Did Anyone Ask:
Is This Good?

Talent + Tech is Closing The Uncanny Valley

Today, skilled filmmakers using the best models are clearing the uncanny valley easier than in the past. That means more realistic content that’s convincing, which is a positive for craft.

The phrase Uncanny Valley is used frequently now but its original use dates back to 1970, when Japanese roboticist Masahiro Mori proposed that as a humanoid figure’s realism increases, our comfort and affinity towards it rises along with it. But only until it gets almost but not quite lifelike, at which point that comfort drops off a cliff into unease, before recovering again as realism approaches true human likeness.

In other words, something that imitates a human without quite being one breaks our trust. We feel deceived before we can say why, and once we feel that way we usually stop watching.

Coca-Cola

Do you remember the 2024 holiday ad Coca-Cola released? The one with the convoy of trucks, the polar bears, the glimpse of Santa’s sleeve handing over a bottle. The creative community crucified the spot and it became advertising’s go-to example of bad AI. A year later, they ran the sequel and their own head of generative AI called the new craftsmanship “10x better than the original,” about as close to an admission of mistake that we saw.

When System1 tested the original 2024 ad with consumers, they loved it. It scored 5.9 stars, the top of System1’s scale. But social sentiment on the 2025 sequel told a different story: By late November, per Sprout Social’s tracking, sentiment on Coca-Cola’s spot sat at 37% positive.

Although their AI craftsmanship actually improved between the 2024 and 2025 ads, sentiment went down. In 2025, McDonalds also released a holiday ad, which didn’t go well and they pulled the spot within 3 days of release.

Failing Loudly and Quietly

A recent IPSOS study observed 3,000 U.S. consumers as they watched 20 ads from major brands, half made by people and half made entirely with AI from the same creative briefs. 25% of consumers were able to identify the AI spots, 75% could not.

The AI-generated ads scored 14% weaker on short-term sales potential and 17% weaker on long-term brand equity than the human-made ones.

Kantar found the same pattern across more than 350 ads that used generative AI in some way. The ones where the AI use was obvious scored poorly on branding specifically.

System1 alongside Peter Field and Adam Morgan found that flat, uninteresting advertising needs somewhere between 2-2.5x the media spend to match the market impact of advertising that makes people feel something. Nielsen’s analysis of nearly 500 campaigns say creative is responsible for 47% of total sales contribution, ahead of reach at 22% and targeting at 9%.

To ground this in real numbers, let’s use a $2M media spend as an example. If you land a 4 or 5 star ad then that spend is worth two to three points of long-term market share growth. Land a 1 or 2 star ad and that same $2M is worth half a point, or nothing.

Looking at it through Field and Morgan’s “cost of dull” means spending $4-5mm on media for the same impact to achieve what $2mm could with good creative.

The Room Agrees with Itself

We recently wrote about how AI Pro-Viz, our AI pitch offering, is helpful for closing the imagination gap when selling ideas. Talented filmmakers can make a video good enough to fool almost anyone; including the people making it. It’s easier now to mistake convincing for persuasive.

At times it seems the AI Yes Machine is capable of achieving unearned agreement, where no one in the room says no. How can we be less deceived or improve our judgement? The teams who put Coca Cola and McDonalds out thought their work was pretty good also, until consumers saw it.

Was there a lapse in judgement? Did they forget that advertising is also about moving hearts and minds, not just craft? Was the new shiny tool a distraction from the larger objective?

What this Means for In-House Agencies

Brands are increasingly being handed an AI mandate and it usually arrives as a number and nothing else like, spend a percentage or a dollar figure on AI. Does a frontier model count? Does hiring a vendor who says they use AI count? It’s often not clear.

The economic benefit for the business is obvious: make more content faster. This objective makes financial sense as long as storytelling and brand equity are not sacrificed.

So it becomes the responsibility of the in-house agency to determine how and where to use AI. The wise teams will spend time defining this and creating a new process for which to create, produce and judge their work.

Know Your Methodology

As teams prepare to make more work with AI, building a methodology for making the work is critical.

Usually a short R&D phase at the beginning of a project or new workflow pays off. Making multiple spots, testing with a small spend and amplifying the best messages is smart business. The tools exist and can often be done in-house.

Consider a walk before you run approach as well. Use AI film for internal work streams where there is more control and less risk if something doesn’t go as planned. Remember a hybrid approach to AI using existing photography and video assets removes some of the technical challenges and brand risks.

Lay out your concerns and your hopes for your AI work stream so staff and partners can help address them.

Determine how to make judgments on storytelling, creative and impact.

Mind the Trap

Every gain AI hands us comes with a new question attached. That’s the trap. AI can fool a room that what it built is persuasive, when all it’s actually built is convincing. Sometimes the room is fooled by an agreement that wasn’t actually earned.

Without a critic willing to say otherwise, that gap becomes a brand problem.

AI for AI sake is not the goal. The goal was always the harder one: faster and cheaper, without losing the way into someone’s heart. Coca-Cola solved the first half. But the second half is the one their holiday ads never quite delivered.

What moves people is what’s always moved people: story, and the feeling it leaves behind.