Knowledge panel

A knowledge panel is the information box Google shows beside or above search results for an entity it already holds in its Knowledge Graph. Google's own definition: "Knowledge panels are information boxes that appear on Google when you search for entities (people, places, organizations, things) that are in the Knowledge Graph."

The key word is automatically. Google states that panels "are automatically generated, and are updated automatically as information changes on the web." A panel is not a profile you fill in. It is a read-out of what Google's systems have concluded about an entity from many sources at once.

Where does a knowledge panel get its facts?

Google says the information "comes from various sources across the web", naming three kinds: data partners, open web sources, and verified entities who have suggested edits to facts on their own panels.

Two of the three are third parties. Only the last is you, and it arrives late: after a panel exists, and as a suggestion rather than an edit. The panel is downstream of the Knowledge Graph, and the graph is downstream of a corroborated web. In May 2020 Google wrote that the graph "has amassed over 500 billion facts about five billion entities", the scale at which one self-description counts for little against everything else that mentions you.

Can you create or claim a knowledge panel?

These are different operations, and confusing them wastes months.

Creating is not available. There is no form, no submission and no paid route. An entity gets a panel because enough independent sources describe it consistently enough for Google to resolve it as a distinct thing.

Claiming is available once a panel exists. Google's support documentation: "If you are the subject of or official representative of an entity depicted in a knowledge panel, you can claim this panel and suggest changes." Claiming gives you a verified channel for corrections, not editorial control. A suggested change is still weighed against the rest of the web.

So the work that produces a panel is not panel work. It is entity work: one canonical name, one canonical home page, structured data that says what the entity is and which profiles are the same thing, and consistent descriptions wherever organisations are described.

What is the AI-answer version of a knowledge panel?

When someone asks an assistant "what is [company]", the answer is a panel written in prose: a summary assembled from sources, with the sources linked underneath. You can see exactly which sources build yours. Here is ours. Attensira asks "what is Attensira" on ChatGPT and Perplexity every day for our own site. ChatGPT's 33 answers between 4 September and 4 October 2026 carried 790 citations, split between our own pages and directories.

attensira.com was cited 564 times, g2.com 60, linkedin.com 58, capterra.com 46 and docs.attensira.com 23. The chart at the top of this page shows the same list.

The pattern matches Google's description of panels almost exactly. Our own site leads, but a review directory, a social profile and a second directory together supplied more citations than our documentation did, and none of those pages is one we write in full. If the G2 description of us were wrong, the answers would repeat it. Answers about us also borrow sources from similarly named organisations; the knowledge graph page shows them.

Why does a knowledge panel matter if assistants do not read it?

A panel is a Google Search feature; an assistant answering a question retrieves documents, not your panel. Treating the panel as the goal is a category error.

Treating it as an indicator is right. A panel is public evidence that a large entity-resolution system mapped your name to one specific thing with attributes it trusts enough to show. Assistants face the same disambiguation problem, and the same signals that resolve the entity for Google, consistent naming, corroboration and machine-readable identity, are what let an assistant retrieve your documents rather than someone else's.

What goes wrong with knowledge panels?

• Chasing the box. You cannot write the output; the input is corroborated entity data.
• Believing a vendor who sells one. Google documents no paid or submitted path.
• Assuming a claim is an edit. Verification opens a suggestion channel; facts still need support elsewhere.
• Inconsistent self-description. Different names, founding dates or categories across your site and profiles give a reconciler no consensus.
• Reading a wrong fact as a Google bug. It is usually a correctly ingested wrong fact from a source. Fix the source.

How does Attensira help with what panels and answers say about you?

It shows you the AI answer's source list, the part of the panel you can act on. For every tracked question, on every engine, Attensira ranks the pages the answers leaned on by how much weight each carried, and marks whether you are on each one: your own domain, or a listing whose placement has been verified. For our own brand that list is the chart above, and three of its top five are directories and profiles we do not write. When a directory decides your answer and your listing is missing or thin, the crew's off-site work prefills it from your confirmed brand profile and waits for your approval before anything is filed. Attensira does not touch your Google panel; claiming that is yours to do with Google.