Why AI Is Making Your Ads Worse (And How to Use It Correctly)

AI was supposed to make advertising easier.

Write the copy faster. Generate the images. Scale the creative. Save hours.

And it does all of those things.

The problem is that everyone else’s AI does the same things. In the same way. Producing the same kind of ads.

And now those ads are failing.

Not because AI is bad. Because most people are using it wrong.

What’s Actually Happening

Meta’s Andromeda update introduced something called Entity ID.

It works like this: Meta’s algorithm now scans every ad before it reaches any user. It checks how similar the ad is to other ads already in the system. If your ad looks or sounds like something that’s been seen before, it gets grouped with those ads and deprioritised.

The result is that your ad never reaches your audience.

Here’s the problem with AI: when millions of brands use the same AI tools, with the same prompts, to produce ads about similar products — the outputs look and sound the same.

Same hooks. Same sentence structure. Same visual style. Same emotional triggers.

Meta’s algorithm sees all of it as the same ad.

Same Entity ID. Same fate.

AI didn’t make ads better at scale. It made sameness at scale.

The Two Ways Brands Get AI Wrong

The first mistake is using AI as a shortcut.

They open ChatGPT and type something like: “Write me 5 Facebook ads for my product.”

The output is technically correct. It has a hook. It has a body. It has a call to action.

But it has no originality. No specific insight about the customer. No angle that hasn’t been used a thousand times before.

It reads like an ad. It doesn’t feel like one.

And Meta’s algorithm and the real humans scrolling through their feeds can tell the difference.

The second mistake is using AI to produce volume, not variety.

Brands take one ad concept and ask AI to generate 20 variations. Different headlines. Different opening lines. Different lengths.

They think they’re testing.

They’re not. They have one idea in 20 different wrappers. Meta assigns them all the same Entity ID. Most never get shown.

Volume is not the same as variety. AI makes volume easy. It doesn’t produce variety on its own. That’s your job.

What AI Is Actually Good At in Advertising

Used correctly, AI saves enormous amounts of time. The problem isn’t the tool. It’s the brief you give it.

Here’s where AI genuinely helps:

Research. Feed AI your customer reviews, sales call transcripts, and competitor complaints. Ask it to summarise the most common frustrations, desires and objections. It will find patterns you’d take hours to spot manually. This is the foundation of good ad copy and AI does it faster than any human.

Drafting from a strong brief. Once you have a specific angle, a genuine insight about your customer that isn’t obvious, AI can turn that into a first draft quickly. The key word is specific. “Write an ad for a busy COO who’s frustrated that their sales AI doesn’t talk to their finance AI” will produce something far more usable than “write a B2B software ad.”

Testing variations of proven concepts. If you have an ad that’s working, a genuine concept with a real angle, AI can help you write variations of the hook or the opening line. This is different from generating variations of the concept itself. The concept comes from you. The execution variations can come from AI.

Editing and sharpening. Paste your copy into AI and ask it to make every sentence shorter. Or ask it to cut anything that a competitor could also say. Or ask it to flag any line that sounds like every other ad. Used as an editor, AI is genuinely valuable.

The Rule That Changes Everything

Before you write a single word of copy or ask AI to write it, you need to answer one question.

What does this person believe right now that is stopping them from solving their problem?

Not what your product does. Not what features it has. Not what pain points your industry typically addresses.

What does this specific person believe right now that is getting in their way?

That belief is your angle. Your ad exists to introduce a new idea that challenges or reframes that belief.

This is original. It cannot be generated by AI without your input. It comes from genuine understanding of your customer, from reading their reviews, listening to their sales calls, watching how they talk about their problems in their own words.

Once you have that, AI can help you write it fast.

Without it, AI will write something that sounds like every other ad in your category.

A Practical Example

Suppose you sell an AI platform for mid-size businesses.

The generic AI prompt: “Write a Facebook ad for a business AI platform that helps companies run more efficiently.”

The output will be something like: “Is your business leaving money on the table? Our AI platform helps companies like yours save time, reduce costs, and scale faster. Book a free demo today.”

Generic. Forgettable. Same Entity ID as a thousand other ads.

Now try this instead. First, do the research. You discover that your customers’ real frustration is this: they’ve already bought multiple AI tools for different departments, and none of them share data. The sales AI doesn’t know what the finance AI is doing. The marketing AI doesn’t know what operations is doing. They have more AI than ever and less clarity than before.

Now you have an insight. Now give that to AI:

“Write a Facebook ad for a COO who has bought Salesforce Einstein, HubSpot AI, and SAP AI, and is frustrated that none of them talk to each other. Their real problem is data silos, not lack of AI tools. Our product replaces all three with one AI that runs across the whole business.”

The output will be completely different. Specific. Pointed. An angle no one else is using.

That’s how AI helps. You supply the insight. AI supplies the speed.

The Simple Test

Before publishing any AI-generated ad, ask yourself three questions.

Could a competitor run this exact ad with just their brand name swapped in?

If yes, the ad has no original angle. Start again.

Does this say something the customer hasn’t already heard a hundred times?

If no, it will blend into the noise. Start again.

Would I stop scrolling if I saw this?

If you’re unsure, assume the algorithm already knows the answer is no.

The Actual Opportunity

Here’s the irony of where we are right now.

AI has made it so easy to produce mediocre ads at scale that the bar for a genuinely original, well-thought-through ad has never been lower.

The brands that still do the human work, the research, the customer insight, the original angle, and then use AI to execute it quickly are performing exceptionally well.

The brands that handed everything to AI and hoped for the best are confused about why their results have collapsed.

The tool didn’t change the game. How you use it did.

AI makes execution faster. It cannot replace thinking.

That’s the work. And right now, very few people are doing it.

…

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.