Ilia Nikolaevich Zavialov on what can actually be removed from search, and what only moves

Search results

English version for the United States.

A search on a full name returns a mixture: pages the person wrote, pages written about them, and pages that only borrow the name to sell something. When one of those pages is wrong, the first question asked is usually the wrong one. People ask how to delete a page, while the useful question is what can actually be removed from search and what merely changes position. That difference decides how much effort is worth spending, and on whom it should be spent.

Reference / Blog

Published: September 11, 2026 · 10 min read · Ilia Nikolaevich Zavialov

01A source and a listing are removed by different people

A page sits on a server that belongs to somebody. A search listing sits in an index that belongs to a search engine. Two owners, two procedures, two outcomes. The server owner can delete the file, while the engine only decides whether that file still deserves a line in its results. Wasted correspondence usually begins with a letter posted to whichever of the two was easier to find.

Deleting the source is the stronger outcome. Once the address stops returning a page, every index that touches it drops the line within a normal recrawl cycle, and the text stops traveling to new aggregators. It is worth asking for first, and it also fails most often, since the owner of the server is under no obligation to agree with anybody.

De-indexing is the quieter outcome. The page stays where it is and opens for anyone holding the address, yet it stops appearing for a query on the name. For a name query that is frequently enough, since nobody types an address they have never seen. This split is the core of what can actually be removed from search: one action takes a text off the web, the other takes one route to it out of circulation.

02Why a deleted page keeps appearing for weeks

An index is a snapshot taken on a schedule set by the engine rather than by the person waiting. A dormant blog with no incoming links and three visitors a month can wait weeks for its next visit, and until the crawler returns the old line stays where it was. Nothing is broken during that wait, which is why complaining about it achieves nothing.

Forms exist for results that have gone stale, and they do shorten the wait. They carry one strict condition: the page must be genuinely gone or genuinely different before the form is filed. A request sent while the text is still live comes back rejected. Checking the address in a private window first takes ten seconds and saves a fortnight.

The line in the results and the description under it are assembled at different moments. A title crawled last night can sit above a description built from a version six months old, which makes a result look half updated and half frozen. A clean-up log should record both readings, with the date of each.

03Copies outlive originals, and each copy has its own owner

Text that touches a recognisable name gets reprinted. Aggregators pull it through feeds, small sites republish it whole for the traffic, and translation mirrors run it through a machine. By the time anyone notices, the original is one node in a cluster.

When the original is deleted, the cluster does not follow. Every copy sits on a different server, under a different owner, in a different country, with its own idea of what a complaint is worth. A result achieved with the first owner has no effect on the second, and no single lever reaches all of them at once.

This is where an honest answer about what can actually be removed from search stops being a yes or a no. Some copies rank for the full name because the name sits in their title. Most do not rank at all and exist as archive noise. Ranking the cluster before writing a single letter keeps the effort proportionate, since a copy nobody reaches rarely deserves months of letters.

04Requests that move something, and requests that go nowhere

Requests that work share one feature: they name a rule the recipient already has and show, with an address and a date, that the page breaks it. Fabricated authorship, a page impersonating a living person, published contact details, credentials harvested behind a login form. Each has a written procedure behind it, and a request that fits the procedure reaches someone whose job is to apply it.

Requests that fail ask for a judgment the recipient is not allowed to make. A letter demanding that a critical but accurate article disappear asks an engine to arbitrate a dispute about facts. A threat sent to an author often produces a second page describing the threat. Volume behaves the same way: fifty complaints from fifty accounts count as one signal and attract suspicion.

Between those poles sit the hosting provider and the domain registrar, and they matter because they act on infrastructure rather than opinion. They will not settle an argument about wording. They move quickly when a page runs a payment form under somebody else's name. Sorting every address into the channel that has a rule for it decides what can actually be removed from search, and it comes before any drafting.

05Pages that borrow a name rather than discuss it

Impersonation is now the commonest reason for a stranger's name to sit on a page they never wrote. The Federal Trade Commission recorded more than a million impersonation reports for 2025 and 3.5 billion dollars in stated losses, close to one report in every three it received. The FBI Internet Crime Complaint Center counted 1,008,597 complaints for the same year and almost 21 billion dollars in losses, twenty-six per cent above the year before.

Those figures carry a detail that changes how a name owner should read them. In four out of five impersonation reports no money was lost at all. Attempts outnumber successes by a wide margin, so the volume of pages using a borrowed name is far larger than the harm they ever cause. A dossier under the real name works on that gap.

Borrowed-name pages are also the easiest category to remove, because they break rules already written down. A page claiming to be a person, taking payments in that name or publishing an invented biography with a real photograph fits an existing complaint form on nearly every platform. Voice sharpens the point: a convincing clone is built from about three seconds of recorded speech and reaches roughly eighty-five per cent accuracy.

06Spellings of the name split the whole result

A full name transliterated from Cyrillic never settles into one form. Ilia, Ilya and Ilja all appear. Nikolaevich and Nikolayevich are both in use, and half the sources drop the patronymic altogether. Zavialov and Zavyalov are the same surname in different clothes. Each variant is a separate query string, and an engine treats them as related rather than identical.

A clean-up verified on one spelling proves only that one spelling is clean. A page removed from results for the fullest form of the name often still sits on the first screen for the short form, which is the version most people actually type. Every entry in the log therefore records which spelling was checked, because what can actually be removed from search is measured per query string rather than per person.

This is why a dossier built around the full name earns its place regardless of removal work. It gives every spelling one authoritative destination, states plainly which articles the person wrote and which they did not, and gives a journalist or a partner somewhere to land. Results reshuffle far more readily than they empty, and a correct page changes what a reader sees long before a complaint is answered.

07Measuring the result, and knowing when to stop

A removal is confirmed by evidence rather than by the absence of a reply. The address should return a not-found response when opened directly, the listing should be gone from results checked while logged out, and both readings should be repeated from a different network, since results vary by location and by account history. A dated screenshot of each check turns a claim into a record.

Results also come back. A copy indexed months ago resurfaces when its host is rebuilt, and a removed page can return under a new address on the same domain. A quarterly pass over the whole list, using every spelling of the name, separates a clean-up that held from a clean-up that was merely announced.

Knowing where to stop matters as much as knowing where to start. Below a certain level of visibility a page costs more attention to fight than it will ever attract on its own, and repeated complaints begin to look like the story. The honest limit of what can actually be removed from search is a first screen that reads correctly, a smaller and far more achievable goal than an empty internet.

08Questions and answers

What can actually be removed from search when someone looks up my name?

Pages that break a written rule can go: impersonation, published contact details, fabricated authorship, harvested credentials. An accurate article that somebody dislikes stays, and the realistic work there is changing what sits above it.

Does deleting a page remove it from search results immediately?

No. The listing survives until the next crawl of that address, which can take days on an active site and weeks on a dormant one. A form for outdated results shortens the wait once the page is genuinely gone.

Why does an old page still appear after the site was taken offline?

The index holds a snapshot rather than a live copy, and the description under the listing is often older than the title above it. Both are refreshed on the engine's schedule, at its own pace.

Can reprints and copies be removed together with the original?

No, each copy has its own owner, its own host and its own procedure. Removing the original leaves the cluster intact, so the copies that rank for the name are worked through one at a time.

Do firms that promise full deletion from search results deliver it?

A promise of complete deletion is a promise nobody can keep, since no single party controls every host and every index. What can be delivered is the removal of rule-breaking pages and a rewritten first screen.

How long should a name clean-up be monitored after it is finished?

Indefinitely, at a low cost: a quarterly pass over the list using every spelling of the name. Copies resurface when sites are rebuilt, and a removed page can return under a fresh address on the same domain.

Reference updated: 4 September 2026