There is a state between "no panel" and "panel" that nobody warns you about. Google has an entity for you in its knowledge graph. If you paste its code into the address bar, the panel appears with your photo and your subtitle. You may have claimed it. And when you type your name into the search box like a normal person, nothing shows. This is a triggering problem, not an existence problem, and it is where most real people's panels sit for months. Here is what decides whether a panel triggers, a worked example from one of our own records, and the things that actually move it.
Existence versus triggering
Google's knowledge graph holds an entity for every person, company and thing it is confident about. Each entity has an identifier, the kgmid, which looks like /g/11abc2defg. Having an entity means Google knows you are a distinct thing with some facts attached. Showing a panel for a search is a separate decision: for this query, in this country, is the searcher most likely to mean this entity, and is Google confident enough in it to put a box on the page?
Those are different thresholds. The first is "do I know who this is". The second is "am I sure enough to answer with it". A name that belongs to several people, an entity with few facts, or an entity whose sources disagree can clear the first and fail the second indefinitely.
The test, in case you have not run it: google.com/search?kgmid=/g/YOURCODE in the address bar, All tab. If the panel appears, you are in the triggering problem. If not, read why your knowledge panel disappeared, which covers the other case. If you have never seen the code, search your name with your company or city added; the moment a panel appears, the code is in the address bar.
What decides triggering
Google does not publish the formula, but the inputs are visible from the outside, and they are the same every time.
Ambiguity of the name. The more people share it, the higher the bar. A unique name can trigger on thin evidence; a common one needs Google to be confident that most searchers mean you, which usually means your entity is far better documented than the others.
Depth of the entity. A panel with a photo, a subtitle and one attribute is a thin entity. Google is more willing to show entities it can say several things about: born, educated, works for, wrote, appeared on. Each of those comes from a source stating it plainly.
Agreement of the sources. If your site, your LinkedIn, your Wikidata entry and three articles all give the same name form, role and facts, confidence is high. If one of them says you hold a job you left in 2019, confidence drops, and that stale page can feed a second, competing entity for the same name.
The absence of a duplicate. When Google holds two entities for one person, it will often show neither on the plain name search, because showing one would be a guess. Duplicates are usually fed by one stale third party bio.
Searcher behavior. Over time, what people click after searching the name tells Google which reading is right. You cannot control this directly; you can make your record the obvious thing to click.
A real example
One of the records we maintain is Dax Hamman's, at dax.fyi. His panel was claimed and verified in September 2026, signed in as an account on his own domain with the domain listed as the panel's website. By early October the kgmid test resolved every time: a headshot, the subtitle "Entrepreneur", and Education as the only attribute shown. A plain search for "Dax Hamman" showed no panel at all.
The diagnosis was textbook. A second, bare entity existed for the same name, fed by an old author bio on a marketing trade site that still listed him as an executive at a company he founded and left years ago. The claimed entity was thin: one attribute, a generic subtitle, and no website or social fields populated. Two entities, neither deep, so Google showed neither.
Three things went in, all on the same day. A feedback report under the claimed panel's Overview box, category "Incomplete", stating that the second entity was a duplicate needing a merge and citing the record page and his Wikidata entry. A second report flagging the subtitle as "Outdated" with the current role and a citing page. A third asking for the website and social fields to be added, with the sameAs list on the record as the source. In parallel, the stale bio was chased by email with a link to the record, and the record's home page was reworked so that it, and not the media kit, owned every fact about him in plain sentences, because the two pages had been reading as interchangeable bios and splitting the entity's weight between them.
There is no dashboard for any of this. The signals are the panel changing, usually inside about three weeks, or an occasional email from Google asking for more evidence. We check monthly: the plain name query in a private window, and both kgmids. That is the whole monitoring process, and it is the one in the course we give every client.
What moves a stuck panel
- Make the entity deeper. Every attribute a panel can show comes from a page stating it in words. Born, educated, citizenship, based, current role, books, shows: put them in plain sentences on your record page, in an "at a glance" block, and mirror them in the hidden layer. Knowledge gain explains why facts that exist nowhere else online carry the most weight.
- Find and kill the duplicate. Search your name with old roles, old companies and old cities. A second panel or a bare header with a stale subtitle is the competing entity. Report it as a duplicate under "Incomplete", citing your record, and fix the page that fed it: email the outlet, update the directory, redirect the old site.
- Make the sources agree. One name form everywhere; the same headshot everywhere; the website field on every profile pointing at your own domain; the same role and dates on your site, LinkedIn and Wikidata. This is tedious and it is where most of the confidence comes from.
- Populate the claim fields. After the claim, suggest the featured image, the subtitle and the website and social links. Those fields being empty is itself a thinness signal, and filling them is the one direct lever you have.
- Add corroboration with context. Interviews and articles that use your full name together with your company or city or subject. Each one is a page telling Google which reading of the name is right. How to get a Google knowledge panel covers the sequence and why it has to be in that order.
- Wait and measure. Monthly, in a private window: the plain name, then the kgmid. Record both. The panel is back when the name query shows it without the code in the address bar.
Where the panel shows when it does
A note for the people who got here by searching "where is the knowledge panel on google", because the answer has changed. On a desktop it is the box on the right of the results. On a phone it sits at the top, often collapsed to a header with a "More" link. In AI Mode and AI Overviews there is no panel as such; the facts from the entity are folded into the generated answer, which is one more reason the entity's depth matters. And on the Web tab there is never a panel, because that tab is designed to show links only. If you are checking for yours, use the All tab, signed out, in a private window, and do not trust a single check; results vary by country and device.
The honest timescale
A triggering problem resolves when Google's confidence crosses its threshold, and that is driven by the evidence changing, not by the calendar. In our experience the duplicate and stale source work moves things within weeks once the source is fixed; entity depth takes a cycle of indexing after the record is updated; and name ambiguity against a well documented namesake can take months or, if they are famous enough, never fully resolve on the bare name. None of that is a promise. It is what the mechanism implies, and what we have seen.