Manual image editing

Photoshop Detection

Check a photo for editing software tags, provenance clues, and visual evidence that the file may have been retouched or composited. FauxGuard gives you a confidence score and the reasons behind it without claiming that every edit is deceptive.

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.

What Photoshop detection can reveal

Photoshop detection asks a narrower question than AI image detection: does this file show evidence of manual retouching, compositing, or generative editing? A camera can produce the original pixels while software changes the subject, removes an object, replaces a background, smooths skin, adjusts a document, or combines several source images. The final file may still look like a normal photograph, but its editing history can leave clues.

FauxGuard begins with metadata that is already inside the uploaded file. A software tag may name Photoshop, Lightroom, GIMP, Affinity Photo, Pixelmator, Capture One, or another editing application. Camera make and model fields can show that a camera produced part of the source material, while a separate software field can show that the file passed through an editor later. Those two facts are not contradictory. Many legitimate photographers edit camera files every day.

The tool also reads generator names and C2PA-related text when available, without validating signatures or confirming attribution. It then requests separate AI-generation and face-manipulation scores from a third-party visual model. Neither score measures every kind of manual retouching. The reasons explain file clues and model signals separately rather than presenting a single probability that Photoshop was used.

Why a Photoshop tag is not proof of fraud

A Photoshop tag suggests an editing workflow; it does not verify that the application was used or that the image is false, because tags can be altered. Newsrooms crop and color-correct photographs. Product photographers remove dust. Families retouch portraits. A person may use the same application to make a harmless repair or to create a misleading composite.

That is why the result should be read as an investigation signal rather than a verdict. A software tag can support a question such as “Has this file been edited since it left the camera?” It cannot explain intent, identify the editor, or establish whether the final image accurately represents an event.

A missing software tag is equally limited. Messaging apps, social platforms, screenshot tools, and export settings can remove metadata. An image can be heavily edited and still show no application name. Conversely, metadata can survive after only a minor crop. Context and an original file are often more valuable than the tag alone.

  • Software metadata can identify a likely editing workflow.
  • Camera metadata can show that a camera file was part of the source material.
  • Visual model signals can add evidence when metadata has been removed.
  • No single signal proves intent, truth, or authorship.

Signals FauxGuard checks for editing

The first layer reads available EXIF fields, software names, generator names, and C2PA-related file text. The browser can resize and re-encode the upload, which may discard original metadata before server inspection. Preserve a source copy for independent metadata review. FauxGuard does not validate Content Credential signatures or SynthID watermarks, so these findings describe unverified clues in the processed file.

The second layer is visual analysis. The processed image is sent to a model that looks for patterns associated with AI generation and face manipulation. It can return low scores for a manually edited photo because ordinary retouching does not always look generative. A low score therefore means “the available model did not find strong generative evidence,” not “this image is untouched.”

The final result separates observations from inference. A software tag, a camera field, model scores, and reasons are reported together. Confidence labels reflect score ranges, not a validated accuracy rate. This helps a reviewer decide whether to request the original file, compare earlier versions, or ask the publisher for a source link. It does not locate a changed region or reconstruct an edit timeline.

How to review a photo that may be edited

Start by preserving the file you received. Do not re-save, screenshot, or send it through another messaging app before checking metadata, because those steps can remove useful fields. Record where the image came from, when it appeared, and whether the publisher provided an original file. If the image is part of a dispute, keep the message, listing, or post that supplied it.

Next, compare the image with other evidence. Search for earlier copies, ask whether the scene matches the claimed location and time, and inspect edges, shadows, reflections, text, and repeated patterns. A visually clean edit can still be exposed by inconsistent lighting or perspective. A strange-looking photo can also be a real camera image affected by motion blur, low light, or heavy compression.

Use FauxGuard as one part of that review. A result that says Photoshop was found can justify asking for the original. A result with a high generative score can justify a closer source check. A low score cannot be used to approve a claim without other verification. The strongest workflow combines file evidence, visual inspection, source history, and human judgment.

  • Preserve the original upload and record its source.
  • Compare the image with independent versions of the same scene.
  • Inspect the full-resolution file rather than a compressed thumbnail.
  • Give the other party a way to explain or appeal the finding.

Photoshop, generative fill, and AI editing

Modern editing tools blur the line between manual retouching and AI generation. Generative fill can add an object, remove a distraction, or extend a background inside a traditional editing application. The final file may contain editing software metadata but no obvious generator name. In other cases, the exported image contains no provenance at all.

FauxGuard handles this mixture by reporting available clues with their limitations. A readable generator name is shown as an unverified family clue. A Photoshop or other editor name suggests a possible workflow but does not identify a changed region. Visual scores are shown separately and can be inconclusive. No heat map, layer reconstruction, or verified editing history is supplied by this tool.

This distinction matters for buyers, moderators, journalists, and support teams. “Edited” is not the same as “synthetic,” and “synthetic” is not the same as “misleading.” A clean report should make the difference visible instead of collapsing every unusual file into one accusation.

When Photoshop detection is not enough

Some edits are deliberately designed to survive a basic metadata check. A person can export an image without metadata, use another application, print and photograph a screen, or create a new file through a screenshot. A detector can still help, but it cannot reconstruct every step in a chain that has been deliberately broken.

Legal, employment, insurance, and marketplace decisions should not rest on an automated score alone. These decisions affect people and require a documented process, a clear policy, a human reviewer, and an opportunity to respond. FauxGuard provides an informational signal. It does not establish who edited a file, whether a person acted with intent, or whether a statement is legally true.

If you need a more reliable conclusion, seek the original file, compare cryptographic hashes where available, request a chain of custody, or consult a qualified forensic examiner. FauxGuard is designed to make the first pass faster and more transparent, not to replace a formal investigation.

Was the background replaced, or just color-corrected?

These edits can have very different meanings, but a software tag cannot distinguish them. Cropping, exposure adjustment, background compositing, and generative fill may all pass through the same editor. A model score also cannot establish which operation occurred. Compare a source photograph with the delivered file when possible, keeping framing, lighting, and export quality in mind before attributing a visual difference to manipulation.

For a product listing or return photo, ask whether the image accurately supports the specific claim rather than whether any editing happened. A clean background may be ordinary presentation work; a changed product defect may be material to a dispute. Those judgments require contextual evidence and a fair review process. A detector score alone cannot make the distinction.

What to check after an editing result

Record which clues actually survived and which were absent. If the model returned a partial result, do not interpret the missing score as evidence that the photo was untouched. If a software tag was found, describe it as unverified file text instead of naming a person or alleging fraud. Then request an original or search for earlier versions to test the claim independently.

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 FauxGuard tell whether Photoshop was used?

It can show a readable Photoshop software tag when it survives in the processed file. That tag is not authenticated and can be changed. A missing tag does not establish that Photoshop was never used; browser conversion, export, or messaging can remove it.

Does a Photoshop result mean the image is fake?

No. Photoshop is used for cropping, color correction, retouching, compositing, and legitimate creative work. The result indicates a possible editing workflow, not the truthfulness or intent behind the image.

Can FauxGuard detect generative fill?

It may find readable generator text or model signals associated with generative editing. It does not authenticate Content Credentials or locate the filled region. A small generative edit can be missed, so a low or inconclusive score does not rule out generative fill.

Why does an edited photo sometimes receive a low AI score?

Manual edits do not always produce the patterns used by a generative-image model. A low score means the visible model evidence was weak, not that the image is guaranteed to be unedited.

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.