Almost everyone has one. A specific old image that has somehow become the canonical picture of them online. A conference photo, an old profile shot, a team page from a company you left years ago. You did not choose it as your representative image. The internet did, and it has been quietly reconfirming that choice ever since.
Understanding why is useful, because it tells you which remedies are worth your time and which are guaranteed to disappoint. Most of the standard advice targets the wrong variable.
Why age is an advantage, not a handicap
The intuition most people carry is that search behaves like a feed, where newer things float to the top. That holds for news queries and almost nothing else. For a query about a person, ranking systems are answering a different question: which result is most likely to be the one this searcher wants? The strongest available signal for that is accumulated evidence. How many other pages point at it, how long it has survived, how often it has been chosen before.
Every one of those signals is a function of time. A photo published in 2014 has had a decade to accumulate them. A photo published last month has had a month. They are not competing on equal terms and were never meant to.
What an old photo has accumulated
Inbound links come first. Every page that ever referenced the original is a vote, ten years of votes is a lot of votes, and they keep counting long after everyone involved has moved on.
Then copies. Old images get syndicated, scraped, hotlinked, re-uploaded and archived, so one 2014 photo may exist at forty URLs. Each copy independently reinforces that this is the image of you.
Host authority is the underrated one. Old photos of people tend to live on institutionally durable sites: a university, a news outlet, a professional body, a conference archive. Those domains carry weight that a personal site or a social profile does not.
Click history is the fourth. If a result has been shown and chosen for years, that behaviour is itself a ranking input, and popularity compounds.
Only the fourth has anything to do with the photo being good, or current, or how you would like to be seen. The other three are inertia.
Why face search makes this worse
Everything above describes ordinary text and image search. Face search behaves differently, in a way that removes even the small comfort of eventual drift.
A face-search engine is not ranking pages by authority. It computes a numeric template, a faceprint, from the photo you upload, compares it against every faceprint in its index, and returns matches sorted by similarity score. There is no recency term in that calculation. A 2014 photo that matches strongly will outrank a 2026 photo that matches slightly less strongly, indefinitely, regardless of which one represents you now.
Two things follow, and they are why this is worth understanding rather than just enduring. Publishing newer photos does not dilute the old one. In ordinary search, flooding the zone with fresh material can genuinely shift the ordering. In face search it does the opposite of helping, because every new photo you publish is another faceprint added to the index, sitting alongside the old one rather than replacing it.
And deleting the source does not remove the match. The faceprint was derived at crawl time and is stored independently of the image. Take the original page down and the template can persist, still matching, still returning the same result.
That is the mechanical reason the usual advice, which is to post more recent photos, produces so little effect. It is aimed at a ranking system that is not the one returning the result.
What moves the result
Go after the copies rather than the original. If one photo exists at forty URLs, removing the source does close to nothing. A reverse image search will tell you where the copies are, and the handful hosted on sites with real authority are the ones that matter.
Ask the durable hosts first. Universities, publishers, professional bodies and conference organisers update or remove pages more often than people assume, and they carry the most ranking weight, so a single success there is worth a dozen elsewhere.
Where you can, ask for a replacement rather than a removal. If an organisation will swap the image on an existing page, you keep that page’s accumulated authority and change what it points at, which beats removal outright on a high-authority domain.
Handle the face index separately, since none of the above touches it. Face-search engines have their own removal processes and they operate on the faceprint rather than the page. Then repeat it, because engines re-crawl, and a removal that is not maintained decays as the old photo gets rediscovered at some copy you missed.
What doesn’t work
Posting more recent photos is marginal in ordinary search and actively counterproductive in face search. Setting old accounts to private does nothing about copies already made, and nothing at all about a faceprint already derived. Deleting the original and stopping there removes one node from a graph with dozens.
Waiting does not help either. The accumulated-weight advantage does not decay on any useful timescale, and in face search there is no time term to decay in the first place.
An old photo outranking a new one is not a problem with the photo. It is a problem with everything that has attached itself to that photo over ten years: the links, the copies, the hosts, and in the case of face search a stored numeric template that has no opinion about what year it is.
Address those and the result changes. Keep publishing newer photos at it and you will be doing steady work against a system that is not measuring the thing you are changing.
The 2014 version keeps winning.
FacePrivacy files removal requests with the major face-search engines and keeps filing as they re-crawl, so an old photo stops being the canonical answer to your face.
Start protection →