General image detection

AI Image Detector

Check a JPG, PNG, or WebP image for AI generation, editing, provenance, and manipulation signals. FauxGuard returns a confidence score and a clear evidence trail.

Still images only. One-time purchase, no subscription or daily free allowance. Save your private recovery code before paying.

$4.99 for 20 image checks

One-time purchase. No subscription. No account required.

Loading your balance…

Already purchased? Restore your checks

Drop a photo here

JPG, PNG, or WebP. Each completed model analysis uses one image check.

A broader view of image origin

An AI image detector can be used for more than asking whether a picture came from a text-to-image generator. Images are created and changed in many ways: a camera produces the original, a photo editor adjusts it, a generative tool adds or removes a region, or a model creates the whole scene. The useful question is not only “Was AI involved?” but “What kind of evidence exists about how this file was made?”

FauxGuard checks the processed JPG, PNG, or WebP file for readable camera fields, editing software tags, and generator names, then requests separate AI-generation and face-manipulation scores from a third-party model. File text is an unverified clue. The service does not verify Content Credential signatures or SynthID watermarks, and a generator name is not confirmed attribution.

This page is the general entry point for that process. More focused pages cover Photoshop detection, deepfake detection, and the direct question of whether an image is AI generated.

How the detector works

The first layer reads available file clues before the visual model is called. EXIF fields may mention a camera make or model, while software and generator names may suggest a workflow. These fields can be changed or copied. The browser resizes and may re-encode an upload before transmission, which can remove original metadata. Preserve the source file separately if its edit history matters.

The second layer sends the processed image to a third-party visual detection model. It evaluates AI-generation and face-manipulation signals separately. A successful analysis returns two scores, not one combined authenticity probability. Metadata alone is never converted into a numerical model score. If model analysis fails, a partial metadata result is shown without using an image check.

The result is not a final label. It is a structured summary designed to help a person decide what evidence to check next. A high score supports closer review. A low score supports caution rather than automatic approval.

  • Metadata and provenance pass
  • Generator family matching
  • Visual generative and deepfake model
  • Plain-language evidence and limitations

What makes a useful result

A useful detector result names what was found and what was not found. It should distinguish readable camera, editor, or generator text from a model inference. If C2PA-related text is found, FauxGuard reports that the signature is not verified. A single percentage without this context is difficult to audit and easy to misread.

A 90/100 AI-generation model score is not a validated 90% probability that a particular image is synthetic. High, moderate, and low confidence labels are based on score ranges, not an independently established accuracy rate. A middle score is inconclusive, while an unavailable score means the model did not provide a usable answer. These situations require different follow-up checks.

The explanation should use ordinary language: metadata survived, a known generator signature was found, the model saw only weak generative evidence, or the file was too processed for a reliable result. That makes the result easier for moderators, support teams, and everyday users to interpret.

Common use cases

Journalists and researchers can use an AI image detector as a first-pass tool before tracing the original source. Marketplace teams can review disputed product images more consistently. Community moderators can identify when an account photo or post deserves a closer look. Individuals can learn whether a suspicious image contains obvious provenance or generative signals.

The detector is also useful for comparing several versions of the same image. A camera file may contain metadata that disappears after a social upload. A generated image may carry a generator signature in one export and none in another. Seeing how the evidence changes across copies can help a reviewer understand which file is more informative.

The service should not be used to make automatic decisions about people or claims. It supplies evidence, not judgment. High-impact decisions need a documented policy, a human reviewer, and a route for correction.

  • First-pass content review
  • Source and provenance investigation
  • Marketplace dispute documentation
  • Media literacy and image education

Accuracy and limitations

No detector can prove image origin with perfect accuracy. Generators change, editing tools evolve, and social platforms remove file evidence. A model that performs well on one style may be less certain on another. A real image can be classified as unusual because of filters, noise, or compression.

This is why FauxGuard does not claim to identify the creator or intent. It also does not decide whether an image is truthful. A camera-origin file can show a staged scene, and a generated image can be used as obvious illustration rather than deception.

The best practice is to use the detector alongside source checks and human review. Ask for an original file when the decision matters. Compare independent copies. Document what was checked. Keep the result open to challenge because automated systems can be wrong.

Getting started

Use the upload area on this page and choose a JPG, PNG, or WebP file. The browser can reduce a large mobile photo before upload. FauxGuard reads available file clues and requests visual model analysis. Compression can discard original metadata, so keep an untouched source copy for any separate provenance investigation.

Read the score, the confidence band, the generator field, the provenance section, and the reasons. Do not stop at the first number. If the result is partial, read the notice to see whether the visual model was unavailable or whether the file simply did not contain enough evidence.

If you need a focused explanation, move to the related pages for editing, deepfake, or direct AI-generation questions. Each page links back to the same detector and explains how to interpret a different type of result.

Checking an AI image shared as a screenshot

A screenshot records what was displayed, not how the displayed picture was originally created. It can contain interface elements, cropped borders, and a second layer of compression. A low model score on that screenshot cannot establish that the original was a camera photo. Likewise, missing camera or generator metadata is expected and should not be presented as evidence of deception.

Ask for the underlying image rather than repeatedly checking smaller screenshots. Record the original post and its caption, then look for earlier appearances. If you analyze two versions, keep the results attached to the files that produced them; differences can reflect processing rather than a change in the underlying scene.

After reading the score: check the source

A detector asks about patterns in a file. Reverse image search asks where matching or similar pictures appear. When investigating a misleading caption, a source search may answer the important question even if the model result is uncertain. When a picture has no searchable matches, that absence does not establish AI generation. Use the two methods as complementary evidence, not competing verdicts.

Detector

Check an image with FauxGuard

Upload a JPG, PNG, or WebP image and review separate model scores, score-based confidence, readable file clues, and reasons. No account is required. A score is evidence to review, not proof of origin.

Related image checks

Frequently asked questions

What is the difference between an AI photo detector and an AI image detector?

The terms overlap. “AI photo detector” usually describes the consumer question of whether a photo came from AI. “AI image detector” is broader and can include generated images, AI-edited photos, deepfake signals, and provenance checks.

What file formats does FauxGuard accept?

The detector accepts JPG, PNG, and WebP files. The browser can resize a large image before upload to improve mobile performance.

Does the detector inspect metadata before using a visual model?

Yes. Readable file fields and generator names are checked before the visual model request. They remain unverified clues, not a numerical AI score. Browser conversion can remove original metadata; FauxGuard does not validate C2PA signatures or SynthID watermarks.

Can an AI image detector be wrong?

Yes. Models can misclassify unusual real images, and generators can produce images that avoid known patterns. Use the confidence band, reasons, and source context rather than treating the score as proof.

Do I need an account to check an image?

No account or login is required. Your image-check balance is accessed through a private recovery code and a browser cookie. Save the code before paying and keep it private: anyone with it can spend your checks. You can restore the balance from the detector or pricing page on another device.

How much does an image check pack cost?

FauxGuard sells 20 image checks for $4.99 USD as a one-time purchase. There is no subscription or daily free allowance. One completed model analysis uses one check. Failed model requests and metadata-only fallback results do not use a check. Review pricing and the terms before paying.

How should I read the detection result?

Read the AI-generation score and the separate face-manipulation score alongside the reasons and file clues. Confidence labels describe score ranges, not a validated accuracy rate. A high score is a reason for further review; a low score does not certify authenticity. An unavailable score is not a low score.

What are the limits of this image check?

False positives and false negatives are possible, especially after resizing, compression, retouching, or changes in image generators. FauxGuard does not verify C2PA signatures or SynthID watermarks, identify a person, or provide a local edit heat map. It accepts JPG, PNG, and WebP still images, not video or audio. Results are informational, not forensic or legal findings.