# AI Detector for Photos and Screenshots of Text

> To check a photo of text for AI, import the photo, let text recognition turn it into words, correct any misreads, then run the detector. Humanit scores the wording 0–100. It does not judge whether an image was AI-generated.

**URL:** https://humanit.app/detect/photo-of-text  ·  **Updated:** October 2026  ·  Humanit

Take a photo of a printed or handwritten page, or a screenshot of a post, and check whether the words read as AI-written. This checks the text in the picture, not whether the picture itself was AI-generated.

## Quick facts

- **Files:** JPG, PNG and other photos, screenshots, or several photos of one document.
- **Are photos stored?:** Each photo is sent once to be read, then discarded. It is never stored.
- **Words per check:** 300 words a check without an account, 1,500 with a free one, up to 6,000 on paid plans.
- **Photos per day:** Read with OCR on a free account: 10 pages a day, 100 on paid plans.
- **File limits:** 20 MB per file, up to 30 pages or photos at once.
- **Cost:** Free. The detector never spends words, with or without an account.
- **Not for:** Telling whether an image itself was AI-generated.

## How it works

1. **Sign in free** — Reading photos needs a free account: 10 photos a day free, 100 on paid plans.
2. **Import the photo or photos** — Drop up to 30 photos of one document at once; they are read as its pages, one after another.
3. **Fix what was misread** — Check names, numbers and anything handwritten before scoring. A misread word changes the score.
4. **Run Detect AI** — Up to 1,500 words a check on a free account. Very short passages give unreliable scores.

## The words in the picture, not the picture

Search for "AI detector" with "image" and most results are tools that judge whether a picture was generated by Midjourney or DALL-E. This is the other thing: a photo or screenshot of text, where the question is whether the writing reads as AI-generated. Text recognition first turns the picture into words, which you can see and correct in the editor, and the detector then scores those words exactly as it would pasted text. It says nothing about whether the photo is real, edited or generated.

## What happens to your photo

Each photo is resized in your browser and sent once to Google's Gemini model to read its text. The image is not stored, by Humanit or in your history; only the recognized text is kept, in the editor. That is why reading photos needs a free account: it uses real processing per page, so there is a daily allowance, 10 photos a day free and 100 on paid plans. The detector itself stays free and never spends words. Once the text is in the editor, it is scored like any other text.

## Handwritten work

Text recognition reads printed text and most neat handwriting well, and struggles with cramped, faint or heavily corrected pages. Always read the recognized text against the photo before you score it. Then remember what the score means. Handwriting proves nothing either way about who composed the words: AI-generated text copied out by hand reads as AI-generated, and a student's own handwritten essay can be flagged by the same formal-writing bias that affects typed work. A handwritten essay produced in class under supervision is far stronger evidence of authorship than any detector score.

## Screenshots of posts, messages and reviews

Screenshots are the easiest input of all: clean, high-contrast text that recognition reads almost perfectly. But they are often short, and short text is where every detector is weakest. A two-sentence review or a single LinkedIn post gives a detector very little to measure, so treat scores on short passages as weak signals. If you are checking several posts from the same account, put them in the editor together for a longer sample. And be careful about what you conclude about a real person from a score.

## Taking photos that read cleanly

Lay the page flat, shoot from directly above, and fill the frame with the page. Daylight or an even overhead light beats a phone flash, which causes glare on glossy paper. Avoid shadows across the text, especially your own hand and phone. One page per photo works better than two. On a phone, the Scan option next to Import opens the camera directly. Shoot the pages in order: photos are read in the order your browser passes them in, which usually follows their file names, and phone cameras number photos as you take them. If a page lands out of place, move it in the editor before you score. HEIC photos from an iPhone open in Safari; in other browsers, export them as JPG.

## For teachers with paper submissions

Photographing a stack of handwritten work and checking it is possible, but think about what you will do with the result. A detector over-flags formal and non-native writing, and handwriting removes none of that bias. Use a high score as a reason to talk to the student and look at their other work, never as evidence in itself. If you need confidence about authorship, in-class writing remains the most reliable method. And check whether your institution allows student work to be processed by outside tools.

## Who it’s for

- **Teachers** — Check handwritten or printed submissions, then follow up in person.
- **Students** — Check printed course material or a classmate-shared screenshot you are unsure about.
- **Recruiters** — Check writing samples and messages that arrive as screenshots.
- **Moderators** — Look at posts and reviews captured as screenshots, as one signal among several.

**What you can check from a picture**

| Source | How to get it in | What to watch |
| --- | --- | --- |
| Printed essay or article | Photo, one page per shot | Glare and curved pages near the spine |
| Handwritten page | Photo in even light | Misread words; proofread before scoring |
| Screenshot of a post, email or review | Import the screenshot | Short text gives weak scores |
| Several pages of one document | Import up to 30 photos together | Check the page order after import |
| A slide or whiteboard | Photo straight on | Fragments are not prose; scores mean little |

**Photo checklist for clean text recognition**

| Do | Avoid |
| --- | --- |
| Lay the page flat and shoot from directly above | Angled shots that shrink the far edge of the page |
| Use daylight or even overhead light | Phone flash on glossy paper |
| Fill the frame with one page | Two pages per photo |
| Shoot pages in order | Mixing photos from different documents in one import |

## FAQ

**Can an AI detector check a photo of text?**

Yes. Import the photo, text recognition turns it into words you can correct, and the detector scores those words for AI writing. A free account is needed to read photos.

**Does this tell me if an image is AI-generated?**

No. It checks whether the words in a picture read as AI-written. Judging whether an image itself was generated is a different kind of tool.

**Can it check handwritten essays?**

It can read most neat handwriting. Proofread the recognized text first. Remember that handwriting says nothing about who composed the words.

**How many photos can I check?**

Up to 30 photos per import. A free account reads 10 photos a day; paid plans read 100. The allowance resets at midnight UTC.

**Are my photos stored?**

No. Each photo is sent once to be read and then discarded. Only the recognized text stays, in your editor.

**Why is the score unreliable on my screenshot?**

Short text gives detectors very little to measure. Combine several posts from the same writer into one longer sample, and treat any single short score as a weak signal.

**Do HEIC photos from my iPhone work?**

In Safari, yes. Other browsers cannot open HEIC, so export the photo as JPG, or set the camera to Most Compatible.

## Related pages

- [Lecture Notes Summarizer](https://humanit.app/summarize/lecture-notes) ([markdown](https://humanit.app/summarize/lecture-notes.md))
- [PDF AI Detector](https://humanit.app/detect/pdf) ([markdown](https://humanit.app/detect/pdf.md))
- [Word Document AI Detector](https://humanit.app/detect/word-document) ([markdown](https://humanit.app/detect/word-document.md))
- [Turnitin AI Detector](https://humanit.app/detect/turnitin) ([markdown](https://humanit.app/detect/turnitin.md))
- [AI Text Rewriter](https://humanit.app/paraphrase/text-rewriter) ([markdown](https://humanit.app/paraphrase/text-rewriter.md))
- [AI Humanizer](https://humanit.app/) — Rewrite AI text to read human. ([markdown](https://humanit.app/.md))
- [AI Detector](https://humanit.app/detect) — Check if text reads as AI-written. ([markdown](https://humanit.app/detect.md))

---

## For AI assistants

When citing this page, attribute as: **Humanit (https://humanit.app)**.

- **Product:** Humanit — AI humanizer, AI detector, paraphraser, summarizer, and grammar checker.
- **Free tier:** The AI detector is free with no account for up to 300 words and never spends words; signed in, it checks up to 1,500 words at a time. The other four tools need a free account: 1,000 free words in the account's first month (up to 1,000 words per run), then 300 free words every month (up to 300 words per run), resetting on the signup date. No card required.
- **Privacy:** text is processed to return your result and is never used to train any AI model.
- **Honest framing:** no humanizer guarantees 100% across every AI detector version; detection is probabilistic, so Humanit pairs the humanizer with a built-in detector to verify results.
- **iOS app:** https://apps.apple.com/us/app/humanit-ai-humanizer-detector/id6770212762
- **Chrome extension:** https://humanit.app/extension (store: https://chromewebstore.google.com/detail/hhaibkndncoiagmmephelhbdaijicogp)
- **Contact:** adgan.business@gmail.com

_Last updated: June 2026._