What is a knowledge panel?
A knowledge panel is the information box Google shows for recognized entities — companies, people, products, places. It is not scraped from your homepage; it is assembled from the Knowledge Graph, Google’s structured database of entities and their relationships. That distinction matters more every year, because the Knowledge Graph is also a grounding source for Google’s AI surfaces: AI Overviews (2.5 billion-plus monthly users, Alphabet Q2 2026) and the Gemini app (750 million-plus monthly users). When Gemini 3 answers ‘who is X’ or ‘is Y a legitimate company’, the graph is one of the first places it checks.
How does the Knowledge Graph feed AI answers?
AI answers about entities run the standard pipeline — retrieval, grounding, generation, citation — but for entity questions the grounding stage leans heavily on the graph. If the graph says your company was founded in 2015 and your site says 2016, that inconsistency either produces a hedged answer or gets your version dropped. If the graph confuses you with a namesake company, the AI answer inherits the confusion. Entity hygiene at the graph level is therefore a GEO task, not just a branding task. For the broader picture, see entities and the knowledge graph.
How do you earn a knowledge panel?
Google creates panels when it can confidently reconcile an entity from multiple independent sources. The levers, in order of impact:
- Organization schema with sameAs. Declare
@id, name,alternateName, logo andsameAslinks to your official profiles. This is the entity anchor — see Organization schema. - A Wikidata entry. Wikidata is one of the graph’s most trusted sources. Follow the Wikidata creation guide.
- Consistent third-party mentions. Press coverage, directories, partner pages and review profiles that all state the same name, founding date and description.
- A Wikipedia article where genuinely notable. Helpful but not required, and risky to force — see Wikipedia and Wikidata.
How do you claim and correct a panel?
Once a panel exists, claim it through the ‘Own this knowledge panel?’ link and verify via Search Console or an official social account. Claiming lets you suggest corrections to the featured image, description and links. Correction speed depends on the underlying sources: if the graph shows a wrong fact, fix the fact at the source it cites (often Wikidata or a major directory), then suggest the edit.
What about engines beyond Google?
Bing maintains its own entity graph that feeds Copilot and, because ChatGPT search depends on the Bing index, indirectly shapes ChatGPT’s brand answers. Registering in Bing’s ecosystem — see knowledge graph registration — covers the second front. Perplexity, which processes over a billion queries monthly with sentence-level citation, grounds entity answers against whatever sources it retrieves in real time, which makes the consistency of your public footprint the controlling factor everywhere.
Common pitfalls
The classic failure is a fragmented identity: a rebranded company whose old name still dominates directories, so the graph merges two entities or builds the panel on stale facts. Second, teams edit Wikidata or Wikipedia promotionally and get reverted or flagged, damaging trust signals. Third, many never claim the panel, leaving wrong logos and descriptions live for months. Finally, some expect the panel to appear on demand — it emerges from accumulated corroboration, typically over months, not from a single schema deployment.
What does a healthy entity footprint look like?
Search your brand incognito and ask three engines about it. A healthy footprint: a claimed knowledge panel with correct facts, a Wikidata item corroborating them, official profiles ranking on page one, and AI answers that state your category, founding facts and offering without hedging or confusion. Anything less is a to-do list, and the fixes always start at the sources — schema, Wikidata, profiles, coverage — never at the answer box itself.
Where to go next
Register your entity step by step in registering in knowledge graphs, connect your profiles with entity linking, and then own the results page with brand SERP optimization.