wyreup
inspect

Scan Image for Suspicious Text

Reads the text in an image, then scans it for prompt-injection phrases, invisible characters, confusable glyphs, and mixed scripts. Run it before feeding a screenshot or photo to an AI.

Loading…

About Scan Image for Suspicious Text

Scan Image for Suspicious Text reads the words in an image and then inspects them for prompt-injection phrases, invisible characters, confusable glyphs, and mixed scripts. Run it before you paste a screenshot or photo into an AI assistant, so a hidden instruction in the image cannot quietly hijack the model. The whole scan happens in your browser — nothing about the image is uploaded.

Category
inspect
Input
Accepts: image/png, image/jpeg or image/webp.
Output
Outputs: application/json (multiple).
Cost
Free, runs in your browser
Memory
medium
Privacy: Scan Image for Suspicious Text runs entirely on your device. Files you provide never leave your browser — no uploads, no server, no tracking. The page works offline once loaded.

Common uses

  • Vet a screenshot from an untrusted source before feeding it to an AI chatbot
  • Catch invisible zero-width characters smuggled into a copied block of text in an image
  • Detect confusable look-alike letters (Cyrillic 'a' posing as Latin 'a') in a logo or banner
  • Flag mixed-script text that may be disguising a malicious URL or command
  • Spot prompt-injection phrases like "ignore previous instructions" hidden in a meme
  • Screen user-submitted images in a moderation workflow for adversarial text

Frequently asked questions

What threats does it look for?

Prompt-injection phrasing, invisible/zero-width characters, confusable look-alike glyphs, and unexpected mixed scripts in the recognized text.

Does the image leave my device?

No. Both the text recognition and the suspicious-pattern scan run locally in your browser.

Which image formats are supported?

PNG, JPEG, and WebP.

What does the output look like?

A JSON report with the extracted text and the specific suspicious findings it flagged.

Does it guarantee an image is safe?

No tool can. It surfaces known adversarial-text signals, but you should still use judgment with images from untrusted sources.

Keywords

  • ocr
  • suspicious
  • prompt injection
  • security
  • image
  • screenshot
  • scan

Try next

OCR
export

Just need the plain text without the security scan? Use straight OCR.
HTML Redact
privacy

Found PII in the scanned text? Redact it from an HTML version.
Describe Image
export

Get a model-written description of what the image actually depicts.
Color Palette
inspect

Extract dominant colors from an image as hex codes.
Image Diff
inspect

Pixel-level diff between two images of the same dimensions.
Image Info
inspect

Extract dimensions, format, and metadata from an image.
PDF Info
inspect

Extract page count and metadata from a PDF.
Hash
inspect

Compute SHA-256, SHA-1, or SHA-512 hashes of files.
JSON Formatter
inspect

Parse and pretty-print JSON with configurable indentation.
Extract PDF Tables
export

Extract tabular data from a PDF as JSON or CSV. Works well for simple tables with aligned columns and no merged cells. Complex tables, rotated text, and scanned PDFs (images) will not extract cleanly.
Topic Keywords
text

Pull the topic-bearing nouns and noun phrases out of any text. Answers "what is this about" — not "what words appear most" (try Text Frequency for that). Runs entirely in your browser.