The decision email says the URL has been removed. The next morning the name search still looks like a problem. There is a grey “Cached” link under a neighbour result. There is a row of sitelinks that includes the old headline. There is a Knowledge Panel on the right that still quotes the paragraph you spent six months getting off the first page. None of this means the removal failed. It means you finished one job and started another.
If you remember one distinction from this piece, make it this one: the index entry versus the chrome around it. The de-indexing guide is about whether Google still treats a URL as a live document. This piece is about the leftovers after that question has been answered — the copies, the extra links, the panel, the image thumb, the autocomplete stem. They are why people come back to the desk convinced Google “put it back”. Usually Google did not. The rest of the page never left.
The cache that outlives the result
Google’s cache is a stored copy of a page the crawler already fetched. It exists so a result can be shown when the host is slow, and so the outdated-content tool has something to compare against the live response. It is not a library. It is also not instant.
After a personal-information or legal removal, the live blue link is what goes first. The cache URL — webcache.googleusercontent.com, or the “Cached” control on a result that has not fully died — can linger. After a publisher 410, the same lag appears. We treat a cache still serving the old text a fortnight later as a freshness failure, not as a new publication, but we still capture it. Recruiters click cached copies. So do journalists.
Two instruments help, and only in that order. First, the live URL must tell the truth: a hard 410, or a noindex on a rewritten page, as the de-indexing piece sets out. Second, the outdated-content tool nominates the stale snippet or the stale cache. Filing the tool against a URL that still returns 200 with the old article is a waste of a receipt. The cache is doing its job. The publisher is not.
A cache is also not the Wayback Machine. Clearing one does nothing to the other. If the leftover is web.archive.org, you have left this chapter.
Sitelinks, and the form that is no longer there
Sitelinks are the extra blue links Google sometimes prints under a domain result — About, Contact, and, unhelpfully, the blog post you thought you had buried. They are algorithmic. They are built from the site’s structure, from the sitemap, and from what searchers click. They are not a menu you get to edit.
Search Console once offered a sitelinks demotion tool. It has been gone long enough that any agency still selling “sitelink removal” is selling a memory. What remains is plumbing. If a sitelink points at a URL that now 410s, it should drop after recrawl. If it points at a URL that still 200s, the sitelink is correct. If the damaging phrase is coming from an old <title> on a page you rewrote but did not recrawl, you have a freshness problem, not a sitelinks problem.
Two mistakes make sitelinks worse. Redirecting the scandal to the homepage can teach Google that the old headline is the homepage. Leaving the article in a tag archive or a “related posts” rail gives the algorithm a live destination. Clean the host before you stare at the chrome. A small site with one notorious URL will grow sitelinks around that URL, because that is what the site is, as far as the crawler can tell.
Knowledge Panels are not a search result
A Knowledge Panel is the card on the right — or at the top, on a phone — that tries to tell a searcher who or what the query is. It is assembled from Wikipedia, Wikidata, news graphs, structured data, and a set of sources Google does not itemise for you. A single de-indexed URL will not empty a panel that is still drawing on a Wikipedia paragraph, a Wikidata description, or a news article that remains live.
This is the leftover that feels most like a conspiracy and is most like a database. People file the Search removal form against the panel itself. The form does not accept a panel. It accepts a URL.
What you can actually do is narrower than the industry pitch.
Feedback on the panel. The three-dot menu and the “Suggest an edit” path will take a factual objection: a wrong workplace, a wrong city, a wrong image. It will not take “please delete me.” Google is not offering a right to be unpublished from the Knowledge Graph.
Claim the panel if you are the organisation or the public person it describes. Claiming unlocks some controls over the official site and the social links. It does not unlock a delete button, and it is a bad idea to claim a panel in order to pick a fight with a news graph you cannot edit.
Edit the sources. If the panel is quoting Wikipedia, the job is at Wikipedia, under the biography-of-living-persons rules. If it is quoting a news article, the job is at the masthead. If it is quoting Wikidata, edit Wikidata with a citation. Google will refresh. It will not refresh because you asked it to stop reading the public web.
Images in the panel. A mugshot or a stolen headshot can sometimes be moved through the image-removal products. The panel will then pull the next image it can find. If the only photographs of you on the web are the ones you hate, a successful image takedown leaves a blank or a worse substitute.
A panel that names a private person who is not a public figure is rarer, and it is the one case where a personal-information product and a legal letter have more bite. Still attach URLs, not a screenshot of the card. The reviewer does not adjudicate chrome.
The smaller leftovers
Several other objects survive a clean de-index and get reported to us as “Google put it back.”
Image results. Google Images is a separate index. A page removal does not always take the thumb. Use the image product, and expect a new scrape to create a new URL.
Autocomplete and “People also search for”. These are query logs, not copies of the article. They fade when the clicks fade. There is no reliable public form that deletes a suggestion for an ordinary Australian name. Paid “autocomplete removal” is, in most of the files we have seen, a waiting room with an invoice.
Discover, news carousels, AMP, Translate. A story that ran as news can linger in surfaces that are not the ten blue links. Wrappers such as translate.google.com can keep serving a copy after the origin has 410’d. If the publisher has taken the piece down, those surfaces should follow; if not, you are in the source business. Nominate the wrapper if it is the URL people are actually opening.
AI Overviews. The generated paragraph at the top of the results is not a cache and it is not a panel. It is a summary of sources that are still being read. That job has its own piece: what AI Overviews do with harmful search results.
A sequence for the week after “removed”
- Confirm the nominated URL is actually gone for the query you care about, in incognito, from Australia, and from one overseas window if that is the harm.
- Open the cache URL. If it still serves the old text, check the live status code, then file outdated content.
- Click every sitelink under the domain. 410 or noindex the ones that still carry the old headline. Do not 301 them to the homepage.
- Read the Knowledge Panel as a bibliography. List the public sources it is quoting. Work those sources. Do not file the panel as if it were a URL.
- Check Images, autocomplete, and any AI Overview. Diary the ones that are still wrong. Do not invent a form that does not exist.
- Recheck in a fortnight, then in ninety days. Panels and sitelinks are slower than the index entry. Brokers and scrapers will also have made new URLs by then.
A removal that only hides the URL is a successful removal. It is not a rebuilt results page. The agencies that send you a screenshot of a blank ten-blue-link view and call the job done are not lying about the URL. They are lying about the page.
If you want the desk to look at a general question about leftovers, use contact. Do not send the intimate image, the court file, or a zip of every screenshot. We publish method. We do not run your case.