The National Telegraph - Wyatt Claypool - Carney sounds robotic - Are Liberals using AI for social media？

00:00:00.000 Hey guys, Wyatt Claypool here, and welcome back to the National Telegraph YouTube channel.

00:00:06.020 Today on the show, I want to talk about something slightly silly,

00:00:10.100 but it is still quite revealing of Prime Minister Mark Carney.

00:00:14.300 Does he use AI to write all of his tweets?

00:00:18.180 Really, the answer is either he uses AI, or he's effectively a robot and already sounds like AI anyways,

00:00:25.040 but I still want to investigate this just for the fun of it.

00:00:28.360 But before we get into it, I just want to remind you guys, if you like the show,

00:00:31.780 make sure to leave a like on this video, subscribe if you are not yet a subscriber,

00:00:36.240 leave a comment about what you think about all this.

00:00:39.120 Is this AI, or is Mark Carney already just a robot?

00:00:42.480 And of course, if you want to support the channel, you can always hit the join button below the video

00:00:46.440 and make a small monthly contribution to allow me to make the channel more sustainable

00:00:51.000 and be less reliant on the YouTube algorithm.

00:00:53.800 Anyways, we're going to start this video off with a conservative MP I really like,

00:00:59.640 first raising the question himself, that MP being the Ontario MP Matt Strauss.

00:01:06.680 He says here,

00:01:07.740 Now, I want to get to this video in just a second, but I actually do want to mention

00:01:28.740 that as somebody who works at the legislature in British Columbia, of course, Victoria,

00:01:35.120 I know from communications people in other parties that MLAs totally use ChatGPT to write their speeches.

00:01:43.960 They are genuinely that lazy.

00:01:46.340 Now, the MLAs in British Columbia don't actually get paid all that much compared to MPs,

00:01:51.340 so maybe we can let them off the hook.

00:01:53.140 But it's absolutely pathetic that MPs who make $200,000 a year minimum with their teams,

00:02:01.420 with communications people and other legislative assistants cannot write their own speeches.

00:02:07.860 You can really tell in the BC legislature that people use ChatGPT in order to extend their speeches,

00:02:14.740 because how is it that whenever you give people five minutes to say something,

00:02:19.440 the statement that they read always takes exactly five minutes?

00:02:23.420 You know, keep it simple, sweetie, should be the standard.

00:02:26.600 Keep it as short as possible, which is what we actually do at 1BC in British Columbia.

00:02:31.180 But now I'm getting to Matt Strauss here, and then I want to go through some of his post examples.

00:02:35.640 Hey, everyone. I'm really happy to be back home for the weekend from Parliament.

00:02:40.820 I seem to have created a bit of a stir on Twitter, calling out the use of ChatGPT by our Prime Minister

00:02:46.080 and basically all of his tweets.

00:02:48.300 It's not just the use of the giveaway and dash.

00:02:50.620 It's the rigid formulation of all of them.

00:02:54.480 We're going to X, Y, and Z, not only because P and Q, but also R and S to ensure A and B.

00:03:01.180 I used to be a newspaper copy editor.

00:03:06.140 I've written in newspapers.

00:03:08.080 My background is an English literature degree from the University of Waterloo.

00:03:11.920 I know when it's a human writing.

00:03:13.760 I know those aren't human tweets.

00:03:16.940 This is really important.

00:03:18.160 I think people have an expectation of authenticity from political leaders,

00:03:21.540 most especially the man who holds the highest office in the land.

00:03:26.620 I think if you have something important to say to people, you should say it to people.

00:03:30.400 I don't think people want to hear what ChatGPT thinks our Prime Minister might say.

00:03:36.440 I think we want to hear what the Prime Minister has to say.

00:03:39.340 So I'm calling on Mark Carney in his office to just stop doing that.

00:03:42.920 We want to hear your words.

00:03:44.260 We want to know what the plan is.

00:03:45.860 We do not want any more ChatGPTs.

00:03:48.300 Thanks.

00:03:49.040 So I do like this master level trolling from Matt Strauss here,

00:03:53.280 because whether it's true or not,

00:03:55.500 it does point out something about the way that Mark Carney does tweet.

00:04:00.320 I just have to bring up some of these examples for you guys here.

00:04:03.680 So Matt Strauss posted this back on February 6th, a couple of days ago.

00:04:09.120 And he shows right here, bring this up on screen.

00:04:13.060 These are some of his examples.

00:04:14.400 We have this Mark Carney tweet that reads,

00:04:18.660 Before recent trade tensions,

00:04:20.380 China represented a roughly $4 billion export market for Canada canola seed.

00:04:25.360 With our new trade arrangement,

00:04:26.800 tariffs are expected to drop about 70% on canola seed by March 1st.

00:04:31.440 M- Unlocking billions in business for Canada's canola farmers.

00:04:36.860 This next one.

00:04:37.640 Canada's new government is moving fast, M-

00:04:40.260 and we need Parliament to move swiftly to deliver for Canadians.

00:04:43.680 We put forward bills to strengthen bail and sentencing laws,

00:04:47.200 build up a stronger economy,

00:04:48.680 cut costs for the middle class,

00:04:50.140 and it just keeps going there.

00:04:52.460 Canada's big city mayor's caucus,

00:04:54.500 and I agree Parliament needs to pass Bill C-14 urgently, M-

00:04:57.760 to strengthen the criminal code

00:04:59.940 and keep violent and repeat offenders off our streets.

00:05:02.660 Good to meet today and discuss how we can make our streets safer,

00:05:05.580 build more homes,

00:05:06.340 and the thing is that the message is always kind of the same.

00:05:09.960 As he is saying, it's like,

00:05:11.540 we need to do X and Y because of P and Q

00:05:14.360 so that A and B don't happen

00:05:16.020 or so that A and B get accomplished.

00:05:18.440 It's all kind of almost impossibly formulaic.

00:05:22.880 You don't really write this way.

00:05:24.580 People have styles,

00:05:25.720 and they will very rigidly stick to their style,

00:05:28.620 but their style doesn't usually require them

00:05:31.040 to write this formulaically in every single tweet.

00:05:35.020 So we have here Matt Strauss saying,

00:05:38.640 ChatGPT is clearly writing the Prime Minister's tweets,

00:05:41.540 and I think Canadians deserve better leadership than that.

00:05:45.080 I think another one where it's an M-

00:05:47.060 where he's talking about, you know,

00:05:48.740 working together to build up Canada's industries or whatever.

00:05:53.060 And so Matt Strauss put this all into ChatGPT or Grok

00:05:58.540 to see what was going on.

00:05:59.940 He says, look, the nation is in crisis,

00:06:03.320 and I think Canadians deserve to hear from their Prime Minister

00:06:05.500 rather than ChatGPT.

00:06:07.600 And he asked about this.

00:06:09.680 He says, look at Mark Carney's last 20 tweets.

00:06:13.420 Are they generated by LLMs?

00:06:16.040 Thoughts.

00:06:16.740 Based on the analysis of Mark Carney's last 20 tweets,

00:06:19.560 covering the period from February 6th, 2026,

00:06:22.160 back to approximately February 4th, 2026.

00:06:25.580 They are strongly indicated that they're generated

00:06:28.540 or heavily assisted by large language models.

00:06:31.980 This conclusion is drawn from stylistic patterns in the text,

00:06:35.440 consistent formatting choices,

00:06:37.340 and external discussions on X

00:06:39.260 about Carney's content showing LLM hallmarks.

00:06:41.660 Now, if you do put ask it about specific tweets,

00:06:46.140 sometimes it will say, well,

00:06:47.460 because the em dashes have spaces around it,

00:06:50.200 it may not be AI

00:06:51.700 because usually what AI does is that

00:06:54.640 it'll have a word like,

00:06:56.660 it doesn't really matter what the word is,

00:06:57.940 and then it'll be em dash, em dash, and,

00:07:00.360 and there won't be any spaces between it.

00:07:02.660 But that doesn't rule out actually using AI

00:07:05.660 considering that, well,

00:07:06.940 if you're his communications director,

00:07:08.540 you're probably not just going to copy and paste it

00:07:11.240 into his Twitter feed.

00:07:13.040 What you're probably going to do

00:07:14.200 is clean it up a little bit

00:07:15.580 and adding spaces around the em dashes

00:07:17.740 may be one of the things that you do.

00:07:20.160 But overall, the tweets do not sound like

00:07:23.040 how a normal human being speaks.

00:07:24.820 And I know that politicians sometimes do speak

00:07:27.880 in a rigid, professional way,

00:07:30.140 but you don't usually also make every single tweet

00:07:34.460 the exact same thing over and over again.

00:07:37.280 And I want to take you guys

00:07:38.680 to Mark Carney's page itself

00:07:41.420 to kind of prove my point as well,

00:07:43.680 because I think the French tweets

00:07:44.960 actually kind of end up revealing something.

00:07:47.660 So we are on Mark Carney's Twitter page here,

00:07:51.060 Mark J Carney,

00:07:52.820 or at Mark J Carney.

00:07:54.580 I'm just making sure that it,

00:07:55.740 like specifying that this is the one

00:07:57.720 that Matt Strauss was analyzing.

00:08:00.160 He didn't get that wrong.

00:08:01.520 So notice also,

00:08:03.360 this tweet right here

00:08:04.620 does not have em dashes in it,

00:08:06.300 but it was a couple of days

00:08:07.500 after he was called out by Matt Strauss.

00:08:10.760 And so suddenly we don't have em dashes anymore.

00:08:14.500 So yeah, also the French ones never have em dashes,

00:08:17.480 but always the English ones

00:08:18.840 before a certain period of time.

00:08:20.640 So we're going to go back.

00:08:23.140 So these tweets are too short

00:08:25.260 to obviously have anything in them.

00:08:27.160 But once we go back to February 6th,

00:08:30.860 the day he was called out by Matt Strauss,

00:08:32.960 you see the em dashes appear.

00:08:35.680 But the French one does not have it.

00:08:38.840 Going down, English one,

00:08:41.480 oh, that one is another short one

00:08:43.180 where it doesn't really apply.

00:08:45.220 But obviously you go to English,

00:08:47.860 em dashes, French, no em dashes.

00:08:50.840 It's English, em dashes.

00:08:54.300 French, no em dashes.

00:08:57.180 English has an em dash.

00:08:59.600 French, no em dashes.

00:09:02.080 English, em dash.

00:09:04.400 French, no em dash.

00:09:06.320 Like it's kind of funny.

00:09:07.680 And at least to me,

00:09:08.520 like if it's a super short tweet,

00:09:10.480 you won't have the em dash

00:09:11.800 because there's just no room for it

00:09:13.480 because he's not listing anything.

00:09:15.120 But then why is it that the French does not have it?

00:09:18.100 I'm pretty sure that em dashes are still a concept

00:09:20.860 in formatting in French.

00:09:22.540 Like you can still use them.

00:09:24.240 So what that at least indicates to me

00:09:26.940 is that Mark Carney has his team chat GPT

00:09:30.880 or some AI generate tweet options.

00:09:33.960 They clean them up a little bit.

00:09:35.680 And then when it gets passed to the person

00:09:37.440 who then rewrites it in French,

00:09:39.180 they remove the em dashes

00:09:40.740 because it's not really their style.

00:09:42.740 Or else why would that not be there

00:09:45.160 if it's supposed to be like one-to-one?

00:09:47.540 It feels like it then gets passed

00:09:49.900 to a second person who's not using chat GPT

00:09:52.980 and they write it up completely differently

00:09:55.100 when they basically look at the tweet

00:09:56.880 and then they just rewrite it somewhere else

00:09:58.640 making sure that it sounds good

00:10:00.420 because it's one of those Google Translate issues.

00:10:03.900 You can't just translate it

00:10:05.260 to a different language one-to-one.

00:10:07.300 You're going to have to rewrite it

00:10:08.540 around how the language usually likes to phrase things.

00:10:12.140 So could it be that Mark Carney is just a robot

00:10:16.100 and this is how he and his team tend to write things?

00:10:19.920 Sure, but you'd think there would be

00:10:21.440 a couple more unique tweets in there.

00:10:23.780 Like obviously there are unique tweets

00:10:25.700 like when he's just saying,

00:10:26.940 hey, good luck to Team Canada

00:10:28.480 at the Winter Olympics in Italy.

00:10:30.940 That's different than like a long sort of policy tweet.

00:10:34.980 Every single policy tweet

00:10:36.680 has the exact same rhythm to them

00:10:39.240 and they're very obnoxious.

00:10:41.140 Like you think that,

00:10:41.920 and also suddenly in some of his longer tweets

00:10:44.540 right after he gets called out by Matt Strauss

00:10:46.620 that the last one he did does not have an M-dash in it

00:10:49.760 even though I can go through the last several

00:10:52.400 and it's completely consistent every single time.

00:10:56.560 Anyways, that's pretty much it for this video.

00:10:59.180 I didn't want to go on too long about this topic.

00:11:03.560 Politicians have to not use AI.

00:11:05.540 Like I use AI a bit.

00:11:07.160 It's usually to answer questions

00:11:08.860 or to process some data

00:11:10.400 where I could process it myself

00:11:12.220 but it would be two hours out of my life

00:11:13.920 to put it in a format that AI can do itself.

00:11:17.380 But no, don't use it.

00:11:19.460 Just being authentic is better.

00:11:21.480 I guess that's a big advantage

00:11:23.100 to the federal conservatives.

00:11:24.700 I think that Carney's inauthenticity

00:11:26.440 will become a big Achilles heel of him going forward

00:11:29.680 because not like the economy is good,

00:11:31.940 not like the job market is good,

00:11:34.000 everything kind of sucks

00:11:35.140 and if you can't even connect with people

00:11:37.040 on a human level,

00:11:38.000 that might start to really aggravate them

00:11:40.100 when you don't seem to actually be able

00:11:41.680 to connect with the problems

00:11:42.860 that the average person is having in this country

00:11:45.540 or I don't know

00:11:48.100 or just keep, they're going to keep doing it.

00:11:51.160 They're not going to learn their lesson.

00:11:53.200 I actually assume that he'll eventually

00:11:54.300 just go back to M-dashes

00:11:55.420 but based on how many people I know

00:11:57.540 in communications and politics use AI,

00:12:00.560 it wouldn't shock me

00:12:01.660 that the prime minister's office was doing it too

00:12:04.280 which I guess means

00:12:05.480 that we could all be communications directors.

00:12:07.840 It's simple enough

00:12:08.700 if all you need to have

00:12:09.920 is a Grok subscription on X

00:12:12.220 to apparently have this six-figure job

00:12:15.000 at the highest parts of Canada's government.

00:12:18.760 Anyways, with that all being said,

00:12:20.500 thank you guys for watching,

00:12:22.120 like, share, subscribe

00:12:23.000 and I'll see you guys all next time.

00:12:25.240 Thank you.

The National Telegraph - Wyatt Claypool - February 08, 2026

Carney sounds robotic - Are Liberals using AI for social media？

Episode Stats

Length

Words per Minute

Word Count

Sentence Count

Hate Speech Sentences

Summary

Transcript