Journalism & Fact-Checking

Location is a core fact. A photo labeled 'protest in Capital X today' is a different story if it was taken in Capital Y last year. Mislocated imagery has driven false conflict reporting, misattributed disaster coverage, and manipulated election narratives.

Editors increasingly require verification notes alongside user-generated content. The geolocation step documents where the verification team established capture location and with what confidence.

Speed matters during breaking news, but accuracy matters more. Tools that return ranked hypotheses — not false-precision single pins — help reporters use qualified language until satellite confirmation completes.

Receive original file from source when possible — request camera-roll export, not WhatsApp forward. Compute hash and store in verification log.

Run client-side EXIF. If GPS present, plot and verify scene match before citing coordinates in copy. Note capture timestamp against claimed event time.

If EXIF absent, run reverse image search and visual clue pass in parallel. Use AI for regional hypotheses when clues suggest geography but not exact address.

Assign second reviewer for satellite or Street View confirmation before 'confirmed' language in publish. First reviewer documents eliminations and alternative hypotheses considered.

Build a verification note template your desk reuses on every UGC image: source contact, file hash, EXIF summary, reverse search results, AI predictions (if run), satellite confirmation screenshots, assigned confidence grade, and reviewer initials. Consistency beats ad-hoc Slack threads when legal review asks for methodology six months later.

Scenario: 90 minutes before polls close, a rival campaign shares a photo on X claiming it shows 'ballot stuffing at polling station in District 7.' Your fact-check desk receives the viral image — a screenshot, no original. Editors want a publish/don't-publish call before the nightly broadcast.

Minute 0–5 — Intake: Save the image, log the post URL and claimant account. Request original from the campaign (likely no response before deadline). Compute hash. Open verification tracker row.

Minute 5–8 — EXIF: Client-side pass returns nothing — screenshot. Document this. Do not infer manipulation from missing metadata alone.

Minute 8–18 — Reverse search: TinEye finds the same scene in a 2023 news article about a polling station in District 3 — a different city, 400 km away. Yandex confirms. You now have a provenance contradiction with the viral claim.

Minute 18–22 — AI hypotheses: Ranked predictions place District 3 city at 78%, District 7 at 6%. AI aligns with reverse search — useful internal signal, not publishable evidence alone.

Minute 22–40 — Visual verification: Crop visible polling station signage and compare to 2023 article photos. Match on facade color, window count, and adjacent pharmacy sign. Street View at District 3 coordinates confirms same building. The viral caption citing District 7 is wrong.

Minute 40–55 — Editorial decision: Confidence grade 'Debunked.' Draft copy: 'Image does not depict District 7. Reverse image search and visual comparison show the scene matches a polling station in District 3 photographed in 2023.' Do not claim AI 'proved' location — cite reverse search and Street View.

Minute 55–70 — Second review: Senior editor confirms side-by-side evidence. Legal clears language. Publish fact-check. Total: ~70 minutes from tip to publish.

Eyewitness originals may contain GPS at the photographer's home before they traveled to the event. EXIF read locally first — understand what metadata exists before any server upload.

Redact GPS from published downloadable files even when coordinates verified internally. Newsrooms publish cropped or metadata-stripped versions to audiences.

When AI analysis is necessary, inform sources if their image will be processed on external servers. Operational security for sensitive sources may restrict AI to open-source copies already public.

Source sends JPEG from iPhone claiming downtown demonstration. EXIF shows GPS on main protest square, timestamp 14:02 — 18 minutes before receipt. Scene shows crowd and municipal building facade.

Visual match: bollards and facade align with Street View at EXIF coordinates. Separate wire photo from agency confirms same square same hour.

Publish: 'Verified eyewitness photo from [Square Name] at approximately 14:00.' Do not publish raw EXIF coordinates in downloadable asset — strip for public file.

Alternate outcome if EXIF pointed elsewhere: treat GPS as hypothesis, verify visually before asserting location in copy.

Key lesson: EXIF GPS accelerates verification but never replaces visual confirmation. A mis-set phone clock or stale location cache can embed wrong coordinates. The Street View match is what earns 'confirmed' language, not the metadata alone.

Modern verification desks combine intake tools, reverse search, satellite platforms, and internal CMS fields. whereisthis.place handles the first two minutes — local EXIF and optional AI hypotheses — before analysts open Earth and TinEye.

Google Fact Check Tools, CrowdTangle (where available), and internal Slack bots each play coordination roles but do not geolocate. Train interns on which tool answers which question: metadata versus provenance versus ground truth.

whereisthis.place complements Google Earth, TinEye, Yandex, and internal verification Slack workflows — it does not replace them.

Export coordinates and notes into your CMS verification field or spreadsheet tracker. Link to full OSINT workflow guide for training new desk analysts.

Use the demo gallery interactive below to train interns on solved examples before assigning live breaking news verification.