How to identify unknown people in old family photos
You find a box of photographs after someone dies. You know it came from your grandmother’s house. You recognize your grandmother in the back row. The young man beside her might be your grandfather at twenty, or it might be his brother.
The first impulse is usually to ask who the face resembles. That is the question most likely to give you a confident wrong answer, because a family is full of people who look alike.
Start somewhere else.
To identify someone in an old family photograph, begin with where the image came from, what date range it supports, and which branch of the family it belongs to. Face-grouping software can help build a short list. A name should be confirmed by context, records, or someone who knew the person.
Start with the source, the date range, and the family branch
A face may help narrow the candidates. The source of the photograph, its date range, and the people around it are usually what separate one candidate from another.
Better opening questions:
- Where did this photograph come from, and who owned it?
- What kind of photograph is it?
- What is the earliest and latest date it could have been taken?
- Which branch of the family is it connected to?
- Who in that branch was the right age, in the right place?
Answer those and the list often shrinks to one or two people. Then a resemblance means something. It is doing the last bit of work instead of all of it.
Preserve the album, captions, and source before you reorganize
Do this before you reorganize anything. Sorting can separate a photograph from the evidence that explains it.
Photograph the album page before you remove a print. The pictures beside it are part of the record. Album order often carries information the isolated photograph no longer does.
Scan both sides. A faint pencil note, a studio stamp, or an address can become readable once you adjust the contrast. The back is sometimes worth more than the front.
Write down where it came from. Whose album, whose house, which box, who handed it to you. The Library of Congress says the same thing in its guidance on treating photographs as evidence. Learn who donated the image. Study the original order of the album pages. Keep notes on why you proposed a name. That applies just as well to a shoebox.
Do not erase or crop away old handwriting, even when it looks wrong. Photograph it first. A wrong caption still tells you who owned the photograph and which side of the family it traveled through.
Keep the record in four parts, so you can see which is which:
Confirmed: Found in Elena Ruiz's album. "Corpus Christi" in pencil on the back.
Possible: The young man at left may be her brother Rafael.
Unknown: The exact date. The other two people.
Conflict: Rafael may have been overseas during part of 1958.
Keeping those categories separate is what stops a possibility from slowly becoming a fact. That is a common way a wrong identification takes hold: someone repeats a maybe until it sounds settled.
How to date an old family photograph using physical clues
Clues are not equally strong, and treating them as though they are is how people arrive at a confident wrong year.
| Clue | What it gives you | How tight |
|---|---|---|
| Dated handwriting, a letter, an event program | A direct date | Often exact |
| Photographer’s studio address | When that studio worked there | A few years, where the address history is documented |
| Album order and nearby dated photos | A context range | Several years |
| Card mount style, backmark, edges | A period inside a format’s run | Broad. Compare with format history and studio details |
| Photographic process and material | An earliest and latest likely period | May not date the individual print |
| Clothing, hair, eyewear | A fashion range | Often broad. Use alongside other clues |
| Buildings, signs, vehicles | Sometimes an exact place and narrow date | Varies, needs confirming |
| Facial resemblance | Nothing about the date | Not a dating clue |
A studio address can be one of the most useful clues, because studios moved. A printed backmark often carries one.
The Smithsonian’s write-up on card photograph backmarks gives worked examples. Samuel Montague Fassett’s Chicago studio ran from 1872 to 1875. His Washington address ran 1876 to 1880. Frederick Gutekunst was at 704 and 706 Arch Street in 1865 and 1866.
If you can date a photographer to an address, you have dated the photograph to a few years. From the back of the card alone.
Clothing can help establish a range, but it is usually weaker than a dated caption, a photographer’s address, or documented album context. Clothes were reused, inherited, regional, and sometimes deliberately old-fashioned.
If the evidence gives you a range, keep the range. “About 1878” or “late 1950s” is worth more than an exact year you cannot support.
What Apple Photos and Google Photos can and cannot tell you
Apple Photos and Google Photos group faces that look similar within your own library. If you have several clear photographs of a relative, the apps may surface older images that seem related. With hundreds of scans, that can save real time.
It is also the part most articles overstate.
Google describes how face groups work in careful language. It builds face models, predicts similarity, and estimates whether two images show the same face. It groups the ones “likely” to match.
It also says the suggestions “aren’t perfect.” When it proposes merging two groups it offers Same, Different, and Not sure, and it warns that a merge cannot be undone. It may group images using signals other than the face, such as photos taken close together in time.
Apple documents recognition, grouping, naming, and correction, and provides a “This is Not” control precisely because the grouping can be wrong.
Neither company publishes accuracy figures, and neither describes these tools as verifying identity. Treat a group as a lead.
Three findings explain why that caution is not excessive:
Age gaps degrade matching badly. In a 2024 study of matching faces across a 40-year gap, human participants matched same-age photographs 82 to 90 percent of the time. Across the age gap that fell to about 66 percent. The study did not test Apple Photos, Google Photos, or damaged snapshots. It shows why a large age gap makes any match less certain.
Relatives are the hard case, and this is a family photo. A 2021 study of face-recognition models on sibling pairs found that telling siblings apart is difficult for these systems, especially when image quality, pose, age, or the visible part of the face varies. A grouping may find the right family and the wrong person in it.
A person who knew them beats the software. In earlier research reviewed by the authors of that 2024 study, former classmates matching yearbook photographs to current images 25 years later did better than strangers, and better still when they had known the person well enough to name them. That is an argument for asking someone who knew the subject at the age shown.
That is the practical finding. If you can, ask someone who knew them at that age before you rely on a software suggestion.
How to ask relatives questions that produce usable clues
“Do you recognize this?” is hard to answer. It asks for everything at once.
Give the person something to work with:
- “This came from Elena’s album and I think it is Corpus Christi. Do you recognize the house?”
- “Was Rafael living in Texas around 1950?”
- “Who would have been at this wedding?”
- “Do you remember this car, this porch, this church?”
- “Does this look like one of the Ortiz brothers, or someone else?”
Ask about the setting, not only the face. People often recognize a kitchen or a street before they recognize a young version of someone.
Avoid asking “Is this Grandpa?” A leading question invites agreement, and a polite yes from someone who is unsure becomes a family fact within a year.
How to record what you confirmed, and what is still uncertain
Professional genealogists do not use a shortcut like three matching clues equals proof. The Board for Certification of Genealogists asks for thorough research and complete citations. It also asks you to resolve anything that conflicts, rather than setting it aside.
In practice, test a name against everything at once. Time, place, apparent age, relationships, provenance, setting, records. Then test it against the other people it could be. That last step is where most identifications quietly fail. If a brother fits just as well, you do not have an answer yet.
Write the confidence into the caption:
Possibly Rafael Morales, left, in El Paso, about 1952. From Elena Ruiz’s album. Not confirmed.
Rafael Morales, left, with his brother Tomás at the family home in El Paso, about 1952. Identified by Rosa Morales in August 2026, supported by a dated letter from the same visit.
And keep a note with the photograph recording who identified the person, when, on what basis, and what still does not fit.
Do not put a guess in the filename. Grandpa_Rafael_1952.jpg becomes fact the moment someone else copies the folder. Unidentified_man_ElenaRuiz_album_c1952.jpg stays honest and stays searchable.
A careful “possibly” leaves room to be right later. A confident wrong name can travel through a family for generations.
Check privacy settings before uploading photos for face grouping
Before uploading a family collection to a service you have not used before, read its current terms.
Google says Face Groups creates face models that “may be considered biometric data in some jurisdictions.” Turning the feature off deletes the models, groups, and labels. Otherwise it keeps them until you delete them or the account sits unused for two years.
Apple says its Photos analysis happens on your device, and that Apple does not access your photos or use them for research and development. Genealogy services set their own rules, and they differ.
A photograph is not automatically biometric data, but a face model derived from one may be treated differently. Some state laws draw that line explicitly.
Practical version: start with the photo library you already use. Think before uploading images of living relatives or children to a public site. Check whether face grouping is optional, and whether you can switch it off.
Keep the original image and the research note together
Do not replace the original scan with an edited copy.
Keep all of it:
- The original front and back scan
- A readable working copy, if you adjusted contrast or cropped it
- The album-page image, if it came from an album
- A note on where it came from
- The names you considered, what supports each one, and what still conflicts
A caption tells the next person what you concluded. The research note tells them why.
An eight-step process for an unknown family photograph
- Photograph the album page before you remove anything.
- Scan the front and the back of the print.
- Record its source, any handwriting, the studio imprint, and the images near it.
- Separate known facts, possibilities, unknowns, and conflicts.
- Establish the widest honest date range before you name anyone.
- Use face grouping and family questions to build a short list.
- Test that list against records, places, known relatives, and the photograph’s context.
- Caption it with a confidence level, and save the research note alongside it.
An unidentified photograph that is scanned, dated, and traced to a family branch is not a failure. It is a photograph someone else can finish.
Common questions about identifying people in old photos
Can iPhone or Google Photos identify people in old photos?
They can group photographs in your library that appear to show the same person, and both let you correct mistakes. Neither verifies identity, and neither publishes accuracy figures for this use. Use a group to decide what to look at next.
Can face grouping recognize someone at different ages?
Sometimes, and less reliably as the gap grows. Research on matching faces across 40 years found accuracy dropped from the eighties into the mid sixties. Childhood to old age is harder still.
What if everyone who would know has died?
Try descendants of cousins and other branches. Local historical societies, town history groups, and church or alumni records also help. People outside your immediate line often kept the same photographs with different captions.
Should I remove photographs from an album before scanning?
Photograph every album page first. Page order, captions, and the images beside one another may all be part of the evidence. If you do take prints out, record where each one sat before you reorganize.
How should I label a photograph when I am not sure?
Use “possibly,” “probably,” or “unidentified,” and include the source and a date range. Do not put an uncertain name in the filename as though it were settled.
Should I keep photographs I cannot identify?
Yes. Scan them, record where they came from, and label them as unidentified. Keep the notes. An unresolved photograph often becomes identifiable when a new relative or record turns up.
If you want somewhere to keep the photographs and what you learned about them together, that is what we build at Perdura. A caption that records who identified someone, and how sure they were, is worth keeping alongside the picture.
Want to keep a story like this?
Perdura is a permanent home for a life story. Start it yourself, free, with real help as you go, or have us build it with you.
Free while we're in beta. No card, and real people to call if you want help. Keeping it online after beta will need payment, and we'll say what it costs this fall.