You search your own name and Google shows you two of yourself. One panel has your headshot and a two-line summary pulled from an old conference bio. The other has your company logo, a link to a podcast you stopped recording in 2023, and a “People also search for” row full of strangers. Neither is wrong enough to report. Together they are a mess.
A duplicate knowledge panel is not a cosmetic issue. It splits every signal Google uses to describe you across two records. Reviews, links, claimed profiles, and citations that should reinforce one entity end up diluted between two, and AI answer engines that lean on the Knowledge Graph inherit the confusion. When ChatGPT or Gemini is asked who you are, it has to pick one, and it does not always pick the one you would.
The good news is that duplicates have a small number of causes and a predictable fix order. This guide walks through both.
Why Google builds two panels for one subject
Google’s Knowledge Graph, launched in 2012, assembles facts about entities from across the web and from licensed and open databases. Google’s own explainer on knowledge panels describes it as a system that “understands facts and information about entities from materials shared across the web.” The key word is understands. It does not ask you who you are. It infers it.
When the evidence points to one subject, you get one record and one panel. When the evidence looks like it might describe two different subjects, the system hedges and keeps them apart. A merge that turns out to be wrong is worse for Google than a duplicate, because a wrong merge hands one person’s career to another. So the default under uncertainty is separation.
That hedge produces duplicates in five recurring ways.
The first is name drift. You publish as “Dr. Sarah K. Lin” on your practice site, “Sarah Lin” on LinkedIn, and “SK Lin” on your book cover. Each variant collects its own cluster of mentions, and if no strong source ties them together, the system treats them as candidates for separate people.
The second is a role split. A founder and a founder’s company get confused, so one record ends up half-person and half-brand. This is common for solo consultants and creators whose company name is their own name.
The third is a source conflict. Two authoritative sources disagree about a fact that matters for identity: birth year, hometown, employer, or job title. A Wikidata item says you were born in 1984 while an old press bio says 1986. Disagreement on identifying facts is the single strongest reason Google keeps records apart.
The fourth is a local and entity overlap. Your business has a Google Business Profile and an entity panel, which is normal, but a second Business Profile also exists for the same location, created by a former employee, an agency, or a data aggregator. Google’s Business Profile help center is blunt about this: “You can have only one Business Profile for each business,” and a profile considered a duplicate “won’t show on Google Search or Maps.” Until one is filtered, both can surface.
The fifth is a project spin-off. Your podcast, book, or show gets its own record, which is correct, but the record borrows your photo or description, so it reads like a second you. This one is often not a true duplicate at all, and the fix is to clarify the relationship, not to merge.
Which kind of duplicate are you looking at?
Before you touch anything, sort your duplicate knowledge panel into one of three buckets, because each has a different owner inside Google and a different fix path.
A Business Profile duplicate shows two local panels with maps, hours, and reviews for the same address. These are handled by Google Business Profile support, and they are the fastest to resolve because a human can act on them.
An entity duplicate shows two panels without map pins, each with a summary, an image, and social links, for the same person or organization. These live in the Knowledge Graph. No support agent merges them on request. You change the evidence, and the system changes its mind.
A related-entity confusion shows two panels that describe different but connected things, such as you and your company, or you and your book. The goal here is separation with clean labels, not a merge.
A quick diagnostic helps. Open each panel and look for its entity ID. On desktop, the share option on a panel produces a URL containing a kgmid parameter, formatted like /m/ or /g/ followed by a string. Two different IDs for what should be one subject confirms a true entity duplicate. If one of the panels has no ID and shows a map, you are in Business Profile territory.
Run the One Entity Audit before you touch anything
At Instant Press we use a simple framework for this, which we call the One Entity Audit. It has four passes, and you run them in order because each one depends on the last.
Pass one is the Name Lock. Pick the exact form of your name, or your company’s name, that you want Google to treat as canonical. One spelling, one middle initial decision, one suffix decision. Write it down. Every later change uses this string verbatim.
Pass two is the Fact Sheet. List the six to ten identifying facts Google is likely to use: full name, role, organization, location, founding year or birth year if public, notable works, official website, and primary social profiles. Then search each one and record every place it appears with a different value. That list is your conflict map. In most audits we run, the duplicate traces back to two or three conflicting facts, not dozens.
Pass three is the Anchor Check. Identify the one page you control that should act as the canonical source for the entity. For a person, that is usually an about page on a personal or company site. For an organization, it is the homepage or about page. That page needs your locked name, your fact sheet values, and structured data with a sameAs list pointing to every official profile. If your anchor page does not exist, or contradicts your fact sheet, nothing downstream will stick.
Pass four is the Source Sweep. Go through the conflict map and fix every source you control first, then request corrections on sources you do not. LinkedIn, Crunchbase, your speaker bios, your author pages, Wikidata if you have an item, and any old press bios hosted on partner sites.
The audit is boring, and that is the point. Duplicates are a symptom of inconsistency, and the audit is how you find it.
Fixing a duplicate Business Profile
If your duplicate is local, the path is procedural.
Start by finding out who controls each profile. If you own both, Google will not merge two verified profiles, so you need to remove the one you do not want from your account first, keeping the profile with the reviews you want to keep. Then contact Business Profile support, choose the option for merging duplicate profiles, and provide both Profile IDs, which you find in the profile’s advanced settings. Explain which one should survive and ask for reviews to be transferred.
If you do not own the duplicate, request ownership through the “Own this business?” flow. If ownership cannot be granted because the listing was created from third-party data and nobody controls it, use “Suggest an edit” on Maps and report that the place is a duplicate or does not exist at that location.
Two cautions. Review replies often do not survive a merge, so screenshot anything you want to keep. And if your duplicate exists because you run two different businesses at one address, that is a separate case in Google’s guidance, and forcing a merge would be the wrong fix.
Fixing a duplicate entity panel
Entity duplicates take longer because you are persuading a system rather than asking a person. The sequence that works is consistent.
Claim the panel you want to keep. Google’s verification flow for knowledge panels lets the subject confirm identity through an official profile such as a YouTube channel, Search Console property, or major social account. Being verified on the right record gives that record a stronger claim and gives you a feedback channel.
Point every owned source at the surviving record. Your anchor page, your social bios, and your structured data should all describe the same entity in the same words. Where your sameAs list currently includes a profile that is feeding the wrong panel, such as an abandoned podcast page that uses your headshot, either update that profile to describe the project clearly as yours, or remove it.
Resolve the identifying conflicts. If Wikidata, an old press release, and your website disagree on a fact, correct the ones you control and request corrections on the rest. A single correction to a high-authority source often does more than ten changes to minor ones.
Use the feedback link last. On the duplicate panel, use “Feedback” to report that it describes the same subject as another panel and include a link to the surviving one. Feedback alone rarely moves anything. Feedback on top of clean sources gives the system the nudge it needs on the next refresh.
Then wait for recrawls. Google has to revisit the sources you changed before it updates its confidence. Pages that are crawled often reflect changes quickly. Old partner bios and directory listings can take weeks.
What not to do while you wait
Do not create a new profile to “replace” the bad panel. A fresh Wikipedia draft, a new Wikidata item, or a new Business Profile adds a third candidate record and makes the duplicate worse.
Do not delete your strongest sources out of frustration. Removing the LinkedIn page or website that feeds the right panel can make the wrong one look like the better match.
Do not pay anyone who promises a guaranteed merge by a date. Nobody outside Google controls the merge decision for entity panels. We have covered the scams around this in detail in our post on buying a Google knowledge panel.
And do not confuse a duplicate with a merge gone wrong. If the second panel belongs to a namesake and Google has started blending your facts with theirs, you need separation, which is a different playbook covered in knowledge panel merged with someone else.
The next step is concrete. Write your Name Lock string and your fact sheet today, before you change a single profile, then fix the anchor page first. Every duplicate knowledge panel we have untangled started there.