The shortlist
6 things that do it. Ours is in there, labelled, not on top.
Where one of our products fits, it sits in the list with an "ours" label, never forced to the top. Ours: made by the same studio that runs this site.
An AI-text detector that also runs image detection scans.
- Free
- Up to 2,000 words a day and 3 image detection scans a day; browser extension, Google Docs and API access included (vendor pricing page).
- Paid
- from $20 per month (Individual plan; less if billed annually)
- Best at
- Daily free checks for text, plus a few image scans, with a browser extension and Google Docs integration.
- The catch
- The free allowance is 2,000 words a day and 3 image scans a day; paid plans are billed in US dollars.
- web
- chrome
- google docs
- api
A free extension and web tools that check images, video, text and audio for AI generation, and guess which generator made an image or video.
- Free
- The Chrome extension is free with no login, per its Chrome Web Store listing; platform-scale use is by contacting sales.
- Paid
- price not checked
- Best at
- Covering every content type in one free tool, with no login.
- The catch
- It shows a likelihood, not a verdict. NewsGuard reported that a Hive result of 96.9 percent "likely AI-generated" was shared online to claim a real video was fake; Hive did not respond to NewsGuard's requests for comment.
- chrome
- web
- api
A detector for images, text, voice, music, video and deepfakes, sold mainly as an API with a file-upload checker on its home page.
- Free
- $5 in free credits to start: 1 million words of text detection free every month and 20 AI image checks.
- Paid
- from $5 per month (Pro, includes $10 of credits)
- Best at
- Covers images, text, voice, music, video and deepfakes from one set of credits.
- The catch
- In NewsGuard's 2026 test it flagged 1 of 15 real photos as AI; its CEO said low image quality can sometimes cause a false positive. Its 98.9% accuracy figure is the vendor's own, from its own evaluation.
- web
- api
Upload an image to Gemini and ask whether it was made or edited with Google AI; Gemini checks for the SynthID watermark.
- Free
- Google says it is rolling out to everyone in the Gemini app; plan limits not checked.
- Paid
- price not checked
- Best at
- Giving a firm answer for Google-made images, because it looks for Google's own watermark.
- The catch
- Per Google's post it verifies images from Google AI only; video and audio support is described as coming soon, and checking other makers' content depends on future C2PA support.
- Gemini app
Upload audio or video to see the probability it was generated with ElevenLabs voice technology.
- Free
- Free, no login required, per the vendor page.
- Paid
- price not checked
- Best at
- Checking suspect audio with no login; it uses the first minute of your sample.
- The catch
- It covers ElevenLabs audio only, and the page says it does not reliably classify audio from the Eleven v3 model.
- web
A free stylometric check for text: sentence-length variance, hedging, punctuation rhythm and phrases models favour.
- Free
- The check is free with a free account, no card (vendor page).
- Paid
- price not checked
- Best at
- Showing which style habits read as machine-written, not just a score.
- The catch
- Text only, and a free account is needed to see the result. It says every detector gives false positives, especially on well-edited prose and non-native writers.
- web
How they compare
Which one suits you, and why.
If you want free daily text checks
Pangram's free plan covers up to 2,000 words a day, which is enough for checking a draft or a pasted article. It also has a browser extension and Google Docs integration, so you can check where you already work.
What you give up is room on images. You get 3 image detection scans a day, and that's it, so sorting through a pile of pictures means hitting the ceiling fast. Paid plans are billed in US dollars.
If you want one free tool for everything, no login
Hive AI Detector's free extension and web tools check images, video, text and audio, with no login needed. It also guesses which generator made an image or video.
The catch is in what it tells you. It shows a likelihood, not a verdict. NewsGuard reported that a Hive result of 96.9 percent "likely AI-generated" was shared online to claim a real video was fake, and Hive did not respond to NewsGuard's requests for comment. A very high number still pointed the wrong way, so what you give up here is certainty.
If you want to check a Google-made image
Google's SynthID check in the Gemini app is built for this. You upload an image and ask whether it was made or edited with Google AI, and Gemini looks for Google's own watermark. Because it's searching for a mark Google put there itself, for Google-made images it can give a firmer answer.
That only covers Google's own images, though, and most of what you'll come across wasn't made by Google at all. For everything else, there are free sites that detect AI images from other places, and they give you a likelihood rather than a verdict. That page covers what those scores can tell you, and where NewsGuard's May 2026 test caught them getting it wrong, flagging real photos as AI and missing heavily manipulated ones.
But it only works for Google AI images. Per Google's post, video and audio support is described as coming soon, and checking other makers' content depends on future C2PA support. Google says it is rolling out to everyone in the Gemini app; we didn't check plan limits or pricing.
If you want to check a voice recording
The ElevenLabs AI Speech Classifier lets you upload audio or video to see the probability it was generated with ElevenLabs voice technology. It's free, needs no login, and uses the first minute of your sample.
It covers ElevenLabs audio only, though, and the page says it does not reliably classify audio from the Eleven v3 model. AI or Not also lists voice and music detection.
If you want to see why text reads as machine-written
ScriptGrain, which is ours, has a free stylometric check that looks at sentence-length variance, hedging, punctuation rhythm and phrases models favour, and shows which style habits read as machine-written rather than just a score. Our own ScriptGrain check is text only, and you need a free account to see the result. ScriptGrain's page (ours) says every detector gives false positives, especially on well-edited prose and non-native writers.
When paying is worth it
Most of these are free to start, so paying is really about volume. Pangram's Individual plan is $20 a month in US dollars, less if billed annually, with a 7-day free trial. It makes sense if the free 2,000 words a day keeps running out and text checking is a regular part of your week.
AI or Not covers images, text, voice, music, video and deepfakes from one set of credits. It starts with $5 in free credits, including 1 million words of text detection free every month and 20 AI image checks, and its Pro plan is $5 per month with $10 of credits. It's sold mainly as an API, with a file-upload checker on its home page. Its 98.9% accuracy figure is the vendor's own, from its own evaluation, and in NewsGuard's 2026 test it flagged 1 of 15 real photos as AI; its CEO said low image quality can sometimes cause a false positive.
What to watch for
The catches worth knowing first.
Treat every result as a lead to check. NewsGuard's May 2026 audit found some tools flagging real photos as AI and some missing heavily edited ones, so a detector can be wrong in both directions. A low score isn't a clean bill of health either.
The University of Sheffield says it does not use GenAI detection tools because of concerns over error rates. That's a reason not to treat one number as the final word: use it to decide what to look at next, then check the source and the context.
The same shortlist, as a table
| Thing | Free tier | Paid from | Platforms | The catch |
|---|---|---|---|---|
| Pangram | Up to 2,000 words a day and 3 image detection scans a day; browser extension, Google Docs and API access included (vendor pricing page). | $20 per month (Individual plan; less if billed annually) | web, chrome, google docs, api | The free allowance is 2,000 words a day and 3 image scans a day; paid plans are billed in US dollars. |
| Hive AI Detector | The Chrome extension is free with no login, per its Chrome Web Store listing; platform-scale use is by contacting sales. | not checked | chrome, web, api | It shows a likelihood, not a verdict. NewsGuard reported that a Hive result of 96.9 percent "likely AI-generated" was shared online to claim a real video was fake; Hive did not respond to NewsGuard's requests for comment. |
| AI or Not | $5 in free credits to start: 1 million words of text detection free every month and 20 AI image checks. | $5 per month (Pro, includes $10 of credits) | web, api | In NewsGuard's 2026 test it flagged 1 of 15 real photos as AI; its CEO said low image quality can sometimes cause a false positive. Its 98.9% accuracy figure is the vendor's own, from its own evaluation. |
| Google SynthID in the Gemini app | Google says it is rolling out to everyone in the Gemini app; plan limits not checked. | not checked | Gemini app | Per Google's post it verifies images from Google AI only; video and audio support is described as coming soon, and checking other makers' content depends on future C2PA support. |
| ElevenLabs AI Speech Classifier | Free, no login required, per the vendor page. | not checked | web | It covers ElevenLabs audio only, and the page says it does not reliably classify audio from the Eleven v3 model. |
| ScriptGrain AI-writing check ours | The check is free with a free account, no card (vendor page). | not checked | web | Text only, and a free account is needed to see the result. It says every detector gives false positives, especially on well-edited prose and non-native writers. |
How we checked
What we read, and what we haven't tested yet.
On 2 October 2026 we read each vendor's own page listed under sources: Pangram's and AI or Not's pricing pages, the Hive extension's Chrome Web Store listing and Hive's detection page, Google's Gemini and SynthID announcement, ElevenLabs' classifier page and the check page for ScriptGrain, which is ours. We also read NewsGuard's published audit and Sheffield's guidance. We did not hands-on test any detector.
If none of these fit
The manual way.
If you only need to know whether something came from one particular maker, check that maker's own tool, such as Gemini for SynthID-marked Google images. For images, a provenance check such as Content Credentials Verify shows details only if the file carries them.