What makes an online image trustworthy in the age of generative media?

A convincing image used to feel like strong evidence. Today, that is not always enough.

Generative media can create realistic scenes, remove objects, change backgrounds, or alter parts of a real photograph without making the result look obviously artificial.

That does not mean every unusual image should be treated as false. It means trust now depends on more than appearance. A reliable check considers where an image came from, how it has been used, whether its history can be traced, and whether other evidence supports the claim attached to it.

Trust starts with the source, not the pixels

The first question should not be, “Does this look fake?” It should be, “Where did this come from?” Visual clues can help, but source and context often tell you far more than closely examining hands, shadows, faces, or background details.

Find the earliest credible source

Start with the account, website, photographer, organisation, or news outlet that originally published the image. A screenshot that has been reposted several times usually provides far less useful context than an original upload with a clearly identified source.

Visual search tools can help trace an image as well. Google Lens and About this image, for example, can provide information about where similar versions of an image have appeared online, how other websites have used it, and when Google may first have encountered a similar version.

This can be particularly useful when an old photograph is being presented as evidence of a recent event.

Check the claim around the image

A real photograph can still mislead people if the caption attached to it is wrong. A flood image from one country might be reposted during a storm somewhere else. A genuine photograph taken years ago could be described as if it were captured this week.

Search important details from the caption separately. Look for reporting from reputable news organisations, official agencies, eyewitness material, or other independent sources. The credibility of an image depends not only on whether the pixels are genuine, but also on whether the surrounding claim is accurate.

Use technical signals, but do not treat them as proof

Technical information can reveal details that are not immediately visible on screen. The mistake is assuming that any one signal, whether metadata, provenance information, or automated detection, can provide a definitive answer by itself.

Look for provenance and Content Credentials

Traditional metadata may contain information about a file’s creator, device, software, or editing history. However, metadata can be changed or removed, so it should normally support other evidence rather than replace it.

Digital provenance can provide another layer of context. The Coalition for Content Provenance and Authenticity (C2PA) develops open technical standards for recording information about how digital content was created and modified. Its Content Credentials system can carry cryptographically signed, tamper-evident provenance information associated with an asset.

Importantly, provenance information does not automatically tell you whether an image is truthful. C2PA itself distinguishes between verifying provenance information and making a judgement about whether the content or its claims are “good” or “bad.”

What should you check?

• Is a creator or source identified?
• Is there information about how the image was produced or edited?
• Are Content Credentials or other provenance records available?
• Does the recorded history match the way the image is currently being presented?

These questions are especially useful when an image relates to breaking news, public figures, disasters, or other situations where misleading visual content can spread quickly.

Treat detection tools as an additional signal

An AI image checker can provide an additional signal when an image has no clear source or accessible provenance information. A detector result may give you another reason to investigate an image more closely, but it should not be treated as proof that an image is AI-generated or authentic.

Automated assessments can be affected by factors such as compression, resizing, editing, image style, and differences between generation systems. 

A useful approach is therefore to combine detector results with source tracing, contextual verification, provenance information, and independent evidence.

A practical five-step check before you share

Most people do not need specialist forensic skills to make a better judgement about an online image. A short process can catch many common problems.

Use this order

1. Identify the source. Find the original post or earliest credible publication you can locate.
2. Check the date and context. Confirm that the image relates to the event, place, or person described.
3. Search for earlier versions. Use Google Lens, About this image, or another visual search tool to see where the image has appeared before.
4. Review available provenance. Look for Content Credentials, creator information, metadata, or relevant platform labels.
5. Cross-check significant claims. For serious or consequential claims, look for independent reporting or official information that supports the same account.

If those signals conflict, don’t force a yes-or-no conclusion. “Unverified” may be the most accurate description until stronger evidence becomes available.

Conclusion: online image trust comes from several signals

A trustworthy online image is not simply one that looks realistic. Its credibility comes from the relationship between its source, context, history, provenance, and supporting evidence.

Generative media has made visual verification more complicated, but the basic approach remains manageable: trace the source, check the context, review available provenance, use automated detection only as an additional signal, and seek independent confirmation when the claim matters.