A photograph can be real and still tell the wrong story. A flood image from five years ago, placed beside a new disaster claim, is not fake in the usual sense. It is false in the detail that matters.
That is why news photo verification starts with the caption, not the pixels. You are checking a package of claims: what happened, where, when, who is pictured, and whether the image was edited or cropped to change its meaning.
Slow down before the photo hardens into proof. A few careful checks can separate a useful lead from a published error.
News photo verification starts before image search
Treat the photograph and the caption as separate pieces of evidence. A caption can be accurate when the photo is a general illustration. A photo can be authentic when the caption borrows the wrong date, town, or event. The match has to hold on both sides.
Read the claim word for word
Copy the caption exactly. Save the headline, image credit, byline, publishing time, account name, and URL. If a post says "today," record its calendar date. If it says "after the attack," identify which attack. Broad language can hide a mismatch.
Then split the wording into checkable claims. "Residents flee City X after Tuesday's earthquake" makes claims about place, time, event, and the people shown. Each part can be right or wrong. Don't let a dramatic frame make the rest feel settled.
Preserve what you saw
Save a screenshot that includes the caption, account, and any visible time stamp. Download the highest-quality file available, rather than relying only on a compressed screen capture. Note when you accessed it. Posts vanish, articles update, and captions can change without preserving the earlier wording.
When we review a photo, we keep a short record of the post URL, image URL, searches used, search date, and results that matter. This is not paperwork for its own sake. It gives another reporter or editor a path they can retrace.
Search the image, then search its details
Reverse-image search is a strong first test. It is not a verdict. Search engines can miss an image, rank a later repost above the original, or find a visually similar photograph with no direct connection.
Run more than one image search
Start with the whole frame in Google Lens. Then try TinEye and another engine, because their indexes differ. The University of Tampa's fact-checking guide includes Google Lens among practical checks for online claims.
Next, crop the parts that leave a trail: a sign, storefront, flag, road marker, uniform, license plate, skyline, or distinctive building. A full-frame search may fail when text overlays, a new crop, or a color change masks the original.
Compare dates, credits, and prior captions
Open the results, not only the preview page. Look for the earliest credible appearance, then read its caption, credit line, and publication date. Check whether it came from a news agency, photographer, government source, local outlet, or an unknown account copying other work.
Finding the same image does not prove the caption is accurate, current, or unedited. It proves only that the image, or a close version, appeared somewhere else. News photo verification requires the image's history and the new claim to agree.
If the caption is already circulating as a public claim, search Google's Fact Check Tools for published checks. A matching headline is a lead, not a final answer. Read the reporting behind it.
Test the place, time, and event
Search results give you a lead. The remaining work is to test the caption's place, time, and event against details in the frame.
Check the visible context
Check signs and languages, road lines, transit logos, building shapes, terrain, weather, shadows, seasonal vegetation, and clothing. Compare them with reliable local reporting, official notices, maps, and dated photos of the claimed location.
One matching palm tree proves little. Several independent details can narrow the answer.
For a claimed breaking event, look for reporting from more than one independent news organization. Compare the scene with agency photographs and official statements, but don't treat an official post as automatic proof. It may repeat the same unverified image.
Treat metadata as supporting evidence
Time often breaks the claim. Ask whether the light, weather, season, damage, vehicles, and public activity fit the stated day. A 2024 wildfire photo can illustrate an article in 2026, but it cannot document a 2026 fire without clear labeling.
Inspect metadata if you have the original file. GPS coordinates, camera model, and a creation date can help. They can also be absent, stripped by a platform, or altered. We treat metadata as one clue, never the final answer. The Data Journalism Handbook's verification guidance draws the same distinction: confirming material is not the same as accepting a claim.
A photo may establish that a scene existed. It does not establish the caption's date, location, cause, or identity without supporting evidence.
Handle images of death, injury, children, private homes, and vulnerable people with added care. Preserve a copy for the record, but don't re-share a graphic frame only to ask strangers to identify it. Crop a non-sensitive detail for a search when that can answer the question. Seek permission and reduce identifying details before publication.
The caption needs its own evidence
Fast checks catch some false captions. Careful checking catches the more common problem, a real image attached to the wrong moment. Preserve the original, search more than once, compare the results, and test each detail in the caption.
Where the record remains incomplete, say so. We would rather publish a limited conclusion that others can check than a smooth answer built on one search result. A photo is evidence, not a conclusion.