How to Tell If a Political Ad Image Is AI-Generated

Updated 5 November 2026 · Standard formula, shown below

You can assess whether a political ad image may have been generated or altered with AI by checking its source, comparing it with other versions, and looking for visual and technical clues. These checks can help you judge how much confidence to place in an image, but no single clue—or detector—can prove who made it.

What You’ll Need

The image or ad, preferably at its original resolution rather than as a screenshot.
A web browser and access to image search or reverse-image search.
A way to zoom in or crop details without heavily recompressing the image.
Optional: an AI-image detection tool and access to the ad’s original post, campaign page, or publisher.

Step-by-Step

Save the image and record where you found it. Note the account, website, date, caption, and any link or disclosure shown with the ad. Save the image itself if possible. A screenshot can preserve context, but it may reduce image quality or remove metadata, so keep the original file too. Do not treat a repost as the source simply because it appeared first in your feed.
Find the earliest available version. Search a distinctive phrase from the ad, the candidate’s name, or a slogan. Use reverse-image search with the full image, then try a crop of its main subject. Compare dates and publishers. An earlier, reputable source may identify the photographer, publication, event, or original image. Search results are incomplete, so failure to find an older copy does not establish that the image is AI-made.
Check the campaign or publisher’s own materials. Look for the same ad on an official campaign site or account, and read its caption or disclosure. See whether the image is described as an illustration, parody, composite, or AI-generated. If it is an attack ad shared by another group, check that group’s original post as well. A familiar logo or candidate likeness does not, by itself, authenticate the source.
Inspect the image closely for inconsistencies. Zoom in on hands, ears, jewelry, glasses, hair, signs, lettering, and objects partly hidden by other objects. Look for warped text, duplicated details, strange reflections, shadows that point in conflicting directions, or edges that blend unnaturally. Also check whether perspective, lighting, and background details agree. These are clues, not proof: ordinary editing, motion blur, small resolution, and compression can create similar defects.
Ask whether the scene makes sense. Check the setting, clothing, weather, event details, and visible signage against the claimed time and place. Search for independent photos or video from the event. An image may be a real photograph used with a false caption, a real photo that has been edited, or a synthetic scene. Those possibilities require different checks; suspicious context alone does not show that AI created the pixels.
Check file details and provenance, if available. On a computer, view the file’s properties or use a reputable metadata viewer. A camera model, creation time, or editing-software entry may offer a lead, but metadata can be stripped, changed, or added. Some content credentials can record information about how an image was created or edited. If present, check their details and validation status; their absence is not evidence of AI generation.
Use AI detectors only as one supporting check. If you try a detector, use a clear, unaltered file and read what the result actually claims. Different tools may disagree, and their scores are not reliable proof of origin. Cropping, resizing, screenshots, compression, and editing can affect results. Do not upload sensitive or unpublished material to an unfamiliar service without considering its privacy terms.
Reach a cautious conclusion and explain it precisely. Separate what you know from what you suspect: for example, “The source could not be verified, and several details look inconsistent” is stronger than calling an image fake without evidence. Link to the original post or reliable reporting, preserve relevant copies, and avoid sharing an unverified image as fact. If the ad is consequential, seek confirmation from the named source or an independent newsroom or fact-checking organization.

Common Mistakes

Treating one visual flaw as a verdict. AI can produce convincing images, while real photographs can contain odd-looking details. Look for multiple independent clues.
Trusting a detector’s percentage as a probability. A score may not mean “this image is that percent AI.” Read the tool’s explanation and limitations.
Assuming metadata settles the question. It can be missing or altered, and an editing record does not necessarily mean AI was used.
Confusing a false caption with a fabricated image. A genuine photo can be recycled from another event or given a misleading description.
Judging a low-quality copy as if it were the original. Social platforms and messaging apps often resize and compress images, obscuring details and creating artifacts.

Tips

Keep an untouched copy before cropping or annotating. Search both the full image and a few useful crops.
Compare claims with independent sources, not several accounts that all repeat the same post.
Check the date and time zone of posts when establishing which version appeared first.
Describe the evidence and uncertainty when sharing findings. “Possibly AI-generated” is not the same as confirmed.
Consider whether the image was partly AI-assisted or edited; “AI or not” is not always a simple choice.

Troubleshooting

Reverse-image search finds nothing: Try a crop containing a face, distinctive sign, or background landmark. Search a slogan separately. The image may be new, private, altered, or absent from the search engine’s index.

The image looks blurry or full of artifacts: Find a higher-resolution copy from the original post or publisher. Avoid enlarging a tiny screenshot and treating the resulting jagged edges as evidence.

Metadata is missing: This is common for images shared online. Continue with source tracing and corroboration; do not infer AI use from missing metadata.

Detectors disagree: Record the disagreement as uncertainty and prioritize verifiable provenance and corroborating evidence. Do not average scores or select only the result that supports your suspicion.

The source denies making the image: Preserve the denial and the image’s link, then look for independent evidence. Contact the account or publisher through a known public channel if clarification matters; avoid assuming that an impersonating account is genuine.

Frequently Asked Questions

Can I tell for certain just by looking?
Usually not. Visual inspection can identify clues, but convincing AI images and edited photographs may be difficult to distinguish without source evidence.

Are AI-image detectors accurate?
They can sometimes provide a useful lead, but results may be wrong, especially on compressed, cropped, or edited images. Treat them as supporting evidence only.

Does missing camera metadata mean an image is AI-generated?
No. Platforms commonly remove metadata, and screenshots may never contain the original camera details.

What if the picture is real but the political claim is false?
Check the original event, date, location, and caption. A genuine image can be presented out of context or altered without being AI-generated.

Can an AI-generated image include real people?
Yes. A synthetic or edited image can depict recognizable people. A likeness alone does not verify that the person was present or did what the ad claims.

What is the strongest evidence of AI generation?
A clear disclosure or verifiable creation record can be strong evidence. Otherwise, a combination of reliable source tracing, corroboration, and technical clues is more useful than any single test.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.