All tools AI Image Checker ● Live Image / Inspection This tool provides heuristic inspection only. It cannot prove whether an image is AI-generated or human-made. How this tool works

AI Image Checker

AI Image Checker helps you inspect images for metadata clues, dimension patterns, file size anomalies, and other lightweight heuristic signals that may warrant closer review. It is an inspection assistant — not a definitive detector. All processing stays in your browser.

AI Image Checker Runs locally

Inspect an image locally

Upload a file to inspect basic metadata and a few lightweight heuristic signals. This does not provide proof.

Heuristic review only. Provenance, context, and visual inspection still matter.

Review summary

Upload an image to see dimensions, file size, and heuristic notes.

Signals to review

  • No notes yet.
Step by step

How to use it

  1. Select an image file from your device.

  2. Review the file format, dimensions, file size, and available metadata clues.

  3. Read the signal summary as a checklist of things to inspect, not as a verdict.

  4. Use the results alongside source context, reverse image search, and manual review.

When to use it

Use cases

  • Screen a social image before sharing it when the source is unclear.

  • Check whether an image has metadata or dimension patterns that deserve a closer look.

  • Teach a review team what lightweight provenance signals can and cannot show.

  • Compare several versions of an image without uploading private files to a detector service.

Limits & privacy

What this tool does and does not do

Private by design

The selected image is inspected locally in your browser. It is not uploaded to Small Web Apps servers.

  • Cannot prove whether an image is AI-generated or human-made.

  • Metadata can be stripped, edited, or missing on both real and generated images.

  • Dimension and file-size patterns are weak signals and can produce false positives or false negatives.

Background

What this can honestly tell you

This inspects an image for signals associated with AI generation and reports what it finds. It does not, and cannot, tell you definitively whether an image was generated. Treat every result as evidence to weigh rather than a verdict.

The strongest signals are metadata rather than pixels. Several generators write identifying information into the file, and some pipelines embed provenance data deliberately. When that is present it is fairly reliable — but it is trivially removed, so its absence proves nothing at all. An image stripped of metadata, or re-saved by any editor, looks identical to a camera photograph that went through the same process.

Visual heuristics are weaker still, and getting weaker as models improve. The artefacts that once gave generated images away are exactly what each generation of models is trained to eliminate, so a negative result increasingly means only that the generator was recent.

Both error directions are real. Heavily edited, filtered, or upscaled photographs can trip the same signals, and a careful generation can pass cleanly. If a decision matters — a claim, a dispute, a piece of journalism — this is a starting point for investigation, never the basis for a conclusion.

Questions

Frequently asked questions

Can this prove an image is AI-generated?

No. This tool highlights signals and inconsistencies that deserve closer inspection. It cannot prove provenance. Human review, reverse image search, and context are still necessary.

Which signals does it check for?

File format, dimensions (including common AI output sizes like 1:1 squares), file size relative to dimensions, EXIF metadata, and heuristic patterns. More signals will be added over time.

Why can't any tool reliably detect AI-generated images?

Modern generators increasingly mimic camera metadata, natural dimension ratios, and realistic compression artifacts. Reliable detection would require access to the generator's training signature or a watermarking standard neither side fully controls yet, so any browser-side heuristic is inherently limited.

Should I use this for content moderation decisions?

No. Treat it as a first-pass screening aid, not a moderation gate. False positives (real photos flagged) and false negatives (AI images that pass) are both expected — pair it with human judgment for anything consequential.

Does stripped metadata mean an image is fake?

No. Many platforms (Instagram, WhatsApp, Twitter) strip EXIF metadata from every photo on upload, real or not. Missing metadata is common and not itself a reliable signal of AI generation.

Can I check multiple images in a row?

Yes — select a new image at any time and the previous result is replaced. Each check is independent and none of the images are stored or compared against each other on a server.

Why does the tool avoid giving a confidence score or percentage?

A numeric score implies more certainty than heuristic signals can support. Instead, the tool lists the individual signals it found so you can judge their relevance yourself rather than anchoring on a potentially misleading number.