Search Engine Deindexing

What an engine can actually do to a page

Deindexing is the removal of a result, not the removal of a page. The distinction sounds pedantic until a business has paid for the first and been told it was buying the second, which happens constantly in this market. The engines themselves are not coy about it. Here is Bing:

“Bing cannot remove content from the internet itself, but Bing and Copilot can update what appears in Bing search results and Copilot experiences when we detect that a URL has been deleted, marked as non-indexable, or replaced.”
— Bing Webmaster Tools, “How to permanently remove content from Bing”, read 12 August 2026

Google says the same thing in its own help documentation: it can stop showing a URL in its results, and it cannot remove that URL from the site hosting it, because the person who owns the site controls it. Its instruction for everything outside its narrow policy categories is to contact the site owner.

So the mechanic is one step removed from the problem. An engine reflects the web; it does not edit it. What deindexing changes is the odds that someone searching a company name encounters the URL. What it does not change is whether the URL exists, whether it can be linked, whether it is quoted elsewhere, or whether the person who published it knows anything has happened at all.

That is a real outcome and it is worth having. It is also, precisely, suppression — which is why this URL carries that verdict rather than the one every vendor uses.

The assistant layer inherits the index

Read Bing's sentence again and notice which product is named inside the limitation. Not just search results — Copilot experiences. The operator has extended the same caveat to its AI assistant, in the same sentence, because the assistant is built on the index. Whatever the index knows, the assistant can repeat.

This is the part of deindexing that has changed since the advice most businesses are reading was written. In 2015 a deindexed page stopped appearing in a list of links. In 2026 the same index feeds an assistant that answers a question about a company in a paragraph, with no results page for anyone to scan, and that paragraph is where a growing share of first impressions now forms. Bing's own wording ties the two together: once it detects the URL is gone or non-indexable, it can update what appears in results and in Copilot experiences that rely on its index.

Note the honest limit in that same sentence, because it matters. It describes what changes going forward. It does not say an assistant's answers are derived only from the live index, and no operator captured in this project publishes that mapping. Two further disclosures show why that gap is not academic:

“We also require licensees in our data licensing arrangements to stop using deleted posts and comments and provide them with compliance tools to help automate public content deletions. We cannot guarantee that third parties have deleted copies of Reddit public content they've made without your or our permission.”
— Reddit, Public Content Policy, read 12 August 2026

Reddit names its licensees as brand-monitoring companies, researchers, and “large language model makers.” And Google's Transparency Report index, read the same day, now carries statutory AI-training-data filings: a “CA Civ Code § 3111 (CA AB 2013) AI Training Data Transparency Summary” and EU AI Act summaries of training content for the Gemini 3 and Gemma 4 model families. There is no defamation removal report in that index, and no removal-from-training-data report either.

Put those side by side and the shape is clear. Ingestion of published material is now documented under statute in two jurisdictions. Removal from it is documented nowhere. That is not a claim that a specific deindexed page persists inside a specific model — nobody publishes that, and I am not going to invent it. It is the reason deindexing is a smaller claim in 2026 than it was a decade ago, and the reason acting early on something at its source is worth more than acting late at the search layer.

The only lever that reliably moves an index

Every engine describes the same route, and it runs through whoever controls the URL rather than through the engine. Delete the page, return a 404 or 410 status code, or add a NOINDEX directive — then the crawler comes back, processes the change, and the URL stops appearing. Bing says the URL may then no longer appear in its index, its results, or Copilot experiences relying on that index. Google's guidance for permanent removal is the same set of moves, with a specific warning against using robots.txt as the blocking mechanism, since a URL blocked from crawling can still be indexed from links pointing at it.

Two Google tools get mis-sold as substitutes for this, and both have the operator's own disqualifier printed on the page:

• Search Console Removals works only on a site you have verified ownership of, and Google states that a successful request lasts only about six months and does not stop Google crawling the URL — only showing it.
• The Refresh Outdated Content tool is for URLs that have already changed or gone. Google's own requirement is blunt: if the URL still exists and the information you want removed is still there, the tool is not for you. Filed against a live page it will be rejected, and a vendor who files it anyway is billing for a rejection.

Which leaves the honest summary. If a business controls the page, deindexing is straightforward and it can be permanent. If somebody else controls it, the request goes to that person or to nobody, and everything else is a narrow policy category most business content does not fall into.

One dated account of the form not working

Filing a removal form and being heard are different events, and there is a dated record of the gap. On 18 February 2026 a user posted to Microsoft's public Q&A forum describing an attempt to use the Report a Concern form for Bing. The submission was made under “Exposed Personal Information”; the reply came back categorized as “Outdated link”, with a message that the system would update itself for old links. A follow-up through the Microsoft account page reporting the mis-categorization drew an instruction to “select the correct reason when submitting the concern.” Both were resubmitted. Nothing came back. Customer service phone numbers were bots; live chat was available for technical support only.

Attribute this correctly, because it matters. That thread is a user forum, not Microsoft policy. The person who answered is an “Independent Advisor”, and a second answer on the same page is machine-generated and carries Microsoft's own disclaimer that AI-generated content may be incorrect. One user's experience is not a pattern, and I have no basis for saying how typical it is. The thread is public and can be read in full.

Its value is narrower and still real: it is a contemporaneous, first-hand account of what this process feels like from outside, and it is consistent with the boundary the operators state themselves. The advisor's answer said it correctly — a search engine cannot remove content directly from a website, and if the information is still on the site, the site owner is who you need. That is the same sentence Bing and Google publish. The form is not a substitute for it.

How many engines you are actually filing with

A proposal that lists an engine-by-engine price — Google, Bing, Yahoo, DuckDuckGo, and whatever else fills the table — is quoting for a shorter list than it looks.

DuckDuckGo says so on its own help page: most traditional search results come from Bing, and to request removal of traditional links appearing in its results, the request goes to Bing first. Once Bing acts, the change is reflected automatically in DuckDuckGo's results. Its page on the subject was read on 12 August 2026.

Other engines syndicate too, and I have not verified each of their current arrangements from their own documentation, so I will not publish a syndication map I cannot support. What is sourced is enough to change how a proposal reads: filings are not a per-engine unit of work, and a line item for each brand name in the list is not describing separate work being done. Ask which engine actually operates the index behind each name on the invoice.

What stays exactly the same afterwards

A deindexed page is still a page, and the list of things unchanged is longer than most people expect:

• The page itself. Live at the same URL, serving anyone who has the link, anyone who bookmarked it, and anyone sent it by a competitor.
• Other engines and other surfaces. A different index, a social platform's internal search, a forum's own search, an aggregator that already copied it.
• Citations elsewhere. Anything already quoting or linking to the page keeps doing so, and those citing pages are separate URLs with their own indexing.
• Copies. Archives, scrapers, syndication feeds and anything licensed for training before the request. Reddit's own policy is the clearest statement that a deletion at the source cannot be promised to reach copies already taken.
• The record of the request. Legal removal requests to Google are forwarded to the Lumen database at Harvard's Berkman Klein Center, where the complained-of URLs are published and searchable.

This is why deindexing is best understood as the last step of a successful removal at the source rather than a route of its own. Get the page taken down or corrected by whoever controls it, then tell the engines so they catch up faster. Reversed — engine first, source untouched — it is a temporary result against a permanent page.

The one case where a business genuinely can remove itself from search

There is a version of this problem where the answer is easy, and it is worth separating from every version where it is not: the page is on property the business owns. An old press release announcing something that went wrong. A stale executive bio. An archived statement written during a crisis in a different decade. A subdomain a former agency built and left behind, still resolving, still indexed, still ranking on the company name.

Here the whole apparatus applies cleanly. Delete or edit the page, serve a 410 where it should be gone for good, add NOINDEX where the page has to stay for another reason, verify the property in each engine's webmaster tools, and use the removals and refresh tools for what they were built for — accelerating a change that has already happened at the source. That is the one situation where removal from search is straightforward, and I find it more often than people expect, usually while auditing what actually ranks on a brand query.

Everywhere else, check the operator's current page before acting. Bing's help documentation, Google's removal tools and Reddit's content policy were all read on 12 August 2026, and every one of them can be rewritten without notice. The mechanics on this page are stable; the forms, the categories and the click paths are not.

Frequently Asked Questions

Published 2026-08-12