What actually makes a web ad work

People have trained themselves not to see ads

In 2007, Nielsen Norman Group ran eye-tracking studies across hundreds of pages and found something blunter than "people ignore irrelevant ads": people almost never fixate on anything that visually reads as an advertisement, whether it's relevant to them or not (Nielsen Norman Group, 2007). This isn't selective filtering based on interest. It's pattern recognition. The eye has learned to identify the shape of "ad" and skip past it, regardless of what's actually in it.

There's a genuinely useful exception buried in that same research, though. Ads that visually resembled the surrounding page, rather than looking like a conventional banner, drew noticeably more attention than ones that announced themselves as advertising (Nielsen Norman Group, 2007). The researchers were careful to flag the ethical problem with pushing this too far: an ad indistinguishable from editorial content blurs a line publishers are supposed to keep clear, and readers who feel misled once tend not to come back. Context helps. Disguise backfires.

This is roughly where HalalSpark's own model plausibly has something going for it, and it's worth stating carefully rather than overselling it. A placement here sits on a specific, manually reviewed publisher page rather than a slot bought in bulk across scraped programmatic inventory. That's not a claim that HalalSpark ads themselves overcome banner blindness; nobody has run that specific eye-tracking study. But a page an advertiser has actually chosen for its relevance is a more native environment for an ad to sit in than an arbitrary purchased impression, and the research above says native context is precisely what banner blindness responds to. Reasonable inference, not a proven result.

You get about fifty milliseconds, so don't make the eye work

A separate, often-cited study measured how quickly people form a judgement about a design's visual appeal: about fifty milliseconds, well before anyone has consciously read a word (Lindgaard, Fernandes, Dudek, & Brown, 2006). An ad has less room to work with than a full page, and it's competing with the actual content around it for that same sliver of attention. If the eye has to hunt for what matters, that hunt doesn't happen. The ad just gets skipped, the same way a banner-shaped ad gets skipped.

Once an ad has been noticed, where the eye goes next isn't left to chance either. A 2017 eye-tracking study on banner ads specifically found that when a pictured face looked toward the product rather than out at the viewer, people were roughly four times more likely to actually look at the product itself, and reported a higher intention to buy it afterwards (Palcu, Sudkamp, & Florack, 2017). Gaze direction works like a pointer. If there's a face in the creative, it should be looking at whatever you want seen, not at the camera.

Both findings point to the same practical rule: one focal point, one path for the eye to follow. A headline, a face, a product shot pulling attention three different directions doesn't triple the impact. It just gives the fifty-millisecond window nothing clear to land on.

One button, doing one job

Every ad needs exactly one call to action, and it needs to look like one. Enough contrast and enough space around it that it reads as clickable at a glance, not as another line of text. The wording matters too: an instruction beats a description. "Start advertising" tells someone what happens when they click. "Learn more" tells them nothing and asks them to guess.

Two CTAs competing for the same click isn't twice the opportunity. It's a decision the ad is asking the viewer to make in under a second, and most people won't bother.

Design for the page you don't control

A web ad rarely renders on a blank canvas. It sits inside someone else's layout, next to someone else's colour palette, on a screen size you can't predict, and increasingly, on a phone held at arm's length. Contrast that would look fine on a white background can disappear against a busy publisher page. Text sized for a desktop monitor can turn illegible once it's scaled down to a phone-width banner.

None of this is exotic advice. Strong contrast between text and background, type large enough to hold up at mobile sizes, and enough breathing room that nothing gets clipped when the creative is resized. The ad that survives being shrunk and placed somewhere unpredictable is the one still doing its job on the tenth publisher site, not just the one it looked good on in the design file.

One idea, not three

Attention is scarce before an ad even gets read, and scarcer still once it has. An ad trying to land a discount, a brand tagline, and a feature list all in the same few square inches usually lands none of them. Pick the one thing that actually matters for this particular ad and say only that. Everything else belongs on the landing page, where there's room for it.

What a normal click-through rate actually looks like

It's worth knowing what "working" actually looks like before judging a campaign against it. WordStream's 2026 display and video benchmark report, built from over 60,000 real advertising campaigns, puts average display ad click-through rates somewhere between roughly 0.15% and 0.3% depending on the industry (WordStream, 2026). That's two or three clicks for every thousand people who see the ad.

That's not a disappointing number. It's the normal one. Display's job, per the research above, is mostly getting seen and half-remembered, not producing an instant click. A campaign sitting inside that range is doing exactly what display advertising is supposed to do.

Skip the generic stock photography too, along with any cartoon or obviously AI-generated imagery. It reads as filler rather than as a real business, and it undercuts trust before anyone's even read the headline. It's also a rule HalalSpark's own brand guidelines apply to its own marketing, for exactly the same reason: manufactured-looking imagery erodes the trust a halal-audience-facing brand depends on. A small, honest coincidence, not a hard sell.

If there's one place all of this comes together, it's a page an advertiser actually chose for its audience, not an arbitrary slot in a programmatic auction.

None of this is complicated, and none of it is platform-specific. Get noticed despite banner blindness, give the eye one clear thing to do once it has, make the ad survive a screen and a background you don't control, say one thing, and judge the result against what display actually delivers, not what a search ad promises.

Insha'Allah it serves you well.