Algorithms are fleeting, so while it used to be possible to hope for isolated “hits” and random success, now it makes no sense. Video advertising on TikTok and Reels was the first to show that it is beneficial for the platform’s feed if videos can hold the viewer’s attention for at least a few seconds and keep them there longer.
The coming year 2026 is about radically testing formats that seemed too bold yesterday, and about creativity finally becoming the main lever, rather than an addition to targeting.
How the algorithm reads ads
The systemic restructuring of video advertising is far from a single update. Initially, TikTok and Meta’s advertising algorithms were based on fairly simple logic:
- more signals → better targeting;
- optimization for individual metrics → stable results;
- more budget → faster learning.
By mid-2025, it became clear to the naked eye: platforms are preparing to live in a new way. Everything is moving towards greater automation, where the system is not only responsible for delivery and optimization, but also looks inside the creative — what exactly to show and when to present it. The feed has become so predictable that users have learned to sense “an ad is coming” even before they realize it logically. Therefore, the development of algorithms was inevitable.
Full-screen TikTok and Reels have stitched everything into one continuous stream: there is no separate content and separate advertising, there is just the next video on the whole screen. The algorithm makes no exceptions here.
Platforms are beginning to reinforce what they consider to be the “new normal”: automation plus constant rapid testing of creatives at scale. In October 2025, TikTok embedded Symphony Automation directly into Smart+ — now the generative engine sits inside Ads Manager and generates, updates, and selects videos itself, based on behavior in the feed, not just campaign settings.
On December 10, 2025, Instagram launched Your Algorithm — a separate interface where users can see which topics the system attributes to their interests and can customize this list for themselves. For advertising, this means tougher conditions: when a person directly indicates what they like and what they don’t. However, this is currently only available in the US, and we can imagine the further direction of movement.
TikTok and Reels: similar formats, different characters
Visually, TikTok and Reels share a similar appearance: they are both vertical, full-screen, and feature endless swiping. Both platforms work as recommendation systems rather than classic social networks — content is selected based on behavioral signals rather than friendships.
But inside, they are different:
- TikTok works in a more cold and mathematical way, almost regardless of the age of the account or brand. If the first impressions do not give strong behavioral signals, the campaign “dies” very quickly.
- Reels still retain some of the social network logic: the reputation of the page and previous interaction with the audience can “soften” the punishment for imperfect creativity. Sometimes what goes in as native UGC on TT sometimes loses out to gloss on Instagram.
In practice, it looks like this:
- If a TikTok ad fails at the start, it means an immediate halt to the campaign;
- in Reels, the same failure more often turns into an increase in CPM, but the video still has a chance to live on.
- Another fundamental difference is the logic behind AI. TikTok actively promotes Smart+ in conjunction with Symphony Automation, artificial intelligence that generates variations of advertising materials and tests them in real time.
Meta is moving more cautiously for now. AI is deeply integrated into optimization and supporting creative tools, but it does not yet control the full cycle of “generate → test → scale” in such a radical format. Against the backdrop of tightening regulations in the EU, some of the “intelligence” is being removed from targeting and transferred to creativity — it is the video itself that must explain to the system who it is really relevant to.
Advertising on TikTok and Instagram is developing in different ways, at different speeds and with different levels of risk. However, the end goal is the same — to hold the viewer’s attention and not fall out of the feed.
Duration as a behavioral metric, not the “right” number
The question of “how long should the video be” is not important. What matters is not the timing, but how the user’s attention moves within the video.
The short format of Instagram Reels wins where:
- the product is simple and already familiar to the audience;
- it is easier to read in 1–2 seconds;
- the payoff can be given in 5–12 seconds.
This is a “see-understand-click” format. The length of Instagram Reels should not provoke any unnecessary actions other than the target one.
When long Reels are appropriate:
Longer videos (25-45 seconds or more) are justified when you need to:
- show a product demo;
- analyze a case study;
- compare “A vs. B.”
Formats that pass algorithmic selection
Minimum rejection at the start
The task of the first few seconds is not to sell, but to avoid triggering an internal “ugh, advertising” reaction. Formats that look like a natural continuation of the feed work:
- POV, where the camera is the user’s eyes;
- face-to-camera with text on the screen that immediately provides context;
- native elements: subtitles, simple overlays, minimal editing.
TikTok and Instagram systematically reinforce subtitles and on-screen text because it’s a quick way to explain “what’s going on here” before the person turns on the sound.
A structure that provokes rewatch
Review is one of the strongest behavioral signals for the platform. It is interpreted not as interest in the topic, but as confirmation of the value of the structure. Rewatch occurs when the video has:
- a loop (the end is stitched to the beginning);
- two levels of reading: image + separate meaning in the text;
- split screen, reactions, comparisons, where the brain follows two streams simultaneously.
Understandable even without sound
Watching “on mute” has become the norm, especially on Reels and TikTok. Therefore, the video should be a complete experience for both those who listen and those who only read. The best options here are:
- strong on-screen text and subtitles with clear context;
- visually readable structure: “step 1-2-3,” checklists, “before/after,” etc.
Metrics that really mean something
Metrics are no longer just a reporting tool. They are the language the algorithm uses to describe the quality of creative content. So how do you make a successful Instagram Reels video?
When it comes to video advertising on TikTok and Reels, it makes sense to look first at:
- 3-second hold/thumbstop — did the video stop the swipe;
- average viewing time — does attention wane after the “main effect”;
- completion rate + rewatch — do viewers watch to the end and come back again;
- drop points — at what moments do viewers leave en masse, and how does this coincide with the script.
Campaign settings are important, but metrics shift most quickly when the following change: the first frame, the hook, the first phrase, the first caption on the screen.
Creativity as a system of variations
TikTok advertising has long since ceased to revolve around a single “perfect” video. The working unit is now a system: when one basic message is broken down into dozens of variations for different attention trajectories.
Typical structure:
- 5 different hooks (different promises, different ways of reaching the target audience);
- 3 demo options: quick, detailed, through an example;
- 2 types of endings: soft CTA vs. direct;
- 2 voiceover or ToV options.
As a result, dozens of videos are produced from a single set of materials, and none of them is “the main one.” The algorithm itself decides which combinations are best for this particular audience and this particular moment.
Why the “most powerful” AI models don’t win
In 2026, the key models will be those that best fit the platform’s advertising cycle. The algorithm doesn’t care which tool generated the variation — it only cares about the result in behavioral signals. The most valuable models are those that work with the recomposition of creative content rather than creating it from scratch. These are tools that:
- repackage the same message into different opening phrases;
- adapt the text to different speeds and lengths of the video;
- generate alternative voiceovers with different emotional tones;
- rearrange subtitles for sound-off viewing;
- simplify or condense the presentation without losing meaning.
A person forms a hypothesis and a framework, the model accelerates it to dozens of options, and the algorithm itself selects the viable ones. TikTok advertising has gone the most radical route in this sense.
When the question arises of how to make an Instagram reel that holds attention and scales, the answer lies not in a single tool, but in a holistic approach to working with creativity. In 2026, it is not individual AI models that “work,” but a combination:
- LLM for variations → video/voice AI for quick replacement → built-in AI platform for selection.
And it is the platform that decides what will survive.
What does all this mean for advertisers in 2026?
Advertising on TikTok and Instagram is no longer just about the “right settings” but is turning into a live system for working with user attention.
To simplify it in practice, in 2026, the following will be decisive:
- different variations of creative content, rather than a single video;
- formats that do not look “foreign” in the feed;
- a structure that provides the user with quick and clear benefits;
- AI that speeds up testing but does not replace strategy.