To work out what’s real online, verify the underlying claim, investigate the media file’s origin and edits, and check whether its caption gets the time, place, and context right. Those are separate questions: provenance credentials can document aspects of a file’s history, and AI detectors can offer clues, but neither proves that a claim is true.
What does “real” mean?
An image or video can be genuine yet misleading. It may show a real event but be old, staged, selectively cropped, or shared with a false caption. A convincing-looking file, meanwhile, does not establish that the event happened as described.
As an Amazon Associate I earn from qualifying purchases.
- Is the claim true? Look for the original source and independent corroboration.
- What is known about the file? Check whether information about its creator, source, or edits is available.
- Does the caption fit? Verify when and where the scene happened and what was taking place.
NIST’s 2024 survey, updated in 2026, describes a range of approaches—including provenance tracking, labeling, detection, testing, prevention, and auditing—rather than one method that answers every question: NIST’s overview of synthetic-content approaches.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →How can I verify something I saw online?
- Preserve the original claim. Save the post’s wording, account, date, link, and media in the highest-quality version available. Record what the poster actually asserted instead of silently rewriting it.
- Find the primary source. Search for the named person, document, organization, event, or statement. Prefer the original record when available, then see whether independent secondary sources confirm the same details.
- Check the media’s history and context. Search for earlier appearances of the image or footage. Independently verify the date and location, and look for signs it was cropped, edited, staged, or reused from another event.
- Inspect provenance information. If Content Credentials or other provenance details are present, note exactly what they say about the file’s source or edits. Check whether the verifying tool or service is conformant.
- Use a detector only as a lead. Check what media type it supports, what its published method and training scope cover, when it was updated, and whether it explains uncertainty.
- Report the limits of the conclusion. Separate what you verified from what remains uncertain, and identify what evidence could change your assessment.
This workflow draws on NIST’s survey, the Reuters Institute’s discussion of miscontextualized footage and detector limitations, and C2PA’s provenance resources: Reuters Institute on detecting AI-generated images, C2PA resources.
#1 Best Overall
- Keep track of everything from attendance to test scores
- Spiral bound
- Measures 8-1/2" x 11"
What do Content Credentials tell you?
C2PA describes Content Credentials as tamper-evident, machine-readable labels that can help audiences discern when media was created or modified by generative AI. Its July 2026 resources also explain its Conformance Program and how organizations can choose, verify, and display conformant tools: C2PA Content Credentials resources.
A credential is evidence about the file’s recorded provenance and edits; it does not prove that the caption is true or that the scene was not staged. Missing credentials do not mean a file is fake: credentials may be absent, stripped as media moves through platforms, or unavailable because adoption is incomplete. Reuters Institute’s 2026 trends report discusses the wider provenance and verification context: Reuters Institute Digital News Report 2026.
Can an AI detector tell whether a photo or video is real?
Not conclusively. Detector outputs may be probabilistic or binary, and they are hard to interpret without knowing the model, training data, target media, and update date. A tool built to detect AI images may not recognize a face-swapped video; noise can interfere with audio detection; unfamiliar subjects and blurred or compressed images can make detection harder. Reuters Institute’s explainer also describes generated or edited images assessed as likely human or not likely AI-generated: Reuters Institute’s detector explainer.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- A high score from an opaque tool is not proof that media is AI-generated.
- A negative result does not establish authenticity.
- Odd-looking details alone are not enough to label an image or video synthetic.
- Even authentic media can be staged or miscaptioned.
There is a related risk: plausible deniability. People may dismiss real evidence by claiming it is a deepfake, so “AI-generated” should not become a reflexive answer to material someone disputes.
When should you ask a person to check?
When a claim could cause serious harm or has legal, safety, or public consequences, seek corroboration from primary sources, subject-matter experts, or experienced fact-checkers. Automated systems can help find and classify material, but they should not be treated as independent arbiters of complex claims.
Reuters Institute’s report on its March 2026 conference describes both the challenge and the potential: misleading content can be produced quickly with generative AI, while tools can help small teams locate and classify claims. It reports that Maldita and Full Fact have developed systems to detect and classify claims across millions of sentences. It also says Aos Fatos used an audience-answering chatbot and was developing a live-newsroom fact-checking tool at the time of the report. These are examples of newsroom support, not proof that automation alone can settle a claim: Reuters Institute’s 2026 conference report.
In that report, Reuters Institute said 16% of the 619 claims Aos Fatos fact-checked in 2025 involved AI-generated content, compared with 7% in the previous year. Those figures describe one fact-checking organization’s workload, not the share of all online claims. The report also quotes Full Fact CEO Chris Morris warning that people may stop trusting what they read, see, or hear, while arguing that technology can help address the problem at scale.
The same report quotes Aos Fatos founder Tai Nalon saying that, in Brazil, AI-generated fast content reached over 32 million TikTok views and AI-powered disinformation related to 2.1 million interactions, including likes and shares, on Facebook and Instagram. Those are figures attributed to Nalon in the report, not a general measure of misinformation exposure.
Best Value
How do you build better verification habits?
Media and information literacy means engaging critically with information, navigating online safely, and building trust in information and technology. UNESCO’s resources include curricula, handbooks, publications, and digital-competency materials: UNESCO media and information literacy and UNESCO learning resources.
UNESCO’s resource page gives several figures, each measuring a different population or behavior: it says two-thirds of digital content creators do not systematically fact-check information before sharing (UNESCO, 2024); 85% of citizens are worried about online disinformation (UNESCO/IPSOS, 2023); 56% of internet users frequently use social media to stay informed about current events (UNESCO/IPSOS, 2023); and 80% of young people use AI tools and services multiple times a day for education (UNESCO, 2024). They should not be combined into a single trend line.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
