Wall of Love
Don't take our word for it.
Here's what professors, editors, and researchers say.
Press · WIRED
Third-party analysis has consistently determined Pangram to be the most accurate, with a near-zero rate of false positives.
The Algorithmic Bridge
Pangram Labs has done something no other AI detector has: achieved a near-zero false positive rate… Pangram is, as I see it, the first successful offensive in humanity's reconquest of the web.
Press · The Authors Guild
Pangram and Originality.ai were the most reliable performers. Pangram returned 0 percent across all ten articles, correctly identifying every piece as human written.
CTO, LessWrong
Pangram noticeably outperforms other LLM content detection services I've tried. If you run a platform with user-generated content, you should probably be using it.
@JordanaSophie
And this is why pangram is THAT bitch 💅
Press · The Times
Independent studies have found Pangram to be largely reliable in distinguishing wholly human-written texts from those partly or entirely composed by software.
@sriramk
Have gotten in the habit of submitting my own 100% human / 0% LLM writing to Pangram pre-emptively to make sure it sees it as human.
Join thousands who trust Pangram to catch AI writing.
Try Pangram free@NatPurser
The day Pangram stops working is the day I leave civilized society.
@mattsclancy
Pangram correctly assessed 118 items I wrote as human authored, and the 1 AI-authored item as AI!
@akarlin
Like many, I was surprised it works so well… I have also found it to be way more helpful than I expected. Much easier to instantly identify and fade slop now.
@orphcorp
“He pangrammed it”, “pangram it first” — wild how quickly pangram became a verb.
@dhillon_p
Use @pangram, it is the most accurate! We have been using it for more than a year and have seen extremely small false positives or negatives.
@Kittywampus
I paid out of pocket for a Pangram account this year. It was worth every penny.
@paularambles
Pangram's chrome extension has ruined Substack for me. wdym all my favorite writers are just three models in a trench coat.
LinkedIn
An English professor I know swears by Pangram. He went back over two years of pre-ChatGPT student papers — they all came back 100% Human Written.
Press · City AM
The City AM sports desk uses Pangram to screen all of our OpEds — amid a huge crackdown on what we call SlopEds.
@mayfer
I formally apologize to @pangram and @max_spero_ for doubting them. It's a strong value add of a product with very low harm-to-benefit ratio… I simply didn't believe AI could be so superhuman at detecting.
@QiaochuYuan
Pangram is significantly more accurate than other AI detectors, afaict; if it says something is AI-generated that is very likely to be true.
@fabianstelzer
Huge fan of the Pangram Labs extension. They dialed in specificity/sensitivity really well, and it's not easy to beat… great model + product.
@squirtle_says
If Pangram Labs has a million fans, then I am one of them. If Pangram Labs has only one fan then that is me… If the world is against Pangram Labs, then I am against the world.
@Simon__Grimm
Pangram is one of the most pro-social, AI-centered technologies of the past year and a great example of defensive acceleration. If you don't use the plugin, do.
@willsolfiac
Installing the Pangram browser extension that automatically scans everything for AI content has been a godsend — you instantly know whether to bother reading something.
@birchlse
Genuinely an excellent product (and people working in academia need to notice and start using it!)
Numb at the Lodge
You really can trust it… we should be incredibly grateful that it works.
“I don't think Pangram is reliable.” “Have you actually used it?” “No.”
@brexton
Notes from day 1 of testing Pangram: onboarding was extremely easy.
Press · Noahpinion
Pangram — the most reliable AI text detector.
Press · Statistical Modeling, Columbia
Pangram, the leading detector whose false positive rate is reported as roughly 1 in 10,000 in its own audits and near zero on medium-to-long passages.
@gleech
I just ran Pangram against my blog and 100 fresh Claude outputs. Great stuff.
@gokulrajan_
I did not believe Pangram at first, I even criticized it, but then quickly compared the results and found @pangram is actually doing way better (no comparison!) over other absolutely useless "AI detectors."
@brexton
Best advertisement for @pangram I've ever seen… Pangram will be one of the world's most valuable companies.
@no_earthquake
Now that we have detectors that actually work (thank u Pangram), this is only a problem in people's head.
Press · Wikipedia Signpost
For samples of at least a few hundred words, Pangram is very good at sorting human-written prose from text that came straight out of an LLM.
anotherpanacea
Pangram is, pretty clearly, the best commercial AI-detection tool on the market. That's a technical claim, not a marketing one.
@chiwilliams
I honestly find Pangram helpful to tell me "no that's not actually AI".
@SvenSchnieders
Using @pangram browser extension over the last few weeks has been an interesting experience. You notice that even popular articles written and shared by known/real people are entirely AI generated.
@DeanLearner
Very grateful for @pangram saving me so much time from reading slop.
@johnhboyer
Just ran an AI paper through 3.3… I'm impressed! Your engineers deserve a round of drinks.
@dingchilling
The new Pangram 3.3 finally catches the declanked text as AI-generated. They actually report much better numbers against humanizers… honestly impressed so far!
@UOzkirimli
According to GPTZero, my fully human-written text is 77% AI. (Pangram score: 100% human.)
@benleo_econ
Love the @pangram plugin. Don't even need to copy/paste across to the website. Can just skip right along.
@foster
The Pangram Chrome AI Detection plugin is REALLY something… This thing is… a lot of fun.
@jadroy2
This is a tool I already can't live without… thank you for making the internet a little more usable.
@Sausius_wx
Most of the AI text detectors I tested said the exposé for my master's thesis — which I typed with my own fingers, not using AI at all — was at least partially written by AI 💀. Shoutout to @pangram for not believing I am an AI.
The Third Hemisphere
Pangram is actually pretty damn accurate even on short pieces of writing. This is a real step forward, and I stand corrected.
@Liface
I use Pangram nearly every day. It's been essential in moderating AI slop on r/slatestarcodex.
@pleometric
I wrote the text, had GPT "fix it" and then deleted a few words. But even then Pangram sees right through it.
@worst_account
I have literally never had something definitely human-written come back as anything but High Confidence 100% Human-Written on Pangram.
@snwy_me
Pangram almost never false-positives; it almost always only false-negatives.
@bryancsk
There are already and will continue to be superhuman LLMs capable of astonishing feats of meticulousness and creativity, and all their writing will nevertheless be caught by Pangram.
@ChorzempaMartin
Very impressed with @pangram. I had GPT 5.4 (thinking) generate a few paragraphs based on an extensive outline I drafted myself… Pangram is not fooled for a second.
@antennaria
Pangram is already one of my fav AI labs solving one of the most important problems of our time.
@shrimp_economys
Genuinely have never once put good writing into Pangram and got back 100% AI.
@chaotictransfem
Pangram is genuinely on a completely different level. I've tested it on my own writing and never had a false positive. ty for protecting us from slop 🫡
Press · Wiki Education
Based on our experiences in the fall term, we have high confidence in Pangram as a detector of AI content, at least in Wikipedia articles.
@elliebababelly
The only reliable one that I've seen rn is Pangram… (also, kinda cool that it can detect mixed AI use!)
@dundereloise
I tried it on 20 pre-ChatGPT pieces of writing… every single piece was correctly identified as 100% human written.
@prrranavv
Discovered @pangram today, blown by the accuracy and the details to which they are able to claim. Plus, their research and case studies page were brilliant!
The Cheat Sheet
Of these three, Pangram was the most accurate and had zero FPR — false positive rate.
Press · WIRED
Third-party analysis has consistently determined Pangram to be the most accurate, with a near-zero rate of false positives.
Press · The Authors Guild
Pangram and Originality.ai were the most reliable performers. Pangram returned 0 percent across all ten articles, correctly identifying every piece as human written.
@JordanaSophie
And this is why pangram is THAT bitch 💅
@sriramk
Have gotten in the habit of submitting my own 100% human / 0% LLM writing to Pangram pre-emptively to make sure it sees it as human.
@NatPurser
The day Pangram stops working is the day I leave civilized society.
@akarlin
Like many, I was surprised it works so well… I have also found it to be way more helpful than I expected. Much easier to instantly identify and fade slop now.
@dhillon_p
Use @pangram, it is the most accurate! We have been using it for more than a year and have seen extremely small false positives or negatives.
@paularambles
Pangram's chrome extension has ruined Substack for me. wdym all my favorite writers are just three models in a trench coat.
Press · City AM
The City AM sports desk uses Pangram to screen all of our OpEds — amid a huge crackdown on what we call SlopEds.
@QiaochuYuan
Pangram is significantly more accurate than other AI detectors, afaict; if it says something is AI-generated that is very likely to be true.
@squirtle_says
If Pangram Labs has a million fans, then I am one of them. If Pangram Labs has only one fan then that is me… If the world is against Pangram Labs, then I am against the world.
@willsolfiac
Installing the Pangram browser extension that automatically scans everything for AI content has been a godsend — you instantly know whether to bother reading something.
Numb at the Lodge
You really can trust it… we should be incredibly grateful that it works.
@brexton
Notes from day 1 of testing Pangram: onboarding was extremely easy.
Press · Statistical Modeling, Columbia
Pangram, the leading detector whose false positive rate is reported as roughly 1 in 10,000 in its own audits and near zero on medium-to-long passages.
@gokulrajan_
I did not believe Pangram at first, I even criticized it, but then quickly compared the results and found @pangram is actually doing way better (no comparison!) over other absolutely useless "AI detectors."
@no_earthquake
Now that we have detectors that actually work (thank u Pangram), this is only a problem in people's head.
anotherpanacea
Pangram is, pretty clearly, the best commercial AI-detection tool on the market. That's a technical claim, not a marketing one.
@SvenSchnieders
Using @pangram browser extension over the last few weeks has been an interesting experience. You notice that even popular articles written and shared by known/real people are entirely AI generated.
@johnhboyer
Just ran an AI paper through 3.3… I'm impressed! Your engineers deserve a round of drinks.
@UOzkirimli
According to GPTZero, my fully human-written text is 77% AI. (Pangram score: 100% human.)
@foster
The Pangram Chrome AI Detection plugin is REALLY something… This thing is… a lot of fun.
@Sausius_wx
Most of the AI text detectors I tested said the exposé for my master's thesis — which I typed with my own fingers, not using AI at all — was at least partially written by AI 💀. Shoutout to @pangram for not believing I am an AI.
@Liface
I use Pangram nearly every day. It's been essential in moderating AI slop on r/slatestarcodex.
@worst_account
I have literally never had something definitely human-written come back as anything but High Confidence 100% Human-Written on Pangram.
@bryancsk
There are already and will continue to be superhuman LLMs capable of astonishing feats of meticulousness and creativity, and all their writing will nevertheless be caught by Pangram.
@antennaria
Pangram is already one of my fav AI labs solving one of the most important problems of our time.
@chaotictransfem
Pangram is genuinely on a completely different level. I've tested it on my own writing and never had a false positive. ty for protecting us from slop 🫡
@elliebababelly
The only reliable one that I've seen rn is Pangram… (also, kinda cool that it can detect mixed AI use!)
@prrranavv
Discovered @pangram today, blown by the accuracy and the details to which they are able to claim. Plus, their research and case studies page were brilliant!
The Algorithmic Bridge
Pangram Labs has done something no other AI detector has: achieved a near-zero false positive rate… Pangram is, as I see it, the first successful offensive in humanity's reconquest of the web.
CTO, LessWrong
Pangram noticeably outperforms other LLM content detection services I've tried. If you run a platform with user-generated content, you should probably be using it.
Press · The Times
Independent studies have found Pangram to be largely reliable in distinguishing wholly human-written texts from those partly or entirely composed by software.
Join thousands who trust Pangram to catch AI writing.
Try Pangram free@mattsclancy
Pangram correctly assessed 118 items I wrote as human authored, and the 1 AI-authored item as AI!
@orphcorp
“He pangrammed it”, “pangram it first” — wild how quickly pangram became a verb.
@Kittywampus
I paid out of pocket for a Pangram account this year. It was worth every penny.
LinkedIn
An English professor I know swears by Pangram. He went back over two years of pre-ChatGPT student papers — they all came back 100% Human Written.
@mayfer
I formally apologize to @pangram and @max_spero_ for doubting them. It's a strong value add of a product with very low harm-to-benefit ratio… I simply didn't believe AI could be so superhuman at detecting.
@fabianstelzer
Huge fan of the Pangram Labs extension. They dialed in specificity/sensitivity really well, and it's not easy to beat… great model + product.
@Simon__Grimm
Pangram is one of the most pro-social, AI-centered technologies of the past year and a great example of defensive acceleration. If you don't use the plugin, do.
@birchlse
Genuinely an excellent product (and people working in academia need to notice and start using it!)
“I don't think Pangram is reliable.” “Have you actually used it?” “No.”
Press · Noahpinion
Pangram — the most reliable AI text detector.
@gleech
I just ran Pangram against my blog and 100 fresh Claude outputs. Great stuff.
@brexton
Best advertisement for @pangram I've ever seen… Pangram will be one of the world's most valuable companies.
Press · Wikipedia Signpost
For samples of at least a few hundred words, Pangram is very good at sorting human-written prose from text that came straight out of an LLM.
@chiwilliams
I honestly find Pangram helpful to tell me "no that's not actually AI".
@DeanLearner
Very grateful for @pangram saving me so much time from reading slop.
@dingchilling
The new Pangram 3.3 finally catches the declanked text as AI-generated. They actually report much better numbers against humanizers… honestly impressed so far!
@benleo_econ
Love the @pangram plugin. Don't even need to copy/paste across to the website. Can just skip right along.
@jadroy2
This is a tool I already can't live without… thank you for making the internet a little more usable.
The Third Hemisphere
Pangram is actually pretty damn accurate even on short pieces of writing. This is a real step forward, and I stand corrected.
@pleometric
I wrote the text, had GPT "fix it" and then deleted a few words. But even then Pangram sees right through it.
@snwy_me
Pangram almost never false-positives; it almost always only false-negatives.
@ChorzempaMartin
Very impressed with @pangram. I had GPT 5.4 (thinking) generate a few paragraphs based on an extensive outline I drafted myself… Pangram is not fooled for a second.
@shrimp_economys
Genuinely have never once put good writing into Pangram and got back 100% AI.
Press · Wiki Education
Based on our experiences in the fall term, we have high confidence in Pangram as a detector of AI content, at least in Wikipedia articles.
@dundereloise
I tried it on 20 pre-ChatGPT pieces of writing… every single piece was correctly identified as 100% human written.
The Cheat Sheet
Of these three, Pangram was the most accurate and had zero FPR — false positive rate.
Press · WIRED
Third-party analysis has consistently determined Pangram to be the most accurate, with a near-zero rate of false positives.
CTO, LessWrong
Pangram noticeably outperforms other LLM content detection services I've tried. If you run a platform with user-generated content, you should probably be using it.
@sriramk
Have gotten in the habit of submitting my own 100% human / 0% LLM writing to Pangram pre-emptively to make sure it sees it as human.
@mattsclancy
Pangram correctly assessed 118 items I wrote as human authored, and the 1 AI-authored item as AI!
@dhillon_p
Use @pangram, it is the most accurate! We have been using it for more than a year and have seen extremely small false positives or negatives.
LinkedIn
An English professor I know swears by Pangram. He went back over two years of pre-ChatGPT student papers — they all came back 100% Human Written.
@QiaochuYuan
Pangram is significantly more accurate than other AI detectors, afaict; if it says something is AI-generated that is very likely to be true.
@Simon__Grimm
Pangram is one of the most pro-social, AI-centered technologies of the past year and a great example of defensive acceleration. If you don't use the plugin, do.
Numb at the Lodge
You really can trust it… we should be incredibly grateful that it works.
Press · Noahpinion
Pangram — the most reliable AI text detector.
@gokulrajan_
I did not believe Pangram at first, I even criticized it, but then quickly compared the results and found @pangram is actually doing way better (no comparison!) over other absolutely useless "AI detectors."
Press · Wikipedia Signpost
For samples of at least a few hundred words, Pangram is very good at sorting human-written prose from text that came straight out of an LLM.
@SvenSchnieders
Using @pangram browser extension over the last few weeks has been an interesting experience. You notice that even popular articles written and shared by known/real people are entirely AI generated.
@dingchilling
The new Pangram 3.3 finally catches the declanked text as AI-generated. They actually report much better numbers against humanizers… honestly impressed so far!
@foster
The Pangram Chrome AI Detection plugin is REALLY something… This thing is… a lot of fun.
The Third Hemisphere
Pangram is actually pretty damn accurate even on short pieces of writing. This is a real step forward, and I stand corrected.
@worst_account
I have literally never had something definitely human-written come back as anything but High Confidence 100% Human-Written on Pangram.
@ChorzempaMartin
Very impressed with @pangram. I had GPT 5.4 (thinking) generate a few paragraphs based on an extensive outline I drafted myself… Pangram is not fooled for a second.
@chaotictransfem
Pangram is genuinely on a completely different level. I've tested it on my own writing and never had a false positive. ty for protecting us from slop 🫡
@dundereloise
I tried it on 20 pre-ChatGPT pieces of writing… every single piece was correctly identified as 100% human written.
The Algorithmic Bridge
Pangram Labs has done something no other AI detector has: achieved a near-zero false positive rate… Pangram is, as I see it, the first successful offensive in humanity's reconquest of the web.
@JordanaSophie
And this is why pangram is THAT bitch 💅
Join thousands who trust Pangram to catch AI writing.
Try Pangram free@akarlin
Like many, I was surprised it works so well… I have also found it to be way more helpful than I expected. Much easier to instantly identify and fade slop now.
@Kittywampus
I paid out of pocket for a Pangram account this year. It was worth every penny.
Press · City AM
The City AM sports desk uses Pangram to screen all of our OpEds — amid a huge crackdown on what we call SlopEds.
@fabianstelzer
Huge fan of the Pangram Labs extension. They dialed in specificity/sensitivity really well, and it's not easy to beat… great model + product.
@willsolfiac
Installing the Pangram browser extension that automatically scans everything for AI content has been a godsend — you instantly know whether to bother reading something.
“I don't think Pangram is reliable.” “Have you actually used it?” “No.”
Press · Statistical Modeling, Columbia
Pangram, the leading detector whose false positive rate is reported as roughly 1 in 10,000 in its own audits and near zero on medium-to-long passages.
@brexton
Best advertisement for @pangram I've ever seen… Pangram will be one of the world's most valuable companies.
anotherpanacea
Pangram is, pretty clearly, the best commercial AI-detection tool on the market. That's a technical claim, not a marketing one.
@DeanLearner
Very grateful for @pangram saving me so much time from reading slop.
@UOzkirimli
According to GPTZero, my fully human-written text is 77% AI. (Pangram score: 100% human.)
@jadroy2
This is a tool I already can't live without… thank you for making the internet a little more usable.
@Liface
I use Pangram nearly every day. It's been essential in moderating AI slop on r/slatestarcodex.
@snwy_me
Pangram almost never false-positives; it almost always only false-negatives.
@antennaria
Pangram is already one of my fav AI labs solving one of the most important problems of our time.
Press · Wiki Education
Based on our experiences in the fall term, we have high confidence in Pangram as a detector of AI content, at least in Wikipedia articles.
@prrranavv
Discovered @pangram today, blown by the accuracy and the details to which they are able to claim. Plus, their research and case studies page were brilliant!
Press · The Authors Guild
Pangram and Originality.ai were the most reliable performers. Pangram returned 0 percent across all ten articles, correctly identifying every piece as human written.
Press · The Times
Independent studies have found Pangram to be largely reliable in distinguishing wholly human-written texts from those partly or entirely composed by software.
@NatPurser
The day Pangram stops working is the day I leave civilized society.
@orphcorp
“He pangrammed it”, “pangram it first” — wild how quickly pangram became a verb.
@paularambles
Pangram's chrome extension has ruined Substack for me. wdym all my favorite writers are just three models in a trench coat.
@mayfer
I formally apologize to @pangram and @max_spero_ for doubting them. It's a strong value add of a product with very low harm-to-benefit ratio… I simply didn't believe AI could be so superhuman at detecting.
@squirtle_says
If Pangram Labs has a million fans, then I am one of them. If Pangram Labs has only one fan then that is me… If the world is against Pangram Labs, then I am against the world.
@birchlse
Genuinely an excellent product (and people working in academia need to notice and start using it!)
@brexton
Notes from day 1 of testing Pangram: onboarding was extremely easy.
@gleech
I just ran Pangram against my blog and 100 fresh Claude outputs. Great stuff.
@no_earthquake
Now that we have detectors that actually work (thank u Pangram), this is only a problem in people's head.
@chiwilliams
I honestly find Pangram helpful to tell me "no that's not actually AI".
@johnhboyer
Just ran an AI paper through 3.3… I'm impressed! Your engineers deserve a round of drinks.
@benleo_econ
Love the @pangram plugin. Don't even need to copy/paste across to the website. Can just skip right along.
@Sausius_wx
Most of the AI text detectors I tested said the exposé for my master's thesis — which I typed with my own fingers, not using AI at all — was at least partially written by AI 💀. Shoutout to @pangram for not believing I am an AI.
@pleometric
I wrote the text, had GPT "fix it" and then deleted a few words. But even then Pangram sees right through it.
@bryancsk
There are already and will continue to be superhuman LLMs capable of astonishing feats of meticulousness and creativity, and all their writing will nevertheless be caught by Pangram.
@shrimp_economys
Genuinely have never once put good writing into Pangram and got back 100% AI.
@elliebababelly
The only reliable one that I've seen rn is Pangram… (also, kinda cool that it can detect mixed AI use!)
The Cheat Sheet
Of these three, Pangram was the most accurate and had zero FPR — false positive rate.
See for yourself.
Run any text through Pangram, or catch AI as you browse with the free browser extension.