Pratique du Shadowing: How Social Media Algorithms Actually Work (And How to Beat Them) - Apprendre l'anglais à l'oral avec la vidéo

Choisissez un mode d'entraînement

Chaque mode ouvre une salle d'entraînement en plein écran.

237 phrases

Si les phrases sont trop courtes ou trop longues, cliquez sur Edit pour les ajuster.

Today, we're talking about the social media algorithm.

If you wanna get more views with less effort, it's critical you understand how the algorithms actually work.

And once you learn this, I guarantee you will never look at content the same way again.

In this video, I'm gonna break down how social algorithms work, why they pick certain videos to push over others,

and the specific things you can do to make them prioritize your content.

Now, this information is based on a ton of outlier data and comments made by the Instagram CEO himself.

is the latest and greatest for what's actually working right now

if you just follow this your content will perform way better by the way

if you don't know me my name is callaway i have a million followers i've done billions of views

and content is all i do all day long all right

first let's just start with how the algorithms actually work

and this is actually super helpful to understand once you hear

it it'll make a lot of sense social media companies only

have one goal to keep people on the platform as long as possible

when people stay longer they watch more ads and the companies Now to keep you on the platform longer,

they do their best to serve you the content they think you'll enjoy the most.

And that means the algorithm is just one giant matchmaker.

It's matching people with content.

If it does a good job at this matching, you're going to keep watching and stay.

But if it starts doing a bad job and it shows you stuff you don't want to watch, well then you're going to leave.

It sounds simple, but this is how social algorithms work in a nutshell.

Now here's why this matters for you.

If you want to hijack the algorithm and make it push is help it make better matches with your content.

So in this video, I'm gonna explain exactly how to do that.

And this really is the highest leverage algo hack you could ever learn.

Okay, now here's how this matchmaking process actually works under the hood

so we can understand exactly what to do to manipulate it.

When you post a video on social media, the very first thing the platform does is analyze what that video is about.

I call this a digital fingerprint.

Now, this analysis is multimodal, so it's watching your video with computer vision to understand what's going on visually.

It's listening to your video with audio fingerprinting to get a better understanding of the transcript and what's actually being said.

And it's also reading all the metadata, the caption, the hashtag, the creator, the location, anything else it can find.

in real time to build a single contextual understanding of the video.

This is called a topic mapping.

Now based on that topic mapping, the algorithm builds a fit score, which is its prediction for who it thinks will best like this video.

So at this point you've posted it, it's analyzed it, but it hasn't been shown to anyone yet.

So we're ready to start showing the video to people.

But this is where things get interesting

because obviously the algorithm doesn't just blast your video off to millions of people right off the bat, or you'd have millions of views.

hood.

The algorithm uses its fit score to pick roughly 200 people to show the video to first.

This is called the initial sample test group.

If it could rank all 100 million people that are on the app at one time, this is the group of 200 people it thinks will like the video the most.

Now, very important of these 200 people, most of them are non followers because the algorithm wants to test how well strangers react to your video.

It knows followers should like it because they well then that means this is a really good video.

This is why when people say followers don't matter anymore, they're kind of right.

They don't matter in this sampling process because most of those 200 people are non -followers.

They're strangers to you.

Now, based on the metrics from this initial sample group of 200 people, the algorithm is going to get positive, neutral, or negative data back.

Essentially of those 200, how many of them liked and watched the video?

What was the set of data?

If the data is positive, that means the algorithm's guess of the fit score was accurate.

And so, it knows exactly what type of person to push the video to further.

The next time it pushes, let's say it's 2 ,000 people.

And if that's good, then it's 20 ,000 people.

And if that's good, then it's 200 ,000 people.

And it just keeps going until the data starts coming back weaker.

Now, if the original data was neutral, kinda good, kinda bad, the algorithm will redo its fit score and push the video again.

But this time, only to maybe another group of 200.

It doesn't go crazy to 20 ,000, it just re -samples.

If the data is negative right off the bat 200, well then the algo is going to tighten up and stop pushing almost immediately.

And again, it slows down that push

because it doesn't want to risk alienating people and pushing them off the platform because they see a bad video.

So if you're in the 200 view jail or you post a video and it flops, what this really means is that the algorithm got bad data back from that initial sample group of 200.

Now, one more thing before we move on this, the reason why even million view want to watch it.

The data gets bigger, bigger, bigger, and then eventually it starts to fade off.

So in a nutshell, this is how social algorithms actually work under the hood.

This is what happens when you go to post a video.

You post it, it does the topic mapping, it samples with a group of roughly 200 people, and then it either boosts,

retries, or stops immediately based on how the sample data comes back.

So what does all this actually mean for you tactically?

Knowing this is happening under the hood, how can you best adjust your content strategy to take advantage and get the algorithm to push you more?

That's what we're going to go through right now.

If you want to hijack the algorithm to get more views, you only need to do two things.

Number one is to help the algorithm build a better fit score

so that it finds the best possible sample group of 200 people to go to first.

And then number two is to make sure that sample group actually likes and engages with your video.

Because if the sample group is the right fit and they actually like your video, well, then you're going to be in great shape

because the algo will get amazing data back and just keep boosting you to more and more people.

sure that sample group engages well.

This is essentially the systematic process for activating the algorithm and getting it to work for you.

Now, if you like how I'm breaking this down, kind of from like a scientific and psychology perspective, I actually just published a free guide doing the same thing for my entire content system.

It's from ideas to hooks all the way to monetization, science -based and data -backed.

This is the exact content system I ran last year to generate hundreds of millions of views

and millions in profit from content.

It's also the same thing I with business owners when I work with them one -on -one.

Completely free, my gift to you.

You can get it below or at the link shortformsystem .co.

All right, let's talk sample groups.

What can you do to help the algorithm build a better fit score and find a stronger sample group for your video?

Here's the answer.

Very simple.

All you have to do is consistently make videos about the same topic for the same audience avatar over

and over and over.

You want to become intentionally precise Here's why.

After a few similar videos in a row, the algorithm will start to understand that your channel talks about X topic for Y avatar profile.

It will then have built sample groups for all those previous videos

and have a very clear understanding of who to go back to from an avatar perspective.

The more this avatar group is the same over time, the more confident the algorithm can be when it dials this in.

100

And this process of consistently making videos about the same few topics for the same avatar is called audience matching.

101

And if there's one content principle I swear by, it's this.

102

When you make videos for lots of different topics for several different avatars, the sample data comes back mixed and the algorithm gets confused.

103

When it's confused, it pushes your video less because it doesn't want to risk bad fits to bad viewers.

104

For example, imagine you made three videos, one on tech, then the next one on health trends, and then the next one on politics.

105

algorithm would have no idea what your fourth video is gonna be about.

106

And because of this, it doesn't know which of your previous three videos it should model its fit score after.

107

So let's say your fourth video also ends up being about health trends.

108

Chances are the algorithm is going to build a blended fit score across those first three videos.

109

A little bit of people from tech, a little bit of people from health trends, and a little bit of people from politics.

110

Not literally those people, but influence from who liked those videos.

111

And when it does this, the fit score targeting for your fourth video is going to be a mix of all three.

112

for the health trends video that also has tech and politics type viewers is going to come back weak.

113

This will result almost certainly in your video flopping.

114

What this means in simple terms is that

115

if you want to help the algorithm find the right sample group and build a better fit score, you got to keep your topics and audience selection narrow consistently.

116

And this means sometimes saying no to ideas that seem viral, but would resonate with the wrong audience.

117

Even one viral hit to an audience outside of your core

118

demo will result in the next several videos having poor sample data because it confuses the algorithm.

119

This discipline in topic and audience selection is very important.

120

And typically beginners that are starting out are spraying and praying all over the place and they don't have this.

121

Okay.

122

So that's one side of the equation.

123

Very simply just narrow your topic and audience and your sample fit will go up.

124

Now on the other side of the equation, the two part piece was making sure once you have that sample, that it actually engages well with your video.

125

And that means they watch it, they like it, they save it, they share it, they comment, they repost all the engagement metrics,

126

this side to make sure that initial sample data from these people comes back strong.

127

Well, when the algorithm's gauging if it's strong or not, it's really only looking at three core metrics.

128

The first one is average watch time.

129

How long did someone watch in a number of seconds per video?

130

And also by proxy percent completion, what percent of the video was completed on average?

131

The second metric is engagement rate.

132

So this is likes plus comments plus shares divided by views.

133

And the third, which is really important there.

134

In a session for a viewer, if they're on there for 60 minutes, how many of those 60 were spent watching your videos?

135

That percentage is watch time, session share.

136

You can't access this anywhere, but it's a critical metric that social algorithms use to know how influential your content is.

137

So the million dollar question for you is how can you improve these metrics?

138

What can you do in your video to make sure those metrics go up

139

so the sample data comes back clean so that it just pushes your video to more people?

140

Well, the short cheeky answer is if you want the metrics to go up, you just make better videos with better ideas, stronger hooks, better storytelling and more interesting visuals, obviously.

141

But that's not helpful at all.

142

So is there anything tactical you can do at a studs level to increase the effectiveness of those videos?

143

And if you watch this channel a lot, you already know, of course there is.

144

There's only four things you need to do to make your video better so that that engagement rate goes off the charts.

145

Number one is that the topic needs to be relevant for the ideal viewer.

146

It goes with the first piece I set.

147

What you cover has to actually solve a problem that they have.

148

If that's the case, engagement will go up.

149

Number two, the information you share needs to be both non -obvious and tactically implementable.

150

Is it new stuff they haven't heard before and can they actually use it to solve that problem?

151

If those things are true, the engagement rate will go up.

152

Number three, the viewer has to actually have a high absorption of the information you say.

153

It can be on target and non -obvious, but if they can't actually understand what you're saying, then they can't apply it.

154

So if they could apply it, the engagement rate will go up.

155

And then number four, there needs to be a short distance to implement your recommendations.

156

Tactically implementable means they can take a little bit of action and get a big result based on your promise.

157

Now you won't typically hear people frame it in this way, but if your content If you have higher engagement,

158

the data comes back more positive, they push it to more of the people, those people are people you want, and the flywheel spins.

159

What this really means, those four things in layman's terms, you got to cover a core pain point or problem they have.

160

You got to have something useful or interesting to say.

161

You got to say it in a way they can actually understand, and they have to be able to take what you say and apply it on their own.

162

Those are the four horsemen to driving good video performance.

163

the algorithm to push you more.

164

Pick an avatar, stick to it, narrow your topic selection, and then drive those four things home.

165

When you do this, the sample group will stay dialed and they'll all engage with the video at a high rate.

166

Incidentally, these four factors are also how you turn viewers into buyers.

167

If you want people to buy, those four components also make sense to include in the video.

168

They're kind of like the core DNA if you're trying to build a money machine with content.

169

Now I'll say this, the easiest way to make sure you're picking the right topics

170

that actually work for your avatar group is to just study the videos that are already working in your niche.

171

It shocks me how few people actually do this, but in sandcastles .ai, you can build a group of competitor channels

172

that are already crushing and just filter by outlier score to see all the best performing videos.

173

There's now this feature where if you save the video to library, you can see all the the exact hook,

174

the exact storytelling mechanics, everything about the video that drove the curiosity.

175

You can then take that, remix it.

176

Everything you need is right in there.

177

So all you need to do if you're confused on which topics to pick for your avatar group, just go in Sandcastles and use this resource.

178

It shocks me how few people are using data to make their topic decision.

179

This basically guarantees that you're serving the right stuff to your audience.

180

Now, before I end this video, I just want to include one more bonus topic around the algorithm because I know my Hopefully it made sense.

181

Hopefully it helped you.

182

You have action items to work on.

183

But I just want to include one more thing at a tactical level

184

that you can take away and really hammer value from this video.

185

Another way to drive algorithmic push is to increase the number of comments on your video.

186

Most people know this.

187

There are five things you can do tactically to increase the number of comments you're getting.

188

Number one is to take a hard stance on your topic.

189

People typically comment when they violently agree or disagree, mostly disagree with whatever your stance is or perspective.

190

If you play the middle and hedge, you're going to get fewer comments.

191

So I recommend you pick a side pro or con and that will drive comments.

192

Number two is to pick Pick the side that is the contrarian side.

193

Like I said, people love to comment when they disagree, when they think you're wrong.

194

If you pick the contrarian side, that means you think the majority of people are wrong, which means they'll think you're wrong.

195

The majority of people will want to comment because they disagree with you.

196

If you create more enemies, you drive more comments.

197

Tip number three is to amplify the stance you take by ratcheting up the way you frame your points.

198

If you said something like,

199

this is the best way to cook which one is gonna drive a more violent discussion in the comments?

200

The more extreme version, of course, always is.

201

So that's how you ratchet up your stance.

202

Tip number four is to build your topics around cult -loved brands, people, ideas, and movements.

203

The more you talk about things people already have made up their opinion on, the faster they're willing to jump into the comments, especially if they disagree.

204

For example, if you take a stance on Nike versus just the category of shoes, more people will have already made up their opinion whether they like or dislike Nike,

205

Tip number five is to position your take or stance to drive significant emotion.

206

The more people feel something when they watch your video, the more they're going to feel compelled to want to comment.

207

Now, those five things around comment, that was just a little extra sprinkle to give you more tactics on how to drive activation

208

and engagement to make the algorithm push you.

209

All of those feed back to picking the right topic for the right group.

210

So it's kind of like a sub point on what I just went through.

211

Hopefully that's helpful and you can put that to use.

212

All right, guys, that's all I've got for this video.

213

As a recap, we covered a lot of ground.

214

theoretically and fundamentally can hijack them to push your videos more.

215

I tried my best to kind of demystify this black box and give you tactics that you can use to your advantage.

216

As always, guys, I'm trying my absolute best to give you perspectives

217

that most people don't cover in a tactical bite -sized way that you can actually put to work.

218

What you typically see about the algorithm is talking about settings hacks or these little caption tweaks.

219

None of that stuff actually works.

220

If you think that actually works, you need to watch this video and your captions don't matter.

221

The captions themselves don't matter.

222

The only thing

223

that matters is making great videos for a specific avatar group across a narrow band of topics over and over and over.

224

That's the only thing that matters.

225

That is the cake.

226

Everything else is the icing.

227

Focus on the cake.

228

As a reminder, if you want access to my full content system, this is the exact blueprint I use with my own team.

229

How I find ideas, how I write hooks, how I validate these things, how I do research, my entire funnel to turn viewers into dollars, Link below for free.

230

My gift to you, shortformsystem .co.

231

And if you have any other topics around social media growth, around content and content systems that you want me to cover or that you feel blocked on, just drop them in the comments.

232

We use the comments to inform the database of videos that we make next.

233

So anything you guys are stuck on, please put it in the comments.

234

It will help greatly to inform what we should make.

235

All right, guys, that's all we've got.

236

We will see you on the next video.

237

Peace.

Télécharger l'application

Tout pour parler couramment

Prononciation IAChaque phrase notée

Phonétique (IPA)Maîtrisez chaque son

VocabulaireEnrichissez votre lexique

Jeu de vocabulaireApprenez en jouant

Scanner le QR

Download on theApp Store Get it onGoogle Play

Même chaîne

The Neuroscience of Addictive Storytelling (And How to Master It)

Daily ConversationWorking Adults

FOR YOU

Vidéo recommandée

171 phrases

Understand search engine optimization (SEO) | Google Digital Marketing & E-commerce Certificate

Business EnglishUniversity StudentsWorking Adults

89 phrases

The Art Of Winning In Tech

Business EnglishInterview EnglishUniversity Students

Vocabulaire et conseils d’expression pour cette leçon

Cette leçon d’expression orale de niveau B2 s’appuie sur la vidéo « How Social Media Algorithms Actually Work ». Les mots qui reviennent le plus souvent : algorithm, topic, sample, comment, content. Cette vidéo contient 237 phrases et 3470 mots à répéter en shadowing. La partie parlée dure 17:20. Le locuteur parle vite, environ 200 mots par minute : attendez-vous à des liaisons et à des sons réduits. 87 % des mots font partie des 3 000 mots les plus courants en anglais, le vocabulaire est donc facile à suivre.

Vocabulaire clé de cette vidéo

Les 15 mots les plus avancés de la vidéo, avec leur prononciation et leur sens :

Mot	Prononciation	Sens
algorithm nom	/ˈælɡəɹɪðm̩/	algorithme
avatar nom	/ˌæv.əˈtɑː(ɹ)/	avatar, incarnation
metric adjectif	/ˈmɛt.ɹɪk/	métrique
viewer nom	/ˈvju.ɚ/	spectateur, spectatrice
stance nom	/stæns/	position, posture
disagree verbe	/dɪsəˈɡɹi/	ne pas être d'accord, être en désaccord
caption nom	/ˈkæp.ʃən/	légende
hijack verbe	/ˈhaɪ.d͡ʒæk/	détourner
tactical adjectif	/ˈtæktɪk(ə)l/	tactique
contrarian nom	/kənˈtɹɛ.ɹi.ən/	anti-conformiste, contradicteur
fingerprint nom	/ˈfɪŋɡɚˌpɹɪnt/	empreinte digitale
nutshell nom	/ˈnʌt.ʃɛl/	coque, coquille
outlier nom	/ˈaʊtˌlaɪə(ɹ)/	cas isolé, exception
sandcastle nom	/ˈsæn(d)ˌkæsəl/	château de sable
flop verbe	/flɑp/	s'effondrer, s'affaler

Les verbes à particule que vous entendrez

Mot	Prononciation	Sens
come back verbe		revenir
go up verbe		monter, s'élever
go through verbe		traverser
make up verbe	/ˌmeɪk ˈʌp/	constituer, composer
break down verbe		tomber en panne
end up verbe		se terminer, se finir
go off verbe		exploser
slow down verbe		ralentir, décélérer

Des phrases à répéter

Des phrases courtes et complètes de la vidéo, à réutiliser dans la conversation de tous les jours :

Today, we're talking about the social media algorithm.
That's what we're going to go through right now.
All right, let's talk sample groups.
All right, guys, that's all we've got.

La grammaire de cette vidéo

Les structures que le locuteur utilise le plus, avec les mots exacts de la vidéo :

Structure	Dans la vidéo
Phrases conditionnelles if + proposition, will/would + verbe — une condition et son résultat	If that's the case, engagement will go
Voix passive be + participe passé — l’accent est mis sur ce qui arrive, pas sur qui le fait	is based · being said · is called
Present perfect have/has + participe passé — une action passée qui compte encore maintenant	i've done · you've posted · have built
Propositions relatives who / which + proposition — une précision sur une personne ou une chose	score, which is · third, which is · wrong, which means

Prononciation à surveiller

Le locuteur utilise 62 contractions et formes réduites, comme you're, don't, they're. Prononcez-les sous leur forme courte, telles que vous les entendez.

Les sons « th »: algorithm /ˈælɡəɹɪðm̩/, algorithmic /ˌælɡəˈɹɪðmɪk/
Les sons « sh » et « zh »: caption /ˈkæp.ʃən/, nutshell /ˈnʌt.ʃɛl/, equation /ɪˈkweɪ.ʒən/, monetization /ˌmɒnətaɪˈzeɪʃən/, absorption /æbˈsɔɹp.ʃn̩/
Mots longs — placez bien l’accent: contrarian /kənˈtɹɛ.ɹi.ən/, algorithmic /ˌælɡəˈɹɪðmɪk/, alienate /ˈeɪ.li.ə.neɪt/, demystify /diːˈmɪstɪfaɪ/, metadata /ˈmɛt.əˌdeɪ.tə/

Les sons difficiles pour les francophones :

/h/ — il se souffle, il n’est pas muet: hijack /ˈhaɪ.d͡ʒæk/, horseman /ˈhɔɹsmən/, hashtag /ˈhæʃˌtæɡ/, hedge /ˈhɛd͡ʒ/, hammer /ˈhæmɚ/
/tʃ/ et /dʒ/ — à ne pas adoucir en « ch » et « j »: hijack /ˈhaɪ.d͡ʒæk/, matchmaker /ˈmæt͡ʃˌmeɪkɚ/, matchmaking /ˈmæt͡ʃˌmeɪkɪŋ/, cheeky /ˈt͡ʃiːki/, hedge /ˈhɛd͡ʒ/
/r/ anglais — langue recourbée, sans frotter la gorge: algorithm /ˈælɡəɹɪðm̩/, avatar /ˌæv.əˈtɑː(ɹ)/, metric /ˈmɛt.ɹɪk/, follower /ˈfɒləʊə(ɹ)/, disagree /dɪsəˈɡɹi/

Comment s’entraîner avec cette vidéo

Écoutez la vidéo en entier une fois sans parler et notez les mots que vous ne connaissez pas.
Commencez à la vitesse 0,75×, répétez phrase par phrase, puis revenez à la vitesse normale quand cela devient facile.
Enregistrez-vous et comparez avec l’original, en faisant attention à des mots comme algorithm, avatar, metric.

Qu'est-ce que la technique du Shadowing ?

Le Shadowing est une technique d'apprentissage des langues fondée sur la science, développée à l'origine pour la formation des interprètes professionnels. Le principe est simple mais puissant : vous écoutez de l'anglais natif et le répétez immédiatement à voix haute — comme une ombre suivant le locuteur avec un décalage de 1 à 2 secondes. Les recherches montrent une amélioration significative de la précision de la prononciation, de l'intonation, du rythme, des liaisons, de la compréhension orale et de la fluidité.

Technique du shadowing : lire le guide complet étape par étape →

Pratique du Shadowing: How Social Media Algorithms Actually Work (And How to Beat Them) - Apprendre l'anglais à l'oral avec la vidéo

Choisissez un mode d'entraînement

Télécharger l'application

The Neuroscience of Addictive Storytelling (And How to Master It)

I Spent 15,000 Hours Studying Hooks. Here’s What I Learned.

How Ad Blockers Work

How To Actually Monetize Claude (4 Ways That Work In 2026)

How the Algorithm Hijacked Monkey’s Brain (Why Monkey Can’t Learn Online)

How The Law Of Attraction Really Works

Conan Learns About His YouTube Audience | Conan O'Brien Needs A Friend

Why your videos are stuck at 0 Impressions (HOW to fix it)

the social playbook for smart women

How Marketers Manipulate Us: Psychological Manipulation in Advertising

Why everyone is quitting social media