AI-generated ads at Oddbox: why gut feel still finds the idea

AI-generated ads lose trust fast in food, so Oddbox mainly uses AI to understand why its ads work. It rarely uses AI to make an ad from scratch. Gastón Tourn, Chief Growth Officer at Oddbox, explained the approach at aiblLIVE London 2026.

In food, ads that look AI-made draw hostile comments and lose trust. So Oddbox keeps real customers in its creative and points AI at its ad library. One AI insight, turned into a new message, cut acquisition cost by 17%. Here is how.

Oddbox in numbers

• 17% lower customer acquisition cost after switching one campaign message, in an A/B test.
• 21% higher conversion rate from the same change.
• Less than a day to make Oddbox’s best-performing ad of the year, for the cost of one box.
• About 70% is how ready Tourn estimates AI is to produce ads from scratch.

All figures as stated on stage by Gastón Tourn.

Do consumers trust AI-generated ads?

Far less than ads they believe people made. Tourn cited recent UK consumer sentiment surveys. About 7 in 10 Brits trust ads they think humans made, against about 2 in 10 for AI-made ads. Food brands that ran obvious AI creative drew comments such as “AI slop” and “scam product”.

Whenever you use AI, and it’s clearly AI in your ad, this kind of sentiment shows up immediately.

Gastón Tourn, Chief Growth Officer, Oddbox

How does Oddbox use AI to improve its ads?

It uses creative analysis tools to explain performance. Motion, and Meta’s own creative analysis feature, can say which hooks drive a higher click-through rate across hundreds of ads. They also read thousands of comments and track what competitors are running, which no team could do by hand every week.

What did the back to routine test show?

That an AI insight plus human judgement can shift results within a week. Motion showed competitors running back to routine messages, while Google data showed most converters were not parents. A marketer on Tourn’s team joined the dots. The new message gave 17% lower acquisition cost and 21% higher conversion.

Where does Oddbox use AI in creative production?

Mostly for edits, touch-ups and extra versions. AI cuts several ads from one piece of raw footage. It also brings real field photos of unusual produce, such as purple broccoli, in line with brand guidelines. Oddbox does not generate full ads or fictional characters from scratch, because customer footage keeps winning its A/B tests.

AI will always find the pattern, but gut feel will find the idea.

Gastón Tourn, Chief Growth Officer, Oddbox

Key takeaways

• Analyse before you generate. Use AI to learn why your ads perform.
• Test real people against polish. Simple customer footage beats AI polish in Oddbox’s tests.
• Use AI to edit and multiply. Cut many versions from one shoot.
• Meet customers in person. Oddbox invites them into the office every month.

About the speaker

Gastón Tourn is Chief Growth Officer at Oddbox, the low-waste food company that began with wonky fruit and veg boxes. Oddbox has more than 60,000 customers in the UK. He spoke on the Growth track at aiblLIVE London on 20 October 2026.

What’s in the full case study

If your board is asking why you are not making ads with AI, this is the answer to borrow. The full PDF goes deeper than this page, in Tourn’s own words.

• The evidence on trust: real audience comments and the survey gap between human and AI-made ads.
• How Oddbox’s best ad was made in under a day for the cost of one box.
• The back to routine test, step by step, from AI insight to A/B result.
• Five actions to take back to your own organisation.

For CMOs, growth leads and creative teams under pressure to use AI for ads. Five pages, about a seven-minute read.