Writing  ·  Essay  ·  June 2026

Why your Knowledge Panel is doing more SEO work than your homepage.

The entity layer that Google and LLMs both read before they read your content. How Knowledge Panels and Wikidata feed AI Overviews, and a five-step audit to find your gaps.

Most B2B marketing teams spend their SEO budget optimising pages that a significant fraction of their target buyers will never directly see. Buyers who search for a brand in Google's AI Overviews, or ask ChatGPT to describe a vendor category, are not reading your homepage. They're reading a synthesised answer drawn partly from your Knowledge Panel, your Wikidata entry, and the entity signals that travel through Google's Knowledge Graph.

Your Knowledge Panel is the interface between your brand entity and every AI-mediated search result. Most teams have never looked at it seriously. Here's why they should, and how to fix what they find.

What a Knowledge Panel actually is.

A Knowledge Panel is Google's structured summary of an entity: a company, a person, a product, a place. It appears on the right side of branded search results and is populated from Google's Knowledge Graph, which itself draws from a mixture of sources: Wikipedia, Wikidata, your own structured data, and a range of third-party data providers.

The Knowledge Panel is not your homepage excerpt. It is not a meta description. It is a separate, structured representation of your entity that Google maintains and updates independently of your website. You can influence it (via Wikidata edits, Knowledge Panel claim verification, and ensuring your schema data is consistent) but you cannot directly control it.

Why does this matter for AI search? Because Google AI Overviews draw from the Knowledge Graph when constructing answers about organisations, categories, and products. When a buyer searches "what is the best observability platform for Kubernetes" and Google constructs an AI Overview, the entity data it has about Datadog, Grafana, or your product comes partly from the Knowledge Graph, not just from crawled pages. Similarly, ChatGPT and Claude have entity representations of well-known companies built into their training data, and those representations often trace back to the same reference sources (Wikipedia, Wikidata) that feed the Knowledge Graph.

The practical implication: fixing your Knowledge Panel and Wikidata entry is one of the highest-leverage technical interventions available in both traditional SEO and GEO simultaneously.

The five-step Knowledge Panel audit.

Run this audit for your brand and for your two or three closest competitors. The competitor audit is often more revealing than your own. It shows which entity signals your category leaders have built that you haven't.

Step 1: Check what your Knowledge Panel currently shows

Search your brand name in Google. If a Knowledge Panel appears on the right: read every field. Check the industry classification, the founding date, the headquarters, the description text, the products listed, the key people named. Compare each field against your current reality. Note every discrepancy.

If no Knowledge Panel appears: your brand is either not established enough as a Knowledge Graph entity, or the signals that would trigger one are absent. This is itself a gap. It means Google doesn't have enough structured confidence about your entity to surface a panel.

Step 2: Audit your Wikidata entry

Go to wikidata.org and search your company name. If an entry exists, check: instance of (should be "business" or the appropriate industry type), industry, founded date, headquarters location, official website, key executives (linked to their own Wikidata entities where possible). If fields are missing or wrong, they can be edited directly. Wikidata is a public, collaborative database.

If no entry exists: creating one is the single highest-leverage GEO intervention available to most B2B companies. A properly structured Wikidata entry with accurate properties and linked references is the primary signal that triggers Knowledge Panel generation and feeds into the entity representations LLMs build during training.

Step 3: Check Wikipedia

Does a Wikipedia page exist for your company? If yes: does the lede paragraph describe your current product and positioning accurately? Are the sources cited credible and current? If the description is out of date, the fix is editing the page with accurate information and updated references. If no page exists: Wikipedia notability requirements are a real constraint, but if your company has genuine editorial coverage in credible publications, a case can be made. Don't create a promotional page. Wikipedia editors will delete it. Create an encyclopaedic entry with proper citations.

Step 4: Verify your Organisation schema

Your homepage should have an Organisation schema block that includes name, url, logo, description, foundingDate, numberOfEmployees (or a range), address, and sameAs links to your Wikidata entry, LinkedIn page, Crunchbase profile, and Wikipedia page where applicable. The sameAs array is the signal that tells Google's entity resolution system that these different sources are all talking about the same entity. Missing sameAs links are the most common schema gap I find.

JSON-LD · Organisation schema with sameAsRecommended
{ "@context": "https://schema.org", "@type": "Organization", "name": "Your Company", "url": "https://yourcompany.com", "logo": "https://yourcompany.com/logo.png", "description": "One sentence. Specific. Uses the category language buyers search for.", "foundingDate": "2019", "address": { "@type": "PostalAddress", "addressLocality": "London", "addressCountry": "GB" }, "sameAs": [ "https://www.wikidata.org/wiki/Q[YOUR_ENTITY_ID]", "https://en.wikipedia.org/wiki/Your_Company", "https://www.linkedin.com/company/yourcompany", "https://www.crunchbase.com/organization/yourcompany" ] }

Step 5: Verify description consistency

Pull the description text from: your homepage, your Wikidata entry, your Wikipedia lede (if one exists), your Crunchbase description, and your LinkedIn About section. Are they saying the same thing about your category, your customer type, and your core value proposition? If they diverge (different vocabulary, different category placement, conflicting scope) you have a description coherence problem that will suppress both Knowledge Panel accuracy and LLM entity representation quality. This connects directly to the entity coverage map framework for diagnosing and fixing description gaps.

Claiming your Knowledge Panel.

If your brand has a Knowledge Panel, you can claim it through Google Search Console, which gives you the ability to suggest edits, though Google retains editorial control over what appears. Claiming is worth doing because it gives you visibility into what Google's Knowledge Graph says about you and a formal channel to flag inaccuracies.

More important than claiming, however, is fixing the underlying data sources that feed the panel. Google's Knowledge Graph draws primarily from Wikidata, Wikipedia, and structured data on your own site. Correcting those sources (rather than trying to edit the panel directly) is the more reliable path to accurate, current Knowledge Panel data.

The same principle applies to LLM entity profiles: you can't directly edit what ChatGPT or Claude knows about your brand, but you can ensure that the sources they draw from describe you accurately, consistently, and in the terms your buyers use.

NOTES
  1. Wikidata editing is publicly available to anyone with an account. The barrier to improving your entry is low. The main constraint is notability and verifiability: claims need to be supported by references to credible, publicly accessible sources.
  2. The schema example above uses a minimal set of properties. Additional schema properties (numberOfEmployees, foundingDate, contactPoint) add value but the sameAs array is the highest-priority addition for most brands that don't already have it.
  3. Knowledge Panel data updates are not instantaneous. After fixing Wikidata or updating schema, expect 2–6 weeks before the changes propagate into the displayed panel.

/ Frequently asked

My company doesn't have a Knowledge Panel. Is that a problem?

It depends on your size and stage. For early-stage companies with limited press coverage, the absence is expected and the right response is to build the signals that would eventually trigger one (Wikidata entry, Wikipedia page if notability criteria are met, strong Organisation schema). For Series B+ companies in competitive categories, a missing Knowledge Panel usually indicates a solvable entity signal gap.

Can I edit my Wikidata entry to say whatever I want?

No. Wikidata is a public database with community oversight. Claims need to be referenced to verifiable sources. You can correct factual errors, add missing properties that are verifiable, and link to credible sources. You can't add promotional claims or unverifiable assertions without them being reverted by other editors.

Does fixing the Knowledge Panel help with ChatGPT citations specifically?

Indirectly. ChatGPT and other LLMs draw primarily from their training data, which was assembled before a specific knowledge cutoff date. Wikidata and Wikipedia, being persistent reference sources, are well-represented in training data. Fixing your Wikidata entry now affects models that refresh their training data in future update cycles and also helps Perplexity and Google AI Overviews, which use live retrieval on top of trained models.

tc
/ Written by

Thomas Cox

Twelve years in B2B SEO, most recently at VP level. Now independent — helping companies stay discoverable as buyer search moves into ChatGPT, Perplexity, and Google AI Overviews. Remote · UK.

About Thomas Work with me
Keep reading

More from the notebook.

June 2026 · 10 min

Entity coverage maps: a working template.

READ →
June 2026 · 9 min

Source-worthiness, explained for SEO teams.

READ →
June 2026 · 14 min

The B2B GEO roadmap I'd build if I started today.

READ →

Want this in your inbox?

Roughly one essay a month. No drip sequences, no upsells.

Subscribe