Botfilter is a free Pangram alternative for everyday reading. It marks likely AI-written posts, email, and articles as you browse, on your device, with no account. It is not a replacement for checking a document. If you need a result on a specific essay or file, Pangram’s free tier covers 2,000 words a day.
Pangram, from Pangram Labs, is a hosted AI detector: you send it a document, a selection, or an API request, and its servers return a result that stays in your history. Everything below describes Botfilter’s current build. Pangram facts come from Pangram’s own pages, read on September 28, 2026.
What Pangram does
Pangram checks text and images for signs of AI generation. Its homepage says it highlights passages that were edited by AI rather than fully generated, and paid plans add a plagiarism check to each scan. It names educators, publishers, law firms, HR departments, and foundations among the groups it serves, and it connects to Canvas, Moodle, Google Classroom, and Brightspace.
You can reach Pangram in four main ways:
- The web dashboard. Paste text or upload PDF, DOCX, or RTF files, up to 100 at a time.
- The Chrome extension. Select at least 50 words on a page, right-click, and choose “Check for AI content.” The extension requires a Pangram account.
- Google Docs and Gmail. Pangram works inside Google Docs, though its pricing and help pages disagree on whether that needs a paid plan. Paid plans add a Gmail integration.
- The API. Developers send text for realtime or bulk checks, priced at $0.05 per 100 words for Pangram 4 realtime checks, paid from prepaid credits.
| Plan | Price | Allowance | Adds |
|---|---|---|---|
| Free | $0 | 2,000 words a day, 3 image scans a day | Extension, file upload, API access |
| Individual | $20 a month | 300,000 words a month, 100 image scans | Plagiarism checks, automatic feed scanning, Gmail |
| Professional | $65 a month | 5 times the Individual allowances | $200 of API credits a month |
Source: Pangram pricing page, checked September 28, 2026. Prices are billed monthly; the page offers a discount for annual billing and a separate sales contact for larger organizations.
The feed scanner is the part closest to Botfilter. Pangram’s pages say it scans posts on X, LinkedIn, Substack, and other feeds, and its help page says it checks posts over 50 words and labels each one AI, Human, or Mixed inline. It is a paid feature.
What Botfilter does instead
Botfilter marks text that is likely AI-written on the sites you visit. It uses a small detection model, a 23.6 MB file that ships inside the extension and runs in your browser. You don’t paste or select anything. As a page loads, Botfilter scores each post, comment, email, or paragraph of 25 words or more and marks the ones that are likely AI-written.
It has dedicated layouts for X, Reddit, LinkedIn, Gmail, YouTube, and Hacker News, and it scores paragraphs on every other website. You choose how a flagged post looks:
- Outline draws an orange outline around the post and leaves it in place. Hover it to see the probability, such as “Likely AI-written · 94%.”
- Collapse folds the post to a one-line bar. Click the bar to read it.
- Hide removes the post from the page. The toolbar popup still counts it, and switching modes brings it back.
In a Gmail inbox list, flagged rows are striped and dimmed instead of folded or removed. A second setting chooses how much Botfilter flags: Fewer, Balanced (the default), or More. It marks only text that looks likely AI-written. It does not label the rest as human.
Spent all Saturday fixing the fence and the dog found the new gap in about four minutes. Back to the hardware store.
Most people will scroll past this. The ones who don't will be 10x ahead in 2027.
Here's the truth: success isn't about talent. It's about consistency, clarity, and relentless execution.
I've distilled 7 lessons from building 3 companies into one simple framework. 🧵👇
our landlord finally replaced the boiler, which means I no longer shower like a raccoon at 6am hoping for three minutes of hot water. small wins. ordering celebratory dumplings tonight
Unpopular opinion: AI won't replace you. But someone who uses AI will.
The key is to embrace change, stay curious, and never stop learning. The future belongs to those who adapt.
What's one skill you're investing in this year? Drop it below 👇
Wrapped our team offsite today. Biggest takeaway for me: we spend far too much time planning the work and not enough time finishing it. Trying two meeting-free afternoons a week and will report back.
Took the long way home along the river and the whole sky went this color for about ten minutes. Nobody else on the path even looked up, which felt like a secret.
Hi Alex,
I hope this email finds you well. I came across your work and was truly impressed by your innovative approach to building products that genuinely resonate with users.
At Brightpath, we help forward-thinking teams like yours streamline their workflows, unlock new efficiencies, and drive meaningful growth. Our platform integrates effortlessly with the tools you already use, empowering your team to focus on what truly matters.
I'd love to explore how we might partner together to elevate your team's productivity to the next level. Would you be open to a quick 15-minute call next week?
Looking forward to connecting.
Best regards, Dana Whitfield Senior Partnerships Manager, Brightpath
For a walkthrough of the modes on each site, see how to filter AI-generated posts on X, LinkedIn, and Reddit.
Where your text goes
This is the largest difference between the two tools. A hosted detector has to receive your text to score it. Botfilter scores text in the browser tab where you are already reading it.
With Pangram
Pangram’s data privacy page says it stores text submitted through its dashboard and browser extension so it can show your results and history, keeps it while your account is active, and deletes it from its servers when you delete it from your history. Pangram says it does not train its models on submitted text and does not share or sell it. It states SOC 2 Type II compliance. Its Chrome extension page says feed-scanner data is stored by default for a Feed Health view, and that you can turn this off in Pangram’s preferences. Its API page lists zero data retention as an option to discuss with its enterprise team.
With Botfilter
Botfilter scores text on your device, and you don’t need an account. Page text and scores never leave your browser.
The model and its runtime load from inside the extension, detection makes no network requests, and Botfilter has no server. On your device it keeps your settings and scores keyed by a hash of the text, never the text itself. For email threads it also keeps each thread’s score and word count under the site’s own thread ID, until you clear it or the detector changes.
How Botfilter runs AI detection on your device covers what it stores, including synced settings, and the steps to check its network use yourself. The privacy page has the same facts in short.
Pangram and Botfilter side by side
| Feature | Pangram | Botfilter |
|---|---|---|
| How you check text | Paste, upload, select and right-click, or call the API | Automatic as pages load |
| Where scoring happens | Hosted: text is sent to Pangram, per its privacy and API pages | Your browser, on your device |
| Account | Required, including for the free extension | None |
| Price | Free tier; paid plans from $20 a month | Free |
| Text kept after a check | Stored in your history until you delete it or close the account | No text; scores on your device only |
| Automatic marks in social feeds | Paid plans; labels each post AI, Human, or Mixed | Included; marks only likely AI-written posts |
| Gmail integration on paid plans | Gmail, included | |
| Minimum length | 50 words in the extension | 25 words |
| Documents and Google Docs | Yes | No; text you are typing is never read |
| Images and plagiarism | Yes | No |
| Languages | Over 20 | Trained on English only |
| API | Yes, priced per word | No |
Pangram facts from its pricing, data privacy, Chrome extension, extension help, and API pages, checked September 28, 2026. Botfilter facts from its current build.
How accurate each one is
Both tools publish their own measurements, taken on different test texts, so the numbers cannot be compared with each other. Pangram has released technical reports on arXiv, most recently the Pangram 4 Technical Report in July 2026, with results on the benchmarks its authors chose. Botfilter reports results on a sealed test split of 51,332 texts that was held out from training and scored once.
On that split, at Balanced, Botfilter flagged between 0.2% and 1.8% of human-written texts depending on the source, and between 94% and 100% of AI-written texts depending on the model group. For a reading filter, the human flag rate decides how often a real person’s post gets outlined, folded, or hidden by mistake.
Ordinary web pages and work email are the human writing Botfilter flags most often. The training data, test sets, and results per model are in how the Botfilter detector was trained.
Pangram has also released Open Pangram, a set of research models and datasets, under the CC BY-NC-SA 4.0 license, which does not permit commercial use. Botfilter did not train or evaluate on any of it.
Botfilter skips texts under 25 words. It was trained and tested on English text. It still scores other languages, and how often it flags them is unmeasured. Human posts from X and LinkedIn were not available to test, so how often Botfilter flags real people there is unmeasured; Reddit posts from 2006 to 2016, flagged 0.5% of the time, are the closest measured source. AI-written texts under 80 words are harder: 91% to 96% are flagged at Balanced. Human writing can be flagged, and text a person and a model both edited is often missed. Each mark is a probability, not proof of who wrote the text.
For what false positive rates mean in practice, and why no detector’s score works as proof, see how accurate AI detectors are.
Which one to use
Pick by the job in front of you.
- Checking a student essay, a submission, or a contract draft. Use Pangram. It gives a report kept in your dashboard history, checks for plagiarism on paid plans, and connects to learning management systems.
- Screening content at volume inside your own product. Use Pangram’s API. Botfilter has no API.
- Checking text in a language other than English. Use Pangram, which lists over 20 languages.
- Reading X, LinkedIn, Reddit, or YouTube comments with fewer AI-written posts. Use Botfilter. It works on every page as it loads, and it costs nothing.
- Reading email or pages you would rather not upload. Use Botfilter. It scores text in the tab you already have open, and no copy goes anywhere.
If you are also weighing GPTZero or Originality.ai, how Pangram, GPTZero, and Originality.ai compare covers them on the same questions: where your text goes, whether you need an account, and what is free.
Getting Botfilter
Botfilter is in development. The install section has the Chrome link as soon as the listing is live. Until then, the Botfilter home page shows the three modes on a demo feed, and Introducing Botfilter explains what it flags on each site.