Is your paid social media agency making every Ad look the same?

AI tools built into platforms like Meta and Google can now generate ad copy, images and video in minutes, and every paid social media agency can use them. This article looks at what the evidence says about how AI-generated ads perform, why sameness is a bigger risk than poor quality, and what the psychology of attention suggests brands should do about it. A May 2026 Ipsos study of 3,000 consumers found ads created by real people outperformed AI-made equivalents, even though most viewers couldn't tell which was which. We also set out the questions worth asking a paid social media agency about how it uses AI.

The ads are getting easier. The attention isn't.

Meta and Google both now offer tools that generate ad creative from a product image or a landing page. Reporting suggests Meta is working towards letting advertisers state a goal and a budget and have the platform handle the rest, although we couldn't confirm how far that has rolled out, and it's worth checking current platform documentation.

The appeal is obvious. Creative is expensive and slow. A tool that produces a dozen variations before lunch looks like a gift.

The catch is what those tools are built from. Generative AI works by predicting what's most likely. Applied to advertising, that means the average of what already exists: the familiar framing, the standard claims, the usual tone.

But being seen isn't the same as being remembered, and when thousands of brands feed similar prompts into similar tools, they get similar results.

What the evidence says

Creativity still wins, even when people can't tell

In May 2026, Ipsos and Syracuse University's Newhouse School published a study of 20 ads from major brands, ten made by people and ten recreated by AI from the same strategy, tested with 3,000 US consumers.

The findings:

• Ads created by people were 14% stronger on short-term effectiveness and 17% stronger on long-term brand health, on average.

• 25% of people who saw an AI ad were at least somewhat confident it was made by AI. About 40% of all viewers weren't sure either way.

• Ads created by people were rated more entertaining, unique and talkable.

• AI did best on direct, product-led ads and struggled most with emotion, narrative and creative leaps.

Ipsos summed it up as: credible is not the same as compelling. The AI ads weren't bad. They just weren't memorable.

Our subconscious notices, even if we aren't aware

An earlier NIQ study combining EEG brain measurement on around 150 participants with a survey of more than 2,000, found that even AI ads viewers rated as high quality produced weaker memory activity than traditional ads. Viewers also tended to describe AI ads as "annoying", "boring" and "confusing", which NIQ suggested could affect how they feel about the brand itself.

People are losing patience

Merriam-Webster named 'slop' its 2025 word of the year, defining it as digital content of low quality produced in quantity by AI. In December 2025, McDonald's removed an AI-generated Christmas ad in the Netherlands after criticism online that it felt soulless. The campaign had been made with AI specialists and a production company over several weeks, which is a reminder that spending effort on AI doesn't guarantee that audiences will accept the result.

Why sameness is the real risk

Attention in a social feed is a contest of distinctiveness. People scroll fast and forget faster. What stops them is something that looks, sounds or feels different enough to register before they've decided whether to look.

Brands are remembered through distinctive cues: a recognisable character, a colour, a voice, a way of framing a joke. When ads become interchangeable, those cues disappear. You aren't just risking a weaker ad. You're losing the thing that made the next ad cheaper to run, because people already knew who you were.

A fair counterpoint

None of this means AI has no place in paid social.

The same Ipsos study found AI performed best on direct, product-led work. It is genuinely useful for resizing and adapting existing assets, producing variations to test, speeding up production and drafting early routes. That's where the savings are real.

It also has limits worth stating plainly. The Ipsos study tested AI ads made with no human creative input, so it says little about work where a skilled person directs and edits the output. It covered 20 ads and US consumers. One study doesn't settle the question, and the technology is changing quickly.

The sensible reading is a division of labour: AI for execution and scale, real people for the idea, the brand and the story.

What to ask a paid social media agency about AI

1. Where do you use AI, and where don't you? A confident answer names both.

2. Who owns the idea? The concept should come from strategy and brand, with AI used to carry it out.

3. How do you stop our ads looking like everyone else's? Look for specifics, such as distinctive assets, a defined voice and a consistent visual identity.

4. How do you test AI-assisted work against human-made work? If they can't answer, they're guessing.

5. How do you handle disclosure and brand safety? Audiences increasingly expect honesty about AI, and a poorly judged execution can hurt the brand.

6. What happens when the ad performs but the brand suffers? Good agencies track brand and quality signals, not just clicks.

If an agency's pitch is mainly about how many ads it can produce, it may be selling volume when what you really need is memorability.

Related reading: What Does a PPC Management Agency Actually Do?

Explore our Paid Social and Brand Development services

Paul McMillan is Marketing Executive at Ouma, a strategic growth partner helping UK businesses build connected growth systems. Paul works as part of Ouma’s senior-led team across digital and performance marketing.