What Happened to My Reach When I Let AI Build the Ad

While you build an ad, the platform gives you an estimate of how many people it may potentially reach.

When I created the ad manually, the projected audience was one number.

When I started accepting the platform’s AI suggestions, that number began to grow.

I added AI assistance with the copy — the potential audience increased.

I allowed the system to work on the visual side — it increased again.

I added more automated options — and the estimate changed again.

At some point, the difference between my original manual version and the fully optimized version became so large that I simply stopped and looked at the numbers again.

This was not a small change.

The projected reach had moved from a relatively modest number of people to figures around one and a half million.

And that raised a very simple question.

Why?

That is important to say immediately.

One personal experience is not enough to conclude:

“The platform prefers AI content and rewards it with greater reach.”

I do not know that.

And that is exactly why this is a Field Note rather than a research conclusion.

There may be a much more practical explanation.

When the system automatically improves an advertisement, it may also make it suitable for more formats and more delivery opportunities.

For example, it may:

• create multiple image variations;
• adapt the creative for different placements;
• generate several versions of the copy;
• categorize the ad more accurately;
• broaden automated audience options;
• create more combinations for the delivery system to test.

So perhaps the platform is not saying:

“AI was used here, so we will show this to more people.”

Perhaps it is saying:

“This ad is now suitable for more situations in which we can deliver it.”

That is a very different thing.

But from the outside, the result looks almost identical:

the more of the platform’s AI tools I enabled, the larger the projected reach became.

And that is what caught my attention.

Because the next question appeared immediately.

If something like this happens in advertising, what happens in other digital ecosystems?

For example, music.

Today we are increasingly asked to disclose whether artificial intelligence tools were involved in the creation of a work.

AI may have been used in composition.

Arrangement.

Individual instruments.

Audio reconstruction.

Mixing.

Mastering.

Visual content around a release.

But what happens to that information after it enters the platform?

Is it simply displayed as part of the credits?

Is it used only for transparency?

Or does it become another signal inside the platform’s internal systems?

I do not know yet.

And that is exactly why I want to look at it more closely.

When we prepared the release of When the Soul Aches, I spent a great deal of time on the credits.

We even discussed the wording with a lawyer because I wanted to separate human authorship, original performances and AI-assisted production as clearly as possible.

The first reason for doing that was transparency.

I believe people should be told honestly who did what.

But now I have another reason to pay attention.

I want to observe what happens to this information after it enters the system.

How does the platform interpret it?

What metadata is passed onward?

What does the listener actually see?

Does the system distinguish between AI-generated and AI-assisted work?

And most importantly:

does any of this information affect how the content is classified or distributed?

At this point, I do not have an answer.

But I do have something I can observe.

This may be the most important point.

Suppose an AI-optimized ad really does receive greater reach.

That still does not mean the platform rewards it because AI was used.

Perhaps AI simply wrote better copy.

Perhaps it identified the audience more effectively.

Perhaps it created more usable formats.

Perhaps it improved the image.

Perhaps it made the ad eligible for more placements.

In that case, the advantage does not belong to “AI content.”

The advantage belongs to better-prepared content.

AI simply happened to be the tool that helped prepare it.

For independent artists, that distinction matters.

A lot.

One case is an observation.

Several controlled cases could become the beginning of a real study.

So I would like to repeat the process with the same basic advertisement and create several versions.

One made entirely by hand.

One using AI-generated copy.

One using AI-assisted visuals.

One using full automated optimization.

At the same time, I would keep the budget, audience, location, campaign objective and duration as similar as possible.

Then I would compare two things:

what the platform predicted before launch

and

what actually happened after launch.

Because projected reach and real reach are not the same thing either.

That is where the question becomes much more interesting.

I constantly see a strange fear around the sentence:

“AI was used here.”

Some people try not to say it.

Some are afraid of the label itself.

Some almost treat an AI disclosure as a warning that the work must somehow be inferior.

At the same time, the platforms themselves increasingly tell us:

Let AI improve your text.

Let AI improve your image.

Let AI create a video.

Let AI find your audience.

Let AI optimize your campaign.

And that creates a curious contradiction.

On one side, users are being asked to disclose their use of artificial intelligence.

On the other, the digital infrastructure itself is becoming increasingly dependent on it.

So perhaps the more important question in the near future will no longer be:

“Did you use AI?”

but instead:

“What exactly did AI do — and what happened to your content after that?”

I do not have a final conclusion yet.

For now, I only have an observation.

But sometimes that is exactly where the most interesting research begins.

— Alex Kryve