Meta Creative Enhancements, Which Ones to Kill and Which Ones You Can Leave On

There's one thing that comes up every single time we take over a Meta account for a brand. We launch the ads, we get the client to sign off on the visuals, everybody's happy. Then two days later the client messages me, slightly panicked, because they just saw their own ad in their feed with a text banner they never approved, music slapped on top, and a crop that chops half the product out of frame.

Meta applies its creative enhancements behind your back, without telling you. And it's by design, it's not a bug on their end. And there are a lot of them. A Jon Loomer article counts 47, and he admits he's probably forgetting some. 47. On a single ad.

Most people's instinct is to turn everything off on principle. I get the frustration, but it's a mistake. Some of these options are useful, some are dangerous, and the real question is knowing which ones you leave on and why. Here's how I sort them.

What these creative enhancements actually are

Basically, Meta takes your visual and gives itself the right to modify it. It can add a text banner, slap on music, generate new AI variants of your image, translate your headline, crop it, tweak the brightness, add product tags, overlay highlights pulled from your site.

The problem is that these settings are scattered all over the creation flow. Some are in Extensions right at the start, just after upload. Some are in the Advantage+ generation section. Some are buried in a block of text you can read ten times without noticing, like Personalized Destinations. And some are on by default without you asking for anything.

So you've got a choice. Either you dig into each setting, or you let Meta decide for you what your audience sees. Both are a pain, but the first is the only serious option.

The stuff I kill almost every time

There's a category of enhancements that doesn't touch your visual but your destination. And for me, that's a hard no.

• First thing I kill, Optimize Website Destination. Meta reserves the right to send people to a different page on your site than the one you chose, because it figures they'll convert better somewhere else. I get the intent, but when I put a URL in an ad, I want people to land on that URL. Otherwise the starting point of your funnel is already distorted. I leave it off.
• Same logic with Product Browsing. Instead of sending to your site, Meta redirects to a browsing experience inside Facebook or Instagram. Until someone proves to me it converts better, I keep control.
• Then there's Browser Add-Ons. Meta constantly wants to stick an instant form in the in-app browser, even when you're sending to a real landing page. I see the value on high ticket where you need to qualify, but by default I remove it.

The other thing I kill without hesitation is video generation when it stages fake humans. An AI building you a character who talks about your brand, for a fashion or lifestyle brand that cares about its image, is just a no. On the other hand, when Meta just animates your static photo, adds a bit of subtle motion to catch the eye, then yes, that's worth it. It's kind of what we do ourselves when we take a static visual and turn it into subtle motion, just enough to catch attention in the feed without going overboard.

The stuff I leave running (and why)

On the flip side, there are enhancements I leave running because they actually deliver.

Text generation, for example. At first it was a disaster, the AI spat out headlines that made no sense. Today it's become acceptable, not always, but acceptable. And more importantly it creates message angle variety for you, for free. And variety is exactly what the Meta algorithm loves. When you stop duplicating your creatives and feed the account different concepts, you go after performance you wouldn't otherwise reach. So I leave it on, I keep what's good, I throw out the rest.

Image generation, same logic. It's gone from "horrible" to "sometimes decent." The variants I accept are the ones that stay close to the original. Two or three clean takes, that's more material to test.

Music is the most divisive setting. Most advertisers kill it by reflex. I often leave it on, because on placements where sound matters (Reels, Stories), a silent ad goes unnoticed. The thing is, you can pick the track yourself instead of letting Meta grab one at random. You keep control of the tone.

Website Summaries and Site Links, same deal, I leave them on but I customize them. If you let Meta decide on its own which links to feature, you're giving it way too much trust. Take ten minutes, pick the right links, the right labels, and you're done.

One setting everyone skips that actually changes everything is Branding. You fill in your logo, your fonts, your colors, your visual style, and even the words you ban the AI from using. That one you don't turn off, and you need to spend time on it. Because every enhancement that generates text or visuals leans on it. If your Branding is empty, Meta improvises. If your Branding is tight, the damage is already limited.

Where it really gets complicated, your reporting

Here's the point nobody looks at and the one I think matters most. When Meta modifies your creatives, it distorts your read.

Picture this. You launch three variants of the same visual to test an angle. One crushes it. You think great, the angle works. Except on that specific variant, Meta added music, an overlay, and cropped the image differently. You can't tell whether it's the angle, the music or the crop that drove performance. You've got plenty of different ads but no clean learning out of any of them.

That's exactly the problem when you want to structure your creative testing seriously. If Meta is tinkering with your variants behind your back, your read of what works is rotten at the foundation. You think you're measuring your concept, you're actually measuring a mix of your concept plus the algorithm's retouching.

My rule here is simple. When I'm in a testing phase, I want to understand what works, so I lock it down. I kill the enhancements that modify the visual to isolate the real variable, the creative angle. Once I've got my winner, once I know the angle holds up, then I can reopen some enhancements in scaling because I no longer need to learn, just to distribute wide.

And without fail, after launch, I go check the actual render. You take your ad, you look at the preview in each placement, and you verify Meta didn't slap something on that you never approved. It's two minutes and it saves you from explaining to a client why their ad has a "terms" banner written sideways across it.

My line, to sum up

It all depends on what you're after at that moment, learning or distributing wide.

In the testing phase, I lock down as much as possible. I kill everything that modifies the visual, because I want a clean read on my angle. I only keep text generation to vary the messaging, and even then, knowing it's one more variable.

In the scaling phase, I reopen. Music, image generation, overlays. There I don't need to learn, I want volume and variety, and that's where these enhancements become allies.

And in every case, Branding filled out properly and a preview check after launch. That's the bare minimum to not let Meta decide for you what your audience sees.

For the rest, judge it case by case. At least you're no longer cutting blind, you know why you're doing it.