]
        if "retry" in (job["error"] or "").lower():  # "service is busy" and "temporarily blocked" both say retry
            time.sleep(5 * (attempt + 1))
            continue
        raise RuntimeError(job["error"])
    raise RuntimeError("Search failed after retries")


def panel_report(query):
    """What entity data a search returns: the knowledge panel and whether an AI Overview showed instead."""
    result = google_search(query)
    panel = result["knowledge_panel"]
    return {
        "query": query,
        "panel": panel is not None,
        "complete": bool(panel and panel.get("type") and panel.get("description")),
        "fields": [k for k, v in (panel or {}).items() if v],
        "ai_overview": result["ai_overview"] is not None,
    }

For four searches it printed:

{'query': 'lionel messi', 'panel': True, 'complete': True, 'fields': ['title', 'type', 'description'], 'ai_overview': False}
{'query': 'nike', 'panel': True, 'complete': False, 'fields': ['title'], 'ai_overview': False}
{'query': 'inflation', 'panel': False, 'complete': False, 'fields': [], 'ai_overview': False}
{'query': 'tokyo', 'panel': True, 'complete': False, 'fields': ['title', 'type'], 'ai_overview': False}

Note inflation. In the 40-search study it had an AI Overview, and in this repeat run it did not. That is the same run-to-run movement we measured in which sites AI Overviews cite, so check an entity more than once before you conclude anything about it.

What to do with it

• Check your entities. Run your brand, your founders and your products through panel_report. A missing panel for a company is the normal case, not a problem you caused.
• Do not expect company facts. Panels for companies returned a title only.
• Validate before you use it. Compare title with the entity you meant. The Central Park example shows how a name can resolve to something else.
• Treat type as a label. It can hold a rating, a year or a review count.
• Track the overview as well. For concepts, the AI Overview is where the answer is.

What it costs

The knowledge panel comes with every google-search response, so there is no extra credit for it. Every request is 1 credit, on every queue. Starter is $1.00 per 1,000 requests and the largest pack is $0.30 per 1,000, see pricing.

Failed requests are not charged.

Next steps

You now have numbers on which searches get a knowledge panel, how complete the data is by kind of entity, and a function that reports it for your own searches. The Google Search API page lists every parameter with a full example response, and What is a SERP API? shows the rest of the response.