AI DETECTOR FOR POWERPOINT · FREE PLAN
AI Detector for PowerPoint: Were These Slides Made With AI?
Upload the .pptx file. We read every slide, tables and speaker notes included, check the images, read the metadata left by slide tools, and give a verdict slide by slide.
Open in beta · 10 free credits a month · Files deleted after 30 days
AI Document Detector
Open in betaWhat you need
Upload a PDF, DOCX or PPTX file
PDF, DOCX or PPTX, up to 20 MB and 50 pages.
At a glance
Per page
Text, embedded images and metadata are checked and summarised page by page.
AnyFact method
3 levels
Probably human, uncertain or probably AI — never a bare percentage — with the reasons and the limits of the method.
AnyFact method
10 credits
Free every month with an account, no credit card. Uploaded files are deleted after 30 days.
AnyFact free plan
Why slides need their own AI check
Slides are short, bulleted and full of images: exactly the content most text-only AI checkers cannot read. Copying text out of a deck loses the notes, the tables and the pictures.
AI slide tools also leave traces in the file itself. Decks exported by generators such as Gamma or SlidesGPT, or built with libraries like PptxGenJS, often say so in their metadata.
AnyFact reads the whole .pptx, checks text and images with the right detectors, lists those metadata signals and shows which slides drove the verdict.
Does this sound familiar?
- You want to check your presentation before a class or a defense.
- A deck was generated with an AI slide tool and you want to see what it reveals.
- You review student or team presentations and need more than a gut feeling.
- The slides are mostly images and short bullets, so pasting text is not an option.
Here's what we check
Upload the file
PDF, DOCX or PPTX, up to 20 MB and 50 pages.
Text, images, metadata
Each part goes to the right detector; unusual authoring software or dates are flagged.
A verdict in three levels
Probably human, uncertain or probably AI — never a bare percentage — with a short explanation and the limits of the method.
How to read the result
The signals point to a person. It is not a guarantee: a careful generator or heavy editing can leave few traces.
The signals disagree or are too weak. We say so rather than guess, and list what was found.
The signals point to generation. It is an indication, not proof — check the context before acting.
Sample layout. Real reports name the signals that were checked and the limits of the method.
What does an AI detector for PowerPoint check?
An AI detector for PowerPoint reads the .pptx file itself: the text in every shape on every slide, grouped shapes and tables included, the speaker notes, the images and the metadata the slide tool left behind. AnyFact scores the text, sends the images to its photo detector, lists the metadata signals and gives the result slide by slide, with an overall verdict of probably human, uncertain, probably AI or mixed.
Presentations are the format people check most often with AnyFact. Of the first 112 documents analysed in the beta, 50 were PowerPoint files, against 31 Word files and 30 PDFs, and most came from students checking their own coursework.
Decks are also the hardest format for a text-only AI checker for presentations. A slide holds a title, a few bullets and a picture; the paragraph you actually say out loud is in the notes; and some decks are produced end to end by AI slide tools or by scripts. An AI PowerPoint detector has to read the whole file to see any of that.
How to check a PowerPoint for AI (step by step)
An AI checker for PPT files works on the .pptx: save the deck with its notes, upload it, then read the result slide by slide.
- 01Get a .pptx. Old .ppt files are refused: open them in PowerPoint and use File > Save As > PowerPoint Presentation (.pptx). From Google Slides, use File > Download > Microsoft PowerPoint (.pptx); from Keynote, File > Export To > PowerPoint.
- 02Keep the speaker notes in the file. They are often the longest text in a deck and give the text check the most to work with.
- 03Check the limits: up to 50 slides and 20 MB. For a heavy deck, compress the pictures (select a picture, then Picture Format > Compress Pictures) or split it.
- 04Upload the .pptx to the document check. It uses 3 credits; a free account gets 10 a month.
- 05Read the overall verdict, then go slide by slide: which slides were flagged, and was it the text, an image or both?
- 06Read the metadata signals, such as an AI slide tool named in the file or a deck built by a programming library, with the explanation of what each can and cannot mean.
- 07Download the PDF report. If you believe the result is wrong, report it with the “the result is wrong” button.
- 08Act on the pattern. If it is your deck, keep the report with your drafts, sources and image credits. If it is someone else’s, use the flagged slides to prepare questions, not conclusions.
Made the deck in an online tool such as Canva or Gamma? Export it to .pptx (or PDF) from the tool and upload that file. The export is what may carry the tool’s name in its metadata. The Canva AI detector and Google Slides AI detector pages cover those two tools step by step.
What does an AI presentation detector read in a .pptx?
AnyFact reads the text of every shape, the tables, the speaker notes, the slide images and the metadata of a .pptx. The table shows what happens to each part and where it appears in the result.
| Part of the .pptx | What is checked | How it shows in the result |
|---|---|---|
| Text in shapes | The text of every shape on every slide (titles, placeholders, text boxes), including shapes inside groups | Scored as part of the deck’s text; flagged text shows on its slide |
| Tables | The text in the table cells | Scored as text, like the shapes |
| Speaker notes | The notes under each slide | Scored as text; notes often hold the longest passages in a deck |
| Slide images | Up to 30 images per deck go through the photo detector; images under 200 px or thinner than 1:6 (icons, logos, divider lines) are skipped | Images that look generated are flagged on their slide |
| Metadata | The application and creator fields, the dates, the total editing time, or the absence of metadata | Listed as signals, each with what it can mean and why it is not proof |
| Text scoring | The text is grouped into blocks of about 1,000 words and scored with the Fast-DetectGPT method and an open-source language model (Qwen2.5) | Under 300 words of text, the result says the text check gives little evidence |
Why a 20-slide deck can hold less text than one page of an essay
Count the words. Twenty slides with a title and four short bullets each add up to a few hundred words, close to the 300 words below which the text check gives little evidence. Because text is scored in blocks of about 1,000 words, the bullets of several slides are read together.
Speaker notes, tables, images and metadata then carry much of the weight. That is why the slide-by-slide view matters more for decks than for essays, and why a deck checked without its notes says less.
What about words inside pictures?
Text that is part of an image, such as a screenshot of a paragraph or an infographic pasted as a picture, is checked as an image by the photo detector, not read as text. The text check covers what is typed into shapes, tables and notes.
Was this presentation made with Gamma or ChatGPT? What the file can tell you
Sometimes the file says so. AI slide tools and assistants can leave their name in a .pptx’s metadata, and AnyFact lists it as a signal when they do: Gamma, Tome, Beautiful.ai, SlidesGPT, Decktopus, MagicSlides, Presentations.AI, Plus AI, ChatGPT, Copilot, Gemini or Claude in the application or creator field.
When an assistant hands you a ready-made .pptx, it may have built the file with a programming library such as python-pptx or PptxGenJS rather than with PowerPoint. A deck produced by a library instead of an office application is a second kind of signal, and it is listed too.
Silence proves nothing. If a generated deck was opened in PowerPoint and saved again, the application field may now name PowerPoint; if its slides were copied into a new file, the original metadata is gone. The answer then rests on the text and images, and the text check measures how predictable the words are. It does not identify which tool, if any, wrote them.
| What you find | What it suggests | What it does not tell you |
|---|---|---|
| An AI slide tool named in the metadata | The deck was made or exported with that tool | Whether the content, the research or the talk is the presenter’s own |
| A library such as python-pptx or PptxGenJS | The file was assembled by a script, as assistants and generators can do | Which assistant was used, or whether a person wrote the text first |
| PowerPoint named, with a normal editing time | The file was built, or at least saved and edited, in PowerPoint | That no AI tool was involved earlier |
| No metadata | The metadata was stripped or never written | Anything about the origin of the content |
Using an AI slide tool is not misconduct in itself: many courses allow it for layout and design. What matters is your course’s policy and whether the use was declared.
PowerPoint metadata signals: what they mean and why they are not proof
A .pptx records the application that saved it, its creation and modification dates and its total editing time. Each of these can point to a generated deck, and each has ordinary explanations.
| Signal | What it can mean | Why it is not proof |
|---|---|---|
| An AI tool named in the creator or application field (Gamma, Tome, Beautiful.ai, SlidesGPT, Decktopus, MagicSlides, Presentations.AI, Plus AI, ChatGPT, Copilot…) | The deck was generated or exported by that tool | The tool may have done only the design; metadata can be edited by anyone with the file |
| Produced by a library (python-pptx, PptxGenJS…) rather than an office application | The file was built by code, as AI generators and assistants do | Learning platforms, report generators and converters build decks the same way |
| 0 minutes of recorded editing time for a long deck | The slides were not edited in this file in PowerPoint | Decks downloaded from Google Slides or Canva, or copied into a fresh file, can show the same |
| A modification date earlier than the creation date | The file was copied, converted or had its dates rewritten | Converters, “Save as” and copying between computers with different clocks do this |
| A creation date in the future | A wrong clock or edited dates | Usually a misconfigured computer clock |
| No metadata at all | The metadata was removed or never written | PowerPoint’s Inspect Document feature, privacy tools and some converters remove it |
How to check a PowerPoint file’s metadata yourself
- 01In PowerPoint for Windows, open File > Info. The Properties panel shows the dates and the author; “Show All Properties” reveals more fields.
- 02In PowerPoint for Mac, open File > Properties.
- 03For the raw fields, make a copy of the file, rename it from .pptx to .zip and open it. docProps/app.xml holds the application name, TotalTime (editing time in minutes) and the number of slides and notes; docProps/core.xml holds the author and the dates.
How to read a slide-by-slide AI detector result
Look at which slides were flagged and by which part of the check: text, notes, images or metadata. Each points to a different question.
| What you see | Likely reading | What to do next |
|---|---|---|
| Only images flagged | Generated illustrations in a deck whose text reads as human | Check whether AI images were allowed and credited on the slide or in the sources |
| Speaker notes flagged, slides not | The script was drafted or polished with an assistant | If it is your deck, keep the drafts of your notes; if not, ask the presenter to talk through those slides |
| An AI slide tool in the metadata, text uncertain | The deck was built with that tool; the text gives little evidence either way | Ask how the tool was used: design only, or content too |
| Most slides probably AI, and a tool named in the metadata | Several signals point the same way | Still an indication: talk with the presenter before anything else |
| Mostly uncertain | Short bullets and many images give thin evidence | Treat it as “no clear signal”, not as a near miss |
| “Little evidence” on the text | Under 300 words of text in the deck | Rely on the notes, images, metadata and the presenter’s explanation |
A mixed deck is the normal case in group work: five people, five writing styles, slides from a template and pictures from three sources. The slide view shows which parts stand out so that the group can explain them, instead of one score for the whole deck.
When the score falls in the grey zone, Claude, Anthropic’s model, reviews the content as a second opinion. It can move the score only a little and never decides alone.
What are the limits of an AI checker for presentations?
An AI checker for presentations gives an indication, not proof, and slides are a format with little text. Keep these limits in mind.
- Up to 50 slides and 20 MB per file. Split long decks or compress the pictures.
- Legacy .ppt files are refused: save as .pptx or PDF first. The conversion rewrites part of the metadata.
- Image-heavy decks give the text check little evidence, and under 300 words of text the result says so.
- Up to 30 images per deck are checked, and icons, logos and thin lines are skipped, so in a long photo-heavy deck some images are not checked.
- Text inside pictures is checked as an image, not read as text.
- Short bullets look alike whoever writes them, and titles such as “Introduction” or “Conclusion” carry no signal.
- Translated or heavily edited slides, and slides written by non-native writers, are harder to judge.
- Detectors disagree: the same deck can get different verdicts from different tools.
- No detector predicts what Turnitin or your school’s own tool will say.
Our guide are AI detectors accurate? explains why results differ between tools. The AI document detector page covers the same engine for all three formats: PDF, Word and PowerPoint.
Checking your own slides before a class or a defense
If you made the deck, what protects you is its history and its sources, not a detector score. Before you present:
- Keep version history. PowerPoint keeps it for files saved to OneDrive or SharePoint (File > Info > Version History); Google Slides under File > Version history > See version history.
- Keep the drafts of your speaker notes, your outline and the research behind each slide.
- Credit your images: note where each photo, chart or illustration came from, and if an image was generated with AI and your course allows it, say so on the slide or on a sources slide.
- In a group project, note who made which slides; any question about a slide is then quick to answer.
- If you used an AI slide tool or assistant, keep a short note of what for (layout, images or text) and declare it as your course requires.
If a teacher questions your presentation
- 01Ask which slides were flagged, by which tool, and whether it was the text, the images or the metadata.
- 02Show your version history, notes and sources.
- 03Offer to present or explain the flagged slides again: knowing the material is clear evidence of your work.
- 04If the metadata names a slide tool you used for design, explain what it did and what you wrote yourself.
- 05Follow your school’s academic integrity policy, and use its appeal process if you disagree with the outcome.
Institutional tools such as Turnitin are separate systems; AnyFact does not replicate or predict them, as our Turnitin alternative page explains. To check essays and reports as well as slides, see the AI checker for students.
How should teachers use an AI slide detector on student presentations?
Use an AI slide detector to choose which questions to ask, not to grade. A fair review of a presentation looks like this:
- 01State the AI policy for presentations in the assignment: which tools are allowed for design, images and text, and how to declare them.
- 02Check the .pptx rather than a PDF export, so that the speaker notes and the deck’s metadata are included.
- 03Read the result slide by slide and note what was flagged: text, notes, images or metadata.
- 04Ask the student to talk through the flagged slides and how they were made.
- 05Weigh the conversation, the version history and the sources above any detector result, and never sanction on a detector result alone.
If the same student also hands in a written report, the AI detector for Word documents and the AI PDF checker read those formats with the same care, section by section and page by page.
Before you check
Is the result proof that a content was made by AI?▼
No. It is an indication, not proof. Detectors can be wrong in both directions, especially on compressed, cropped, edited or very short content. Use the result to decide what to check next, not as a final judgement on anyone.
Why no percentage?▼
A single percentage suggests a precision no detector has. You get one of three levels — probably human, uncertain or probably AI — with the signals behind it and the limits of the method.
How does the analysis work?▼
When the content carries provenance data, such as C2PA content credentials or metadata, we read it first. Then an open-source detector scores the content. When the score is borderline, Claude reviews it as a second opinion, and the result is explained in plain words.
What happens to my files?▼
They are processed to run your analysis — including, for borderline cases, a review by Claude, Anthropic’s model — and deleted after 30 days.
What does it cost?▼
There is a free plan: a free account gets 10 credits every month, with no credit card. Pro ($15 a month) gives 1,000 credits a month; Business ($49 a month) gives unlimited Business Checks and AI checks under a fair-use limit.
Can an AI detector check a .ppt file?▼
Not the legacy .ppt format: it is refused. Open the file in PowerPoint, use File > Save As > PowerPoint Presentation (.pptx), or save it as a PDF, then upload the new file.
Does the AI detector for PowerPoint read speaker notes?▼
Yes. The speaker notes are read and scored along with the text of every shape, including grouped shapes and tables. Notes are often the longest text in a deck, so keep them in the file you upload.
Can it tell if a presentation was made with Gamma?▼
Sometimes. When Gamma or another AI slide tool, such as Tome, Beautiful.ai, SlidesGPT, Decktopus, MagicSlides, Presentations.AI or Plus AI, is named in the file’s metadata, AnyFact lists it as a signal. If the deck was re-saved in PowerPoint or its slides were copied into a new file, that trace can disappear.
Can I check Google Slides or Canva presentations for AI?▼
Yes, by downloading them first. In Google Slides, use File > Download > Microsoft PowerPoint (.pptx); in Canva, download the design as a PowerPoint or PDF file. Then upload that file to the document check.
Are the images on my slides checked for AI?▼
Yes. Up to 30 images per deck go through the photo detector, and images that look generated are flagged on their slide. Images under 200 px or thinner than 1:6, such as icons, logos and divider lines, are skipped.
How many slides can the AI detector for PowerPoint check?▼
Up to 50 slides and 20 MB per file. A check uses 3 credits, and a free account gets 10 a month.
Why does my presentation get an uncertain result?▼
Slides hold little text: short bullets and many images give the text check thin evidence, and under 300 words of text the result says so. Uncertain means no clear signal, not a hidden “probably AI”.
Will the result predict what Turnitin says about my slides?▼
No. Turnitin and other school tools are separate systems, and AnyFact does not replicate or predict them. The result shows how your deck reads to AnyFact’s detectors, with the reasons.
Check a PowerPoint, slide by slide
What you need
Upload a PDF, DOCX or PPTX file
Free plan — 10 credits every month, no credit card.
Each result is an indication, not proof.
Files deleted after 30 days.
You might also want to check
AI Checker for Students: Check Your Assignment Before You Submit
Check your essay, report or slides for AI before you submit: upload the file or paste the text and see which parts look AI-written, and why.
AI Detector for Word Documents: Upload the .docx, Get a Verdict
Upload a Word .docx instead of pasting text: up to 50 pages, section-by-section results, images and editing metadata checked, and an explained verdict. Free plan.
Was This Document Made With AI?
Check the text, images and metadata of a PDF, DOCX or PPTX file, page by page, with a verdict explained in plain words. Free plan.