What TikTok search reads
The TikTok marketing pillar describes search as one of the surfaces alongside the For You feed and the Following feed, where TikTok surfaces videos by their captions, on-screen text, hashtags and sounds, and notes that search rewards the same things the feed rewards while adding a reason to describe a video in the words people would type. The Creator Portal in the sources is where TikTok explains these features to creators, and it is the place to check when the screens change.
The content formats and hooks guide adds the detail that matters most for planning. TikTok can generate captions automatically, and creators can edit them before posting; captions are text the platform can read when it decides what a video is about. The caption field should add something the video does not say, in the words people would type, with a few relevant hashtags, because TikTok counts hashtags among its signals and documents nothing that rewards a wall of them.
That is the whole of what this guide relies on about how search works. TikTok does not publish how search ranks the videos it finds, and nothing here claims to know. The practical consequence is modest and useful: a video that is good for the feed and is described clearly, in the words a searcher uses, can be found by someone looking for it. A video described vaguely, or in the brand's own jargon, can only be found by the feed.
| What search can read | Where it comes from | What to do with it |
|---|---|---|
| Caption | The text written in the caption field when posting | Describe what the video answers in the searcher's words; add what the video does not say |
| On-screen text | Text added with TikTok's text tool | State the question or the premise in plain words within the opening |
| Captions of speech | Automatic captions, editable before posting | Say the words aloud, then check and correct the generated captions |
| Hashtags | Tags added to the caption | A few that describe the topic as people would; never a wall of them |
| Sound | The sound used in the video | Original sound for most search-led videos; the sound page is a feed route, not a description |
Search-led and feed-led videos
Not every video should be planned for search. The feed recommends on interest, so a video can travel because it is funny, surprising or satisfying without answering anything. Search works the other way round: someone arrives with a question, and the video that answers it clearly is the one worth watching. The difference is Search IntentThe goal behind a search query: to learn, to find a site, to compare or to act., the reason behind a query, and the guide to classifying keyword intent sets out how to label it consistently for web search. Its categories adapt to TikTok with little change: people look for how to do something, for what something is, for which product to choose, and for a particular brand or account.
A search-led video answers one of those questions and would still be useful to someone who found it months later. How-tos, explanations, comparisons between options, reviews of the brand's own products in use and answers to the questions customers ask before buying are search-led by nature. A feed-led video earns its views in the first days and has done its job: a trend adaptation, a behind-the-scenes moment, a reaction, a joke. The two kinds need different planning. A feed-led video is judged on its opening and its timing, which the posting cadence and trends guide covers. A search-led video is judged on whether it answers the question it is named after, and on whether it keeps answering it.
Most accounts benefit from both. A useful way to plan the mix is to give search-led videos their own Content PillarOne of a small set of recurring themes that a brand's content is planned around. in the content strategy framework's sense: a recurring theme of answers, with its own slot in the calendar, rather than an occasional how-to fitted in when the trend slot is empty. The organic TikTok hub frames the account's content promise; a search pillar is that promise expressed as the questions the account exists to answer.
Finding the words people use
The words that matter are the customer's, not the brand's. A brand calls its product a "fermentation vessel"; the customer searches for how to stop a brew going cloudy. The work is to find those phrases and keep them.
Start with the brand's own records, which this guide treats as the main research tool:
- Questions asked before buying. Sales conversations, live chat, emails and the questions shop staff hear every week are the clearest record of what people want answered and the words they use to ask.
- Comments on the brand's videos. The same questions recur, and the guide to Duets, Stitches and video replies explains how to turn the shared ones into a series. A question that keeps arriving in the comments is often a question people are also typing into search.
- TikTok's own search bar. The formats guide names search suggestions as a source for the idea bank. Typing the start of a customer's question into the app and reading what TikTok suggests shows how people phrase it, and the results show what already answers it and how well.
- Reviews, returns and support tickets. Problems customers describe after buying are questions future customers will search before they buy.
TikTok does not publish search volumes, and this guide does not estimate them. Judge a phrase by how often it turns up in the brand's own records and in the app's suggestions, not by a number from a tool that cannot see TikTok. The methods in the guide to keyword research when volumes are low transfer well, because they rely on customer language and validation rather than on volume figures. The wider keyword research guide covers grouping similar phrases into clusters, which on TikTok means deciding which questions one video can answer and which need a video each. Specific, longer phrasings, the long tailA specific, usually longer query searched rarely on its own but often as a group., are usually the right target for a brand: a question narrow enough that a clear answer from someone who knows the subject stands out.
Keep the list in the same place as the idea bank, one row per question, with where it came from, the exact wording customers use, any variants, and the video that answers it once there is one.
Saying, showing and writing the words
A search-led video puts the searcher's words in three places, and each has a second job besides being read.
The spoken line. Open by saying the question, or the answer to it, in the customer's words. The viewer who searched for it hears at once that they are in the right place, which is the hook's job; the formats guide covers hook craft and nothing here replaces it. Saying the words also puts them into the automatic captions, which TikTok can read once they are posted.
The on-screen text. State the premise in plain words in the opening, in the area of the screen the interface does not cover. The formats guide explains the clear area and timed text; for search the point is simply that the text names the topic as a searcher would, not as a slogan.
The captions. Check the generated captions before posting and correct them, especially product names, technical terms and anything the speaker said quickly. Captions also serve people watching without sound, and the guide to accessible social media posts explains why accurate captions matter for viewers who rely on them; an accurate caption serves both purposes at once.
The caption field. Write a sentence that says what the video answers, in the searcher's words, and adds what the video does not say: a detail, a variant of the question, the next step. Add a few relevant hashtags that describe the topic. Do not list every variant of the phrase; a caption written for a machine reads as one, and the person deciding whether to watch reads it too.
The profile carries the same words. The social page optimisation guide covers names, bios and pinned content across platforms. On TikTok, a bio that says in plain words what the account helps with, and pinned videos that answer the most common questions, give a searcher who lands on the profile a reason to stay.
Keeping search-led videos current
A search-led video is evergreenContent that stays relevant and useful long after it is published, rather than being tied to a date or a news moment. only while its answer is true. Products change, prices change, instructions change, and a confident how-to that is now wrong costs more trust than it earns. Treat the search pillar as a small library to maintain rather than a stream of posts to forget.
- Review on a schedule. On a regular date the team keeps, and whenever a product or process changes, list the search-led videos and check that each answer is still right.
- Replace rather than pile up. When an answer changes, post a new video that answers the question correctly, then archive or make private the old one if it would mislead, and move pins to the new one.
- Fill the gaps the comments reveal. Questions in the comments under a search-led video show what the answer left out. A follow-up video, or a video reply to the comment, answers them.
- Keep the question list alive. New questions arrive with new products and new seasons; the list is never finished.
Reading search views in analytics
A Business Account's analytics report, per video, where views came from: the For You feed, Following, the profile, search, sounds and hashtags. The organic TikTok hub includes the reading in its metrics table ("Did people find it by looking?"), and the TikTok measurement hub sets organic analytics beside the paid measures and a web analytics property.
For search-led videos, read three things. First, the share of views from search, compared with other search-led videos rather than with feed-led ones, since a trend post with little search traffic is doing its job. Second, how views arrive over time: a feed-led video usually collects most of its views soon after posting, while a video that search keeps finding can go on collecting views from search after the feed has moved on. Third, what the search viewers did: average watch time, follows and profile visits from search-led videos say whether the answer satisfied the people who looked for it.
Read the results at the level of the pillar as well. If search-led videos as a group bring a growing share of the account's views and followers, the pillar is working. If a video carries a search phrase in all the right places but earns no search views, the question may be one few people ask on TikTok, or one already answered well; try the next question on the list rather than rewriting the same caption repeatedly.
A home-brewing kit seller's search plan (illustrative)
Dimbrava Brewing Kits builds a search pillar from its customers' questions
Dimbrava Brewing Kits, an invented seller of home-brewing kits that trades online, has posted feed-led videos for a year: bottling time-lapses, tasting reactions and the occasional trend. Before planning anything new, the owner reads 3 months of pre-sale emails and live chat and lists the 14 questions that come up most, in the customers' own words: "why is my beer cloudy", "how long before I can drink it", "do I need to sterilise everything". Typing each into TikTok's search bar shows how people phrase them and which already have strong answers.
The shop starts a Wednesday answer slot. Each video opens with the owner saying the question as customers ask it, shows the question as on-screen text in the clear area, demonstrates the answer at the brewing bench, and ends with the next step. The owner corrects the automatic captions before posting, writes a caption that answers a variant of the question, and adds 3 hashtags. The 3 strongest answers are pinned to the profile, and the bio now says the account answers beginners' brewing questions.
After 2 months the shop compares the 8 answer videos with its feed-led posts in analytics. Search brings 9% of the answer videos' views in their first week and 31% across the 2 months, and still brings views weeks after posting; the feed-led posts stay near 2% from search throughout and have done their work within days. The answer videos bring 140 of the 360 new followers in the period. The shop keeps both kinds, adds the questions from the comments under the answer videos to its list, and books a quarterly check of every answer against the current kit instructions.
Common mistakes
- Writing captions for a machine. A list of keyword variants reads as spam to the person deciding whether to watch; one clear sentence in the searcher's words does both jobs.
- A wall of hashtags. TikTok documents nothing that rewards them; a few relevant tags are enough.
- Using the brand's words. Customers search in their own language, not the product catalogue's.
- Leaving automatic captions uncorrected. Mistranscribed product names and terms help neither the viewer nor search.
- Planning every video for search. Feed-led videos have their own job; give search its own pillar instead.
- Inventing search volumes. TikTok does not publish them; judge questions by the brand's own records and the app's suggestions.
- Letting answers go stale. An outdated how-to that still collects views from search does lasting harm.
- Reading search views on day one. Search-led videos earn over weeks; compare them with each other over time.
