Why your ad is invisible

5 min read·Lesson 1 of See like a media buyer

By the end of this lesson you will have six real ads saved on your phone, three that stopped you and three you scrolled past, plus Claude's first read on what the stoppers have in common. Those six screenshots are the raw material for the next three lessons.

Nobody decides to watch your ad. People are on a bus, in a queue, half-watching TV, thumbing through a feed with the sound off. Your ad appears between a friend's baby photo and a video their group chat is arguing about, and it gets the same treatment as everything else there: a fraction of a second of attention, then gone.

Most first-time advertisers carry one belief into this: "my product is good, so the ad will work." It is the most expensive belief in advertising. The feed does not know your product is good. The feed only knows whether the thumb stopped.

Three facts about the feed

Sound off. Somewhere around 80 to 85 percent of feed ads are watched muted. If your ad only makes sense with the voiceover, most people never find out what it was.
Two seconds. That is roughly what you get to earn a stop. Not to explain, not to build up to the good part. The good part is the opening or it does not exist.
You are not competing with other ads. You are competing with friends, memes and gossip. The bar is not "better than the other meal-prep ad". The bar is "more interesting than what their cousin just posted".

Take Sunday Hour, the meal-prep shop we will use all through this course. Its product is genuinely useful: one tray cooked on Sunday becomes three office lunches. Its first ad was a slow pan over a roasting tray while a voiceover explained the method. Muted, that ad is 15 seconds of anonymous food. Nothing on screen says what it is, who it is for, or why anyone should care within two seconds. Good product, invisible ad.

Same product, same feed, same two seconds. The left ad introduces a brand; the right one makes a promise you can read on mute.

Save six ads

Tonight, scroll like you normally do, but catch yourself in the act. When an ad actually stops your thumb, save a screenshot. When you blow past one, go back and screenshot that too. You want three of each: three that stopped you, three you skipped. Take screenshots, not bookmarks; you will paste these into Claude in the next two lessons.

Then open a free claude.ai chat on your phone and describe them in your own words.

I am learning to judge ads. Here are six ads from my feed today, described in my own words. Ads 1 to 3 stopped me. Ads 4 to 6 I scrolled past.
 
AD 1 (stopped me): [what you saw first, and any text you remember]
AD 2 (stopped me): [same]
AD 3 (stopped me): [same]
AD 4 (skipped): [same]
AD 5 (skipped): [same]
AD 6 (skipped): [same]
 
Tell me:
1. What the three stoppers have in common, using only what I wrote. Quote my own words as evidence.
2. What the three skipped ads have in common, same rule.
3. One sentence on what earns a stop from me specifically.
 
If my descriptions are too thin to find a real pattern, tell me which ads need more detail instead of inventing a pattern.

You should see

Two short pattern lists, each point quoting your own descriptions back at you, then a single sentence about what stops you. If it starts theorising about ads in general instead of your six, reply "only use my six descriptions" and run it again.

Check it

Six screenshots saved on your phone, three stoppers and three skipped
Every pattern Claude names quotes something you actually wrote
Claude asked for more detail if your descriptions were thin, rather than forcing a pattern
You can now say in one sentence what earned the stop, and it is about the first two seconds, not about the brand

If you see

If you think no ad stopped you this week, one did; you just did not register it as an ad. Check your watch history or your saved posts. If truly nothing, screenshot any three posts that stopped you, ad or not; the stopping mechanics are identical. If Claude produces a grand psychological theory from six thin lines, it is guessing: add one more sentence of detail to two of the ads and run it again.

Keep the six screenshots. Next lesson gives you the names for what you are looking at.

What you just learned

The feed is muted, gives you about two seconds, and pits your ad against friends and memes, not other ads.
"The product is good, so the ad will work" is the belief that burns budgets; the feed only measures whether the thumb stopped.
Your own scroll behaviour is data: six screenshots and one prompt turn it into your first pattern.

practice

Open Claude Cowork on your own computer and try this on real files. Each module ends with a checkpoint where you submit what you made.

Open in Claude

The anatomy of an ad