If Every Social Ad Looks the Same, Who Remembers the Brand?

Social ads can fit the feed so closely that the brand disappears. Zappi's analysis of more than 3,300 videos found creator-style ads were 30% less likely to be top performers, while using three or more distinctive brand assets made ads 30% more likely to perform well. Strong hooks still mattered, but recognisability mattered too.

That finding lands at an awkward moment for marketers. Instagram is getting better at telling creators why one post beat another and suggesting the next hook, script or caption, while agencies are being pushed to make more content across more channels. The industry already has plenty of optimisation; the concern is whether the same advice starts producing the same-looking work.

Fitting the feed has a cost

Zappi analysed more than 3,300 social video ads across TikTok, Instagram, Facebook and YouTube. Ads designed to resemble creator content were 30% less likely to rank among the top performers in the study and had the weakest brand connection.¹ Strong visual hooks still helped, with those ads 2.4 times more likely to be top performers.

The more useful finding was what happened when the brand stayed visible. Ads using three or more distinctive brand assets were 30% more likely to perform well and produced the strongest brand connection.¹ That gives marketers a better brief than the familiar instruction to make every advert look native to the platform.

The creator look became popular for a sensible reason. People scroll past anything that feels stiff, over-produced or obviously transplanted from another medium. Marketers responded by borrowing the grammar of the feed: handheld video, direct-to-camera delivery, looser editing, conversational captions and formats associated with individual creators.

A format can become so familiar that the business inside it disappears. If the colours, language, product cues and recurring details that make a brand recognisable are stripped away in the name of looking native, the advert may earn a few more seconds of attention while leaving less behind. Being easy to watch and being easy to remember are different jobs.

That distinction matters more as AI lowers the cost of producing variations. A team can now generate dozens of versions of the same idea with different hooks, openings and visual treatments. If the system optimises only for what holds attention in the moment, it can make the content more platform-shaped while making the brand less visible.

Instagram can now teach the pattern

Instagram's new AI creative assistant inside Edits can use likes, views, retention and shares from a creator's account to explain why one video performed better than another. It can also suggest hooks, scripts, captions, audio and future content ideas based on those signals.² For a small team, that can turn performance data into usable guidance without needing an analyst to unpack it.

The capability is genuinely useful. A restaurant owner can learn that a food close-up held viewers longer than an exterior shot. A salon can see that a faster opening kept more people watching. A small fashion brand can identify which product angle earned more shares instead of relying on instinct alone.

The tension starts when a useful observation becomes a repeated formula. If Instagram tells thousands of accounts that a particular hook length, edit pattern or audio treatment improves retention, many of them will follow the advice. The recommendation can be individually sensible while collectively making the feed more uniform.

Platforms already shape creative behaviour because distribution rewards some formats more than others. Adding an AI assistant makes that feedback loop faster. The platform can observe what performs, explain the pattern and recommend the next version, all inside the same system that will later decide how widely the content travels.

For marketers, that creates a new discipline: separate the lesson from the imitation. "People stayed longer when the product appeared in the first two seconds" is a useful lesson. "Make every video look like the same high-performing creator format" is imitation. The first can sharpen a brand's content; the second can slowly erase its identity.

Native does not mean anonymous

This is where the Zappi result becomes practical. Three or more distinctive brand assets improved the odds of strong performance in its study, which suggests that brands do not need to choose between fitting the platform and remaining recognisable.¹ The better target is content that belongs in the feed but still carries obvious signs of who made it.

Those signs do not need to be a large logo in the corner. They can be a recurring colour combination, a familiar phrase, a product shown from the same useful angle, a recognisable setting, a particular style of demonstration or the way the business talks about its customers. Distinctiveness is often built from small repeated cues rather than one formal brand element.

This matters especially for smaller businesses because they rarely win by outproducing large advertisers. Their advantage can be that customers recognise the owner, the room, the product style or the voice before they consciously process the account name. AI content marketing works better when it strengthens those cues instead of replacing them with whatever visual style is currently common across the feed.

That is also the logic behind Asteris.ai: start with a business's own photos, products and brand cues, then use AI to help shape that material into Instagram content rather than asking a model to invent a generic identity from scratch. The software has more useful material to work with because the business supplies the distinctive ingredients. The result can still follow the conventions of Instagram without pretending that every business should sound or look alike.

For a small business, this is good news. It does not need to imitate every creator trend or invent a completely new visual language every week. A recognisable colour, phrase, product detail, setting or point of view can do more long-term work than another generic talking-head video.

The commercial reason is straightforward. Social advertising has to do more than win the next second of attention; it also has to leave enough memory for the customer to recognise the business later. If the format earns the view but the brand cues vanish, the team can end up optimising the easiest metric to measure while weakening the thing the campaign was meant to build.

The same principle applies to Instagram AI content management more broadly. A useful system should know which parts of a post can vary and which recurring cues deserve protection. If every high-performing suggestion pushes the content further towards the average, the tool is improving the individual post while weakening the account over time.

More content makes sameness cheaper

The pressure to produce more is not going away. Digiday's agency research found that 75% of respondents saw creator marketing as more important than a year earlier, while 89% reported increased client spending in AI search.³ Marketing teams are being asked to cover more formats, more surfaces and more discovery systems at the same time.

AI is already making that volume easier to produce. Wyndham Hotels & Resorts told Digiday that a five-person marketing team was producing 15 times as many assets with AI, while cutting production and approval time by 75%.⁴ Opella said its team had produced more than 20,000 pieces of content, including localised variants and animated brand assets.

Those numbers are impressive, but they make brand discipline more important, not less. One forgettable post disappears quickly. Fifty forgettable posts can teach an audience that the account has no particular reason to be remembered.

Four Pillars Gin offers an interesting counterexample from the same week. Its latest campaign was deliberately made by hand, with physical sets and an unbroken camera shot, because the craft of making the advert supported the craft story of the product itself.⁵ The production choice became part of what the campaign was saying about the brand.

That does not mean every brand should reject AI or start building physical sets. Wyndham needs large numbers of variations built from existing hotel imagery, so AI can be useful. Four Pillars wanted the making of the advert to reinforce the thing the brand claims to value, so a handmade execution made sense.

Scale is useful when the brand survives the scaling. Once the distinctive parts disappear, producing more simply distributes the same anonymity more efficiently. A fifteen-fold increase in output only helps if the additional content still gives people a reason to know who it came from.

What should social teams protect?

Performance data still matters. Social teams should use it aggressively, especially when AI can surface patterns that would otherwise take hours to find. Every optimisation should also be checked against a second question: does this still look and sound recognisably ours?

That check can be simple. Look at the last five posts without the account name or logo visible. If a customer who already knows the business would struggle to identify who made them, the problem is unlikely to be posting frequency or hook length.

For teams using AI captions for Instagram business posts, the same test applies to language. A model can produce a clean caption in seconds, but "clean" is not a brand characteristic. The useful details are the words the business uses repeatedly, what it refuses to say, how formal it sounds, what it notices about its customers and which claims it can genuinely support.

The more recommendation systems influence creative decisions, the more valuable those constraints become. Instagram can tell a business that a shorter opening retained more viewers. It cannot decide whether repeating the same opening for the next 30 posts will make the account feel monotonous.

Using performance data without outsourcing identity is the useful middle ground. It allows the platform to teach the business something about attention while leaving the business responsible for what people should remember. That distinction will matter more as recommendation tools become more capable.

Remember the brand, not the format

The Zappi study should not be read as a call to make social ads that feel like television commercials. Strong hooks still mattered, and content still needs to work in the environment where people encounter it.¹ The useful lesson is narrower: copying the look of creator content is not a substitute for giving people something distinctive to remember.

Instagram's new AI assistant makes that lesson more urgent because the feedback cycle is getting faster. Marketers can learn more quickly, make more versions and respond to performance with less manual work. That is valuable, provided the system is helping the brand become more itself rather than more like the statistical average of the feed.

The next wave of social marketing will probably contain far more AI-assisted content than the last one. Volume will be easy. Pattern-matching will be easy. The brands that remain recognisable will be the ones that know which patterns to borrow and which parts of themselves never to hand over.

If every social ad starts to look the same, attention alone will not rescue the brand. People still need something specific to remember. That is the part no optimisation metric should be allowed to erase.

Sources