How to verify a photo before you share or publish it

To verify a photo before you share or publish it, check two things separately: where the picture came from, and whether the caption's claims (event, place, date, people, source) match what that original source says. A real photograph with a wrong caption is the most common kind of misleading image, so a reverse image search match alone does not verify the story.

The checks below take five minutes for a casual reshare and a little longer for publication. They work for news photos, viral posts, newsletter and slide images, client assets and community reports.

The five-minute check before you share

1. Open the post, not the preview. Read the full caption and note exactly what it claims. Save the page URL and today's date.
2. Reverse search the image. Use Google Lens and TinEye at minimum. In TinEye, sort by Oldest. If the image is a screenshot, crop away the app interface first.
3. Look for a copy that predates the claimed event. One credible earlier publication of the same image is enough to show the caption is at least partly wrong.
4. Read the earliest credible source in full. Its caption, article and credit tell you the original context.
5. Compare claim by claim. Mark each part of the caption as supported, contradicted or unverified.
6. Decide. Share with the original context, share with a clear qualification, or do not share.

If you cannot find the source, that is a result too: "unverified" is a good reason to hold off on a dramatic image. The full tracing workflow, including dates, watermarks and credits, is in how to find the original source of an image.

Split the caption into claims you can check

A caption usually bundles several claims. Write each one down and check it on its own.

A genuine photograph of a 2019 flood shared as "this morning's flood" is a real image with a false date and place. Finding the 2019 source proves reuse; it does not prove there was no flood this morning. Report the specific mismatch rather than calling the whole story fake.

Is it an old photo presented as new?

This is the most frequent problem with viral images during breaking news, storms, conflicts and protests.

1. Search the full image and a crop of the main subject. Reposts are often cropped, flipped or filtered, and a crop can match the original better.
2. Open candidates that predate the claimed event. Ignore results that only repeat the new claim.
3. Confirm it is the same photograph. Compare object placement, people's positions, cloud shapes, signage and edges. A recurring festival, a city skyline or a regular storm can produce near-identical scenes in different years.
4. Check whether the earlier page labels it as an archive or file photo. That page may itself be reusing an older image.
5. Check the old page's real date. Use the Wayback Machine to see whether the page and image existed at an earlier capture date; a search-result date may only be an update or crawl date.
6. Check whether fact-checkers already covered it. Search the claim in Google Fact Check Explorer, which collects published fact-checks.

Visual clues can support or weaken a date: season and foliage, weather, shop signs, vehicle models, election posters, construction that has since finished. Treat them as supporting evidence. They rarely settle a date on their own.

For clips rather than stills, grab a few clean frames and search those; see reverse searching a video frame.

What a reverse image search cannot prove

An image match supports a relationship between a picture and a page. It does not, by itself, answer these five questions:

1. Who created the photograph. The first site you find may be a repost.
2. Who or what it shows. A face or building that looks similar is not an identification.
3. Who runs the account using it. Stock and stolen photos appear on many unrelated accounts.
4. What permission exists. Finding a picture online is not a licence to reuse it.
5. Whether the caption is true. Only the original context and independent reporting can test that.

Before you act, write the conclusion you are tempted to draw, then underline the part your evidence actually supports. The gap between the two is what still needs checking: a creator's record for attribution, a licence for permission, an official channel for an account's authority.

When the image may be edited or AI-generated

Context checks assume the picture is a real photograph. If it might be composited, retouched or generated, check that too: look for inconsistent shadows and reflections, garbled text, odd hands and edges, and repeated textures, and check for Content Credentials with Content Credentials Verify. Missing credentials prove nothing. The AI image and deepfake checks guide covers this in depth. A genuine photo can still be cropped to remove context, so compare against the widest original you can find.

Workflows by role

Blog, news and newsletter editors

Give every image a job before it goes in: documents the event, illustrates a concept, or depicts a named person or place. Then:

1. Trace the image to a credible source and record the URL.
2. Confirm the caption's date, place and names against that source.
3. Check the licence and the credit wording it requires.
4. Label illustrative and file images as such ("File photo", "Illustrative image").
5. Re-read the final layout: a headline or logo beside a generic photo can make it look like evidence about a specific company.

A generic office photo can illustrate a story about remote work. It should not be captioned as a particular company's team.

For images that arrive inside a newsletter or email, follow the visible credit or article link before searching. Images hosted on an email platform's content server were not made by that platform. Save only the image and the relevant caption; do not upload whole emails with subscriber data to a search service.

Designers receiving client assets

Keep an intake record for each asset: filename, where it will be used, source, credit and usage limits. For files with an unknown source, search for the likely original, compare with any stock asset ID the client supplied, and ask the client to confirm their permission in writing. If the source stays unresolved, use a replacement asset.

A client may send an image taken from a competitor's site, believing public means reusable. Finding that source is the easiest way to explain why a licence or replacement is needed. Never mark an asset "cleared" because a reverse search found no copies; an empty result is not a licence. Creative Commons files must follow the specific licence terms, including attribution.

Community moderators

Start from the rule and the report, not from the person.

1. Name the alleged issue: reused event footage, a misleading caption, or someone posting another member's photos as their own.
2. Search only the material in the report.
3. Record the source pages and dates you relied on.
4. Compare the evidence with the specific community rule.
5. Explain the decision without posting unrelated personal information.

A reposted photo may be fine with a credit and against the rules when presented as a member's own recent experience. If a match is a lookalike or an inaccessible preview, do not call it confirmed impersonation. Use the platform's escalation process, and keep short evidence notes so moderators decide similar cases the same way.

Students and researchers

Record the source when you choose the image, not the night before the deadline. Keep: title or description, creator, publisher (separately from creator), source-page URL, access date and licence. A search results page is not a source; click through to the page that publishes and explains the image. Prefer museums, archives, creators' sites and clearly licensed collections such as Wikimedia Commons, and read the item's own reuse note.

If you already have an uncredited file, trace it with image search and caption clues. Do not cite the first matching site as the creator. If the origin stays uncertain, pick a different image with a clear record.

Images in presentations and reports

Treat slide images as clues stripped of their context. Check the slide's citation first, then search the image or any caption phrase, open the original publication and compare its description with how the slide uses it. Crop out confidential notes, names and charts before uploading a slide image anywhere. If the reference is missing, ask the presenter rather than inventing an attribution from a similar result.

How to write up the result

Use wording that matches your evidence:

Link to the source when you share a correction, explain the specific mismatch, and do not replace one unsupported story with another. Do not name private people, guess motives or supply a location that the evidence does not support.

Correcting an image after publication

If you published the image and later find the context or credit was wrong:

1. Confirm the new source on the creator's record or another primary publication. Do not swap one uncertain credit for another.
2. Fix the visible credit and link.
3. Fix the caption, not just the credit. If a file photo was presented as current, the date is the bigger error.
4. Update everything you control: social preview images, newsletter archives, downloadable PDFs and syndicated copies.
5. Add a dated correction note when the change affects what the image appeared to show.
6. Keep the evidence in your editorial record.

Our own approach to sourcing and corrections is described in the editorial policy.

When the photo shows a person

Most caption checks never need to identify anyone. When a photo of a person is the claim, for example a profile picture that may belong to someone else or your own photo being used under another name, a face search is the right tool rather than a copy search. Facehunch runs a pay-per-search face search ($4.99, no subscription) that lists each match with its source link and similarity score, so you can check the pages yourself. Use it to protect yourself or check someone you are already dealing with, never to name a stranger in a viral photo. For copy-oriented checks on any other picture, start with a reverse image search.

Written by Facehunch with AI assistance and checked against the cited sources. Our editorial policy.