Originality, Play Duration, Search Value, Engagement: Building to the Formula

Most creators can name two of TikTok's four scoring inputs without looking them up — originality and engagement feel intuitive, the kind of thing you'd guess even without reading the documentation. Fewer can name the other two, and almost nobody scripts deliberately for the one that's grown the fastest in the last two years. TikTok's own Creator Rewards Program page names all four plainly: originality, play duration, search value, and audience engagement, applied to content over a minute long. Three of those four are things you can bake into a script before you ever hit generate. One of them almost never gets scripted for at all — it gets left to whatever the algorithm happens to notice, which is a strange way to treat the input platform data says is growing fastest.

Four inputs, stated plainly by the platform

It's worth reading the four terms as the platform actually defines them, because two of the four names are more specific than they sound. Originality isn't "nobody has made this before" — it's content that reflects a creator's own point of view or creative process in a way that resonates with their specific community, which is a lower and more achievable bar than pure novelty. Play duration covers two different behaviors bundled into one score: how long people actually watch, and what share of viewers make it to the end. Search value is assigned based on how well a piece of content matches terms people are actually searching for on the platform. Audience engagement is the familiar one — likes, comments, and shares. Four named things, and the honest starting point is that only two of them get any deliberate attention from most creators writing a script: originality, because it's the creative brief anyway, and engagement, because everyone already optimizes for a strong hook and a clear ask. Play duration and search value are the two that get treated as outcomes rather than inputs — things that happen to a video rather than things a script can be built around.

Play duration is two different pressures, not one

Watch time and finish rate sound like the same metric measured two ways, but they reward different structures. Versely's own reference on the retention curve makes the useful distinction: the curve is read by its shape, and a steady, gentle slope through the middle of a video is a different achievement from a hard landing at the very end. Watch time rewards a video that holds attention throughout — no dead stretch where the curve goes flat. Finish rate, which completion rate measures as a single number, rewards a video that's been cut to the length its content actually supports rather than the length the format technically allows. A video can hold a strong middle and still lose on finish rate if it runs thirty seconds longer than its idea justified, and a tightly-cut video can post a strong completion number while still under-performing on total watch time if it's too short to accumulate much of it. Scripting for play duration means treating both halves as separate constraints: build a middle that doesn't sag, and cut the whole thing to the length the material earns rather than the length you're allowed.

Search value: the one almost nobody scripts for

This is the input most scripts skip entirely, and the platform behavior data suggests that's an increasingly expensive gap to leave open. Adobe's research, fielded in January 2026, found that 49% of US consumers say they've used TikTok as a search engine — up from 41% in Adobe's 2024 report. That's not a niche behavior anymore; close to half the platform's audience is arriving at content the way they'd arrive at a search result, typing or speaking a specific question and expecting a specific answer, not just scrolling a feed and hoping something relevant appears.

Scripting for that is mechanically different from scripting for a hook. A hook is written to stop a scroll; a search-value line is written to match a query someone would actually type — which means naming the thing, in words a search system would match, early and plainly, rather than teasing it. "The mistake everyone makes with X" performs as a hook. "How to fix X when Y happens" performs as both a hook and a search match, because it contains the actual phrase a person searching would use. The gap between those two openings is the whole difference between scripting for engagement alone and scripting for engagement and search value together — and it costs nothing to close, because it's a phrasing decision, not a production one.

The practical extension of this matters just as much as the line itself: a search system can only match text it can read, and a video's spoken audio isn't text until something transcribes it. A script that states the exact query out loud, paired with automatic captions that put those same words on screen as searchable, readable text, gives a search-matching system two chances at the phrase instead of one buried in an audio track it may or may not process the same way.

Audience engagement is the one you can only invite

Of the four, engagement is the input with the least direct scripting lever, because likes, comments, and shares are ultimately someone else's decision made after the video's already done its job. What a script can do is create the specific conditions that make engagement likely — a clear point of disagreement worth commenting on, a moment worth sharing because it's genuinely useful on its own, a direct question at the end rather than a vague sign-off. That's an invitation, not a guarantee, which is the real distinction between engagement and the other three inputs: originality, play duration, and search value are properties of the content itself, decided before a single viewer sees it, while engagement is a property of how the audience actually responds — correlated with good scripting, but never fully controlled by it.

Scripting for three inputs in one pass

Put together, a script written with this formula in mind is doing three jobs at once rather than one. It states a specific point of view early, which is the originality input. It's cut to the length its content supports and holds a steady middle, which is the play duration input, covering both halves. And it names the actual thing a viewer would search for, in plain words, spoken and captioned, which is the search value input almost nobody bothers with. Engagement isn't ignored in this pass — a script with a real point of view and a real answer to a real question tends to earn engagement as a side effect, which is a better outcome than chasing it directly with a generic call-to-action bolted onto content that doesn't otherwise invite a response.

A Versely walkthrough: scripting the search-value line deliberately

The concrete version of this, run through the agent: draft the script with the target query stated as an actual sentence early in the piece — for example, prompting "Write a 45-second script answering 'how do I stop my houseplants from getting leggy,' state that exact question in the first five seconds, and keep the middle moving so nothing sags before the answer lands" — which handles originality (a specific answer, not a generic tip list) and play duration (an explicit length target with a pacing instruction) in the same brief. Generate the video from that script, then run add_subtitles_automatically against the finished clip: it transcribes the spoken audio and burns timed captions onto the video automatically, which means the exact search phrase you scripted into the opening line now exists twice in the finished asset — once spoken, once as on-screen text — rather than once, buried in an audio track. None of that touches how much the video earns, which the payout structure at /make-money covers on its own terms; this is entirely about which of the four scored inputs the script was actually built to satisfy before a single viewer sees it.

Takeaway

Originality and engagement get scripted for by habit; play duration and search value mostly don't, and search value specifically tends to get skipped entirely because it doesn't feel like a creative decision. It is one — a script that states the actual query in plain, spoken, captioned words is doing something a vaguer, teasier hook can't, at a moment when close to half of US consumers say they're already using the platform the way they'd use a search bar. Four named inputs, three of them genuinely available at the scripting stage. Most scripts are still only reaching for two.