Entity ID Explained: Why Your Ads Are Being Killed Before Anyone Sees Them

You set up the campaign.

You write the copy.

You choose the audience.

You hit publish.

And nothing happens.

No clicks. No conversions. Budget barely moves.

You assume it’s the targeting. Or the offer. Or maybe the creative just isn’t good enough.

But what if the problem happened before any of that?

What if your ad was rejected before a single human being ever saw it? That’s what Entity ID does. And most advertisers have no idea it exists.

What Entity ID Actually Is

Every ad you upload to Meta gets two identifiers.

The first is a Creative ID. This is the unique number Meta assigns to every individual ad. It has always existed. It just means “this is a separate file.”

The second is the Entity ID. This is new.

Entity ID is assigned by an AI that scans your ad. It looks at the image or video and asks: how similar is this to other ads we have already seen?

If your ad looks similar enough to another ad in terms of layout, concept, and visual approach, it gets the same Entity ID as that ad.

And here is the problem.

Meta does not test ads with the same Entity ID against each other. It sees them as essentially the same thing. So it picks one. Or neither. And moves on.

Your ad never gets a fair shot.

Why This Is Different From What Came Before

Before the Andromeda update, Meta was relatively simple about this.

It saw two ads as two ads. Different Creative IDs meant different tests. The algorithm would run both and let user behaviour decide which performed better.

That system had a flaw. Brands figured it out fast.

They would take one ad that worked and make twenty slight variations of it. Change a word here. Swap a colour there. Upload all twenty. Meta would test all twenty because they all had different Creative IDs.

It looked like proper testing. But it wasn’t. It was just the same ad, slightly remixed, over and over.

Users started seeing the same things constantly. Quality dropped. Experience got worse.

Meta’s answer was Entity ID. Now the algorithm groups ads by how similar they actually are. Not just by whether they were uploaded as separate files.

Twenty variations of the same concept now get the same Entity ID. Meta runs one at most. The other nineteen are invisible.

The Part Nobody Talks About

Here is what makes Entity ID more complicated than most people explain.

It is not just about visual similarity.

Meta’s AI is also scanning for conceptual similarity. Two ads can look completely different visually and still get the same Entity ID if they are making the same point in the same way.

Think about that for a moment.

You could change the image entirely. New colours. New layout. New font. And still get the same Entity ID, because the underlying idea is the same.

This is why brands that tried to fix their results by simply refreshing their creative assets are still struggling. They changed how the ad looks. They didn’t change what the ad says or the angle it takes.

Entity ID sees through that.

What Gets a New Entity ID

The good news is that this is actually quite clear once you understand it.

A genuinely different concept gets a new Entity ID.

Not a different background. Not a different headline font. A different angle entirely.

Here are some examples of what that looks like in practice.

An ad that opens with a customer problem gets a different Entity ID from an ad that opens with a product result. Same product. Different entry point. Different Entity ID.

An ad that uses a before-and-after format gets a different Entity ID from an ad that uses a comparison format. Different structure. Different concept. Different Entity ID.

An ad that speaks to one type of customer gets a different Entity ID from an ad that speaks to a different type of customer. Even if the product and the visual style are identical.

The question to ask yourself is simple: is this a different idea, or is this the same idea wearing different clothes?

If it is the same idea, Meta’s algorithm will see that. Even if you cannot.

The Quality Filter On Top of That

Entity ID is not the only thing standing between your ad and your audience.

The Andromeda update also gave Meta’s algorithm a quality check.

Even if your ad gets a unique Entity ID, it still has to pass a standard for user experience. The algorithm asks: is this ad something users will find useful or interesting? Or is it something they will want to scroll past?

Ads that look like AI-generated content, ads that feel jarring or out of place in a feed, ads that make people feel bad about themselves, these get deprioritised. Sometimes completely.

This means you can have a genuinely original concept and still fail to get distribution if the execution is poor.

Originality gets you through the Entity ID filter.

Quality gets you through the experience filter.

You need both.

A Simple Test You Can Run Right Now

Look at your last five ads.

Ask yourself honestly: are these five different ideas? Or are they the same idea presented five different ways?

If you are unsure, that is your answer. They are probably too similar.

Now ask: would I actually stop scrolling if I saw this? Not as someone who made the ad. As someone who had never heard of the brand. Would this earn a second of attention?

If the answer is no, the algorithm’s answer will be the same.

What to Do About It

Stop making variations. Start making concepts.

Before you produce anything, write down the angle. What specific belief, problem or desire is this ad speaking to? What point is it making?

If that sentence sounds the same as the sentence you wrote for your last ad, you have not created something new. You have created a variation.

The goal is to have a portfolio of genuinely different ideas, different enough that a stranger could look at them and immediately see they are making different points.

That is what passes the Entity ID filter.

That is what gets your ads in front of people.

That is what makes your budget actually spend.

…

If this was useful, follow me for more on copywriting, creative strategy, and what actually works in 2026. I am documenting my own journey learning this — the wins, the mistakes, and everything in between.