Guide
Yes. That's the easy part of the answer, and it's also the part that gets people in trouble — because a face that wins attention can spend all of it on itself.
Humans are unusually good at finding faces. We find them in electrical outlets and cloud formations, we spot them in peripheral vision, and we lock onto them faster than almost any other category of thing. Eye-tracking research on advertising has found this over and over: put a face in an ad and the eye goes there, quickly, often before the viewer has consciously decided to look at anything. In the context of banner ads — where people have trained themselves into near-total blindness — studies have found that faces help cut through that learned skipping.
So the headline is true. The reason it's not very useful on its own is that "the eye went there" is a much smaller claim than the one people build campaigns on. Getting looked at is the price of entry, not the win. It's the same distinction behind what makes an ad scroll-stopping: stopping someone and holding them are two different mechanics, and only one of them is solved by a face.
Researchers have a name for what happens next, and it's unflattering: the vampire effect — when the face drains attention away from the product instead of toward it. The mechanism is straightforward. You have a finite amount of viewer attention. A face is the most magnetic object you can put in the frame. If you put a striking one next to your product, the face wins, and the product becomes set dressing behind a person people are busy looking at.
This is why "we hired an attractive creator and the ad still flopped" is such a common story. The ad worked exactly as designed — it captured attention — and then spent every bit of it on a face rather than on a reason to buy. The face was never the problem. The assumption that attention automatically flows through the face to the product was.
The most actionable thing in this literature is gaze cuing. We don't just look at faces — we look where faces look. Studies on gaze direction in advertising have found that a model whose gaze is turned toward the product pulls more attention onto that product than one staring down the lens, and that averted gaze can lift attention to the ad's text and product rather than just parking it on the person.
That converts a vague rule ("use faces") into a real edit. If you need the viewer on the person — a founder making a claim, a creator delivering a line — direct-to-camera is correct, because the face is the message. If you need the viewer on the product, have the person look at it, hold it, react to it. The face buys the glance; the gaze spends it. Which of those you want depends on what the shot is for, and it can change between the opening and the demo inside the same ad.
A face in second 14 of your ad is a composition choice. A face in the still that represents your ad, or in the opening moment, is a gate — it's doing the work described in first frame vs first three seconds, and it's the most reliable lever you have there. This is a large part of why UGC-style creative tends to earn attention on feed surfaces: it opens on a human, close, talking, in a room that looks like a room. It matches what the surrounding organic content looks like, and it exploits the face reflex before the viewer has classified the video as an ad.
The corollary is that a face in the opening is not a free pass. It buys you a second or two of attention that you then have to earn the rest of with something to say — which is the whole argument of the testimonial ad format: front-load a specific claim, because the face got you the hearing and now the content has to keep it.
Everything above is about where eyes go. None of it is about whether anyone buys. A face can win the glance and lose the sale; an ad can be looked at intently and still fail because the offer is weak or the product never registered. Those are different questions, measured differently, and conflating them is how "faces increase attention" mutates into "faces increase sales" on a slide somewhere.
The practical move is to stop treating it as a rule and start treating it as a variant. Make the cut with the face and the cut without. Make the version where she looks at the camera and the version where she looks at the product. PreTestAds predicts how each cut holds attention across its runtime and ranks it against a benchmark of top-performing short-form ads, so you can see which framing actually held before you pay to find out — the same discipline as any other ad pre-testing decision. It's a screen, not an oracle: it reads the creative, not the offer, and live results still get the final say.
Upload both cuts and compare predicted attention before you spend — first analysis free.
Score Your AdYes — eye-tracking research consistently finds that a human face draws the eye quickly and reliably, which is why faces are effective at overcoming the habit of skipping past ads. But attention to the face is not the same as attention to your product or message. Researchers describe a 'vampire effect', where a compelling face absorbs the attention an ad needed to spend elsewhere.
Eye-tracking studies on gaze cuing find that viewers tend to follow the direction a face is looking, and that a model gazing toward the product draws more attention to that product than a model looking straight at the camera. The practical takeaway: a face is a pointer. Direct-to-camera gaze holds the viewer on the face, while a glance toward the thing you're selling passes attention along to it.
It makes an ad more likely to be looked at, which is a necessary condition for performing, not a guarantee of it. A face can win the glance and still lose the sale if the product never registers or the offer is weak. Treat 'add a face' as a hypothesis to test rather than a rule — score the version with the face against the version without it and let the comparison decide.