How to identify people in old family photos

The short answer

To identify an old photo, record the object before changing it, preserve the front and back, trace where it came from, estimate a date from several independent clues, and ask relatives focused questions. Treat facial resemblance as a lead, not proof. Record confirmed, probable, possible and unknown identities separately.

When people search for how to identify people in old photographs, they often begin with the face. The stronger evidence is usually around the face: the album page, handwriting, studio address, clothing, buildings, neighboring photographs and the person who owned the collection.

The aim is not to produce a confident name quickly. It is to leave an evidence trail that another relative can inspect and improve.

Start with one photograph: Open the free photo-identification notes template (.txt). Save a copy and fill it in on your device. No account, email address or photo upload is required. A worked example below shows how to keep a plausible name separate from a supported conclusion.

1. Record the photograph before separating or editing it

Photograph the album page, envelope, frame or bundle exactly as you found it. Then capture the full front and back of the photograph. Record its dimensions, mount, surface, condition and any text without correcting spelling.

Give the physical item a stable identifier such as `BOX01-PH-0042`. Use the same identifier for the front scan, back scan, research note and any share copy. A name should not be the identifier because the proposed identity may change.

If the photograph is stuck to a mount, flaking, moldy or unusually valuable, do not pull it apart. The US National Archives guidance for family archives explains basic handling and storage; a qualified conservator should assess damaged originals.

2. Start with provenance: whose collection was it?

Provenance means the history of where the photograph came from. An unidentified portrait from a named aunt's album has a narrower candidate group than the same portrait bought loose at an estate sale.

Record:

• who owned the box or album;
• which household or family branch it came from;
• whether nearby photographs are identified;
• the order of pages, envelopes and bundles;
• who gave the collection to its current owner;
• any earlier captions, including question marks.

Do not reorganize everything by your current theory. Original groupings can preserve relationships that are not obvious yet.

3. Estimate the photograph's date independently

Date the object before testing a person's name. Useful clues include:

• photographic format and mount;
• photographer or studio name;
• studio street address;
• tax stamps, postcards and printed designs;
• clothing, hair and uniforms;
• buildings, signs, vehicles and landscape;
• the age of the paper, ink or annotation.

A studio address can often be checked in digitized city directories. Clothing is useful for confirming a range, but ordinary clothes were worn for years and should not set an exact date alone.

Use at least two independent clues. Write the range and reasoning, for example: “Cabinet card; studio at this address 1888–1891; clothing consistent; probably 1888–1891.”

→ Use the complete method for dating an old family photograph

4. Build a candidate list from records, not resemblance

Once you have a likely date and place, list people who could realistically fit. Compare each candidate against:

• birth and death dates;
• residence at the estimated time;
• age range in the photograph;
• family relationships to other people pictured;
• military, school, occupation or wedding evidence;
• securely identified photographs from another source.

Eliminate impossible candidates first. A person who died before the studio opened, lived on another continent or would have been the wrong age should not remain the leading identification because the face looks familiar.

5. Compare groups and recurring details

One face can be ambiguous. A set of photographs may show the same person across several ages, the same brooch, uniform, porch or studio chair. Compare the whole collection rather than placing two cropped faces side by side.

Keep comparisons transparent. Note which photographs are securely identified and why. If the reference image itself comes from an uncertain family caption, it cannot independently confirm another image.

Automated face-comparison tools can sort a large archive into possible groups, but they should only generate candidates for human review. Image quality, aging, pose and family resemblance can all create false matches.

6. Ask relatives one photograph at a time

Do not send fifty unlabeled thumbnails with “Who is this?” Send one to five numbered images and ask concrete questions:

• Where might this have been taken?
• Whose home or album did you see it in?
• Could the people be from the Carter or Mills side?
• What detail makes you think that?
• Who else might recognize the place, uniform or event?

Record the reply in the relative's own words, along with their name and the date. Keep your interpretation in a separate field. If relatives disagree, preserve both accounts instead of silently choosing one.

7. Use explicit confidence labels

Use a small, consistent vocabulary:

Confirmed: supported by direct reliable identification or multiple independent records.

Probable: several clues agree and no strong contradiction is known.

Possible: a reasonable candidate, but evidence is incomplete or based mainly on resemblance.

Unknown: no candidate currently has enough support.

Write the evidence beside the label. “Probably May Carter” is less useful than “Probably May Carter: studio date, residence and estimated age agree; identification not independently confirmed.”

8. Preserve the evidence trail in the caption

A research caption can contain:

1. stable photograph ID;
2. identity and confidence level;
3. estimated date and place;
4. exact original writing;
5. source of the identification;
6. unresolved questions;
7. physical and digital storage locations.

Keep the untouched master scan. Make edits only on a separate access copy. Scan the back even when it appears blank: faint pencil, studio embossing or paper patterns may become useful later.

→ Scan both sides of old photographs at home

→ Name scanned family photo files consistently

Worked example: why “Aunt May?” is still a question

This is a fictional teaching example. The people, studio and records below are invented to demonstrate the method, not findings from a real family collection.

Imagine a mounted portrait in an aunt's album. Pencil on the back reads “Aunt May?” The front names a studio at 18 Market Street. In this example, directories list that studio at that address in 1890 and 1893; the years between and outside those entries have not yet been checked.

Separate observations from interpretations

• Observed: the original caption contains a question mark. Preserve it exactly.
• Observed in the example records: two directory entries place the studio at the printed address. They do not establish its opening or closing date.
• Interpretation: the person appears to be a young adult. Visual age is approximate and should not become a birth year.
• Working hypothesis: the portrait may date from around the early 1890s. A later reprint, reused mount or incomplete directory run could change that estimate.

Test two candidates without forcing a winner

Suppose the family tree contains May Carter, born in 1871, and her cousin Mary Carter, born in 1850. If the early-1890s estimate is right, May would be around 19–22 and Mary around 40–43. May fits the visual estimate better, but that does not identify her. The date estimate and apparent age could both be wrong.

A relative says, “I think it is May; that is what the album says.” Record the statement, but do not count it as a second independent identification. It repeats the same uncertain caption. Ask whether the relative has a separately identified photograph or remembers who wrote the note.

Write the conclusion and the next useful check

An honest research caption is: “BOX01-PH-0042. Unidentified young adult, possibly May Carter. Original reverse: ‘Aunt May?’ Working date around the early 1890s, based on two studio-directory entries; full operating dates not checked. Identification remains possible, not confirmed.”

The next task is specific: check a wider run of directories and locate a separately identified photograph of May. Keep Mary in the research notes until the evidence rules her out. Do not rename the master scan after May or post the tentative caption as an established fact.

Copy these fields into your own research note

The editable photo-identification notes template (.txt) contains the same fields. Use one note per photograph and keep it beside the front and back scans.

1. Item and location: stable ID, album page or box, original and scan locations.
2. Exact observations: front/back writing, studio/address, format and dimensions.
3. Source history: who supplied the item and which statements come from the same source.
4. Date hypothesis: range, evidence for it, alternatives and remaining uncertainty.
5. Candidates: name, supporting evidence, contradictions and untested assumptions.
6. Current conclusion: confirmed, probable, possible or unknown, with reasons.
7. Next check: one record, person or comparison that could change the conclusion.
8. Review and sharing: researcher, review date and permission before sharing private details.

A practical first session

Choose five photographs, not the whole box:

1. photograph their original container;
2. assign five stable IDs;
3. scan each front and back;
4. transcribe every mark exactly;
5. estimate a date range;
6. list no more than three candidates per image;
7. send two numbered images to one relative;
8. record their exact reply and your confidence level.

This creates a repeatable method. Even if none of the five people is identified immediately, the archive is safer and the next question is clearer.

What not to do

• Do not publish a living person's identity without permission.
• Do not present an AI face match as a confirmed identification.
• Do not remove a photograph from a mount if it resists.
• Do not overwrite the only scan.
• Do not erase question marks from inherited captions.
• Do not let a family story outrank conflicting physical evidence.

The Library of Congress guidance on caring for family treasures provides additional preservation resources. For valuable or damaged originals, use a qualified conservator rather than an online cleaning technique.

The standard to keep

A good identification is not simply a name under a face. It is a conclusion whose source, date, uncertainty and competing evidence remain visible. That is what allows the next generation to trust the answer—or correct it without losing the photograph's history.