A practical verification guide

How to Tell If a Photo Is AI

Preserve the source, check the context, inspect the photo, and read model evidence with its limitations. Follow these steps instead of turning one unusual detail or score into a verdict.

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.

Step 1: preserve the photo and write down the claim

Before checking whether a photo is AI, identify what you actually need to establish. Is the question about a synthetic scene, a modified face, an altered background, or an incorrect caption? A real photograph can illustrate a false story, and a generated picture can be clearly labeled artwork. Defining the claim prevents a detector score from being used to answer a different question than the one that matters.

Keep the received file and record its source URL, publisher, caption, and the time you encountered it. Avoid resaving the only copy or taking a screenshot as a replacement for the original. These steps preserve useful context even if later analysis is inconclusive. Ask for a higher-quality source when appropriate, but do not request private documents or unnecessary identifying details merely to satisfy curiosity about an online image.

Step 2: look for earlier versions and a credible source

Investigate whether the same scene appeared before the claimed event. A source search can lead to earlier captions, alternate crops, or a creator's explanation. Open the matching pages and compare the images rather than relying on search thumbnails. A similar mountain, room, or face is not automatically the same photograph. The earliest copy you find is a useful lead, not necessarily the first publication anywhere.

If there are no matches, keep the origin unresolved. Private, new, or altered files may have no searchable counterpart. If there are many matches, do not treat repetition as independent confirmation: several pages may be copying one source. For a news claim, seek independent coverage; for a product dispute, compare the image with the actual order and correspondence. Source research checks the story around the pixels, something an AI-generation score cannot do.

Step 3: inspect visible details without using an artifact checklist as proof

Look at lighting, perspective, repeated objects, reflections, and boundaries between a subject and its background. Write observations precisely: an inconsistent reflection is something to investigate, not proof of a specific generator. Many ordinary camera effects can look unusual, including motion blur, lens distortion, low-light noise, shallow focus, and compression. A plausible hand or readable sign also does not establish that a photo came from a camera.

Compare any suspicious region with another version of the scene if one exists. Do not repeatedly enhance, sharpen, or crop a file and then assume the resulting differences were present in the source. A visual review is most useful when it produces a clear follow-up question, such as whether an object existed in an earlier version, rather than a verdict based on a collection of strange-looking details.

Step 4: upload once and read the actual model evidence

Use the detector on this page for a JPG, PNG, or WebP still image. The browser reduces large images to a maximum edge of 2048 pixels and may convert the file before upload. FauxGuard reads available file clues and sends the processed image to a third-party model. Because conversion can remove original metadata, preserve your source separately rather than expecting the result to reconstruct its entire history.

Read the AI-generation score separately from the face-manipulation score. A high AI-generation score supports closer review; a low score does not certify a real or untouched photo. Confidence bands reflect score ranges, not a measured accuracy guarantee. A generator name or software tag is unverified text, and C2PA-related markers are not authenticated signatures. If the model is unavailable, the partial notice means there is no usable model answer, not that the image passed.

Step 5: decide what remains unknown before sharing a conclusion

Summarize the evidence in separate categories: what the source says, what you observed, what the model returned, and what remains unknown. Avoid translating a model score into a statement that a named person lied or committed fraud. Scores cannot identify the creator, establish consent, prove a date or location, or determine whether an image accurately depicts an event. Important decisions need independent evidence and a human review path.

You may find that the next useful action is asking for the original, comparing another source, or leaving the claim unresolved. That is a legitimate result of verification. If a model check would help address an unresolved origin question, return to the upload area above and scan with FauxGuard. Review the one-time pack and recovery instructions first. Repeating the same scan does not turn uncertain evidence into a guarantee, and this tool is not a substitute for a qualified forensic examination.

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

Can I tell whether a photo is AI just by looking at hands or text?

Not reliably. Visual inconsistencies can suggest follow-up questions, but real photos can have artifacts and synthetic images can look consistent. Combine source context, comparisons, and model evidence instead of treating any single visual detail as a conclusive test.

Should I scan a screenshot or ask for the original photo?

An original or less-processed source is generally more informative for a review. You can scan a supported screenshot, but its result describes that file. Cropping, resizing, and compression may change the evidence. Keep the original separately because FauxGuard's browser conversion can also remove metadata.

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.