AI street interview ads

What an AI street interview ad looks like

A presenter holds a handheld microphone and stops a member of the public on a city street or at an event. In the examples here the question appears as an on-screen caption, and some clips cut to a reveal shot of a product or a result.

Both vertical and landscape versions appear. The framing is a two-shot at chest height, with street life moving behind.

Why brands use the format

The question-and-answer shape makes viewers answer the question in their own heads before the person on screen does. That gives the product a reason to appear: it is the answer, or the surprise after the answer.

It also lets a script put an objection in a stranger's mouth and answer it, without a presenter selling directly.

The question-and-answer structure

1. Cold open on the question, spoken by the presenter and captioned. No greeting or intro.
2. The stranger's first answer: a guess, a confession, or a common belief.
3. The presenter's follow-up, which narrows the question toward the product's angle.
4. The reveal: the presenter shows the product, or the stranger tries it.
5. The stranger's reaction, held long enough to read on their face.
6. The presenter's closing line to camera, plus the offer on screen.

For a longer cut, add two or three quick answers from different people between steps 1 and 2.

Writing the question

The question is the hook, so write ten versions and pick the one a viewer could answer in one word. Specific beats general: a question about one habit or one price lands faster than a question about a whole category.

The stranger's lines should sound unscripted: short, a little hesitant, with one filler word. Write them, then cut a third of the words.

Casting the presenter and the stranger

Never base the presenter or stranger on a real person.

Ratio and length

Most street interview ads run as 9:16. Keep both faces and the mic inside the vertical frame by standing the two people close together, angled toward each other.

A single exchange suits a short cut. Several answers, a reveal and a reaction suit a longer cut. Make both from the same shots.

Make it with AI

Start with stills: the presenter and the stranger together on the street, approved before any motion. Use those stills as references to animate each shot with image-to-video, so faces stay consistent through the exchange.

Write the dialogue, generate the two voices, and lip-sync them. Add street sound, a little wind on the mic, and captions for the question in the editor. Model notes are in Veo 3 for ads and Higgsfield for UGC ads.

Compliance checks

This format imitates real passers-by, so disclosure matters more here than in a studio format.

• The people are AI-generated. Do not present them as real customers, real reactions or a real survey.
• Label the ad as AI content wherever the platform requires: see AI ad labels on Meta and TikTok.
• Do not imitate a real presenter, a real show or a real news programme, including its logo or set.
• The stranger's words are still the brand's claims. Keep them within what the brand can support.
• No real businesses, signs or logos in the background other than the brand's own.

Street interview as one of many versions

A tested script maps onto this format by turning its hook into the presenter's question and its main objection into the stranger's answer. The rest of the script becomes the reveal and the close. The UGC script template helps you split a script into those parts, and one winning script made in every format shows how this version sits next to the others.