Every guide to tracking brand mentions gives you the same stack: set up Google Alerts, run keyword searches, check your notifications, maybe pay for a monitoring tool that does the searching for you. I have no argument with the stack. I have an argument with its definition of a mention, because every tool in it reads text, and on TikTok and Instagram the mention that matters is very often not text. It is a product on camera, or a brand name said out loud, above a caption that says nothing.
This guide covers the text layer quickly, because you should still build it and most of it is free. Then it spends its time on the part the other guides structurally cannot cover: how to catch the mentions nobody typed.
What a brand mention actually is on short video
On the web, a mention is a string. Your brand name appears in an article, a forum post, a review, and a crawler can find it. Every tool in the standard stack was built for that world, and it carried its assumption over to social: a mention is your @handle in a caption, or your brand name typed somewhere a search can reach.
Short video breaks that assumption into three pieces.
The tag
In your notificationsThe creator types your @handle in the caption or tags your account.
The only form the platform delivers to you. Also the rarest.
The typed name
Findable by searchYour brand or product name appears as plain text in the caption, no tag.
Keyword search catches it, if you search the right variants often enough.
The shown and spoken
No text at allThe product is on camera or named out loud. The caption says nothing.
No alert, no search result, no notification. This form only exists on video.
The platforms only deliver the first form to you. The second is findable if you search often and creatively enough. The third form generates no alert, matches no keyword, and appears in no tab, and as you are about to see, it is not an edge case. It is a fifth of everything, sitting inside an untagged majority.
Step 1: Build the free text layer in an afternoon
Do this first and do not overthink it. The text layer is table stakes, not the strategy.
- Google Alerts for your brand name and flagship product names. This covers the web: articles, blogs, forums. It sees almost nothing that happens inside TikTok or Instagram, and that is fine, because that is not its job.
- Notifications and the tagged tab on both platforms. Check them, respond to them, but write down what they are: a feed of the creators who chose to tell you. That is a courtesy some creators extend, not a census.
- Saved in-app searches. This is the one part of the manual text layer worth real discipline, so it gets its own step.
Step 2: Search the platforms like you mean it
Once a week, search TikTok and Instagram directly. Not just your brand name once: the variants. A checklist beats inspiration here.
| Search for | Why |
|---|---|
| Your brand name, spaced and unspaced | Captions write "glow recipe" and "glowrecipe" interchangeably |
| Common misspellings | Creators type from memory, not from your style guide |
| Flagship product names and nicknames | Products get mentioned without the brand far more often than with it |
| Your branded hashtag, plus the community's variants | The tag the community invents rarely matches the one you launched |
| Your handle without the @ | Catches the half-tag that never reached your notifications |
Log what you find in a spreadsheet, one row per post: creator handle, follower count, date, views, tagged or not. The rows are the asset; the searching is just how you get them. Twenty minutes a week at a small scale.
This is where every ranking guide stops, and it is worth being precise about what stopping here buys you. Here is what the full stack, tags plus diligent keyword search, would have found of the brand posts our tracking index surfaced on the same platforms.
6,000 detected brand posts, by what would have found them
4,572 creators
Your notifications · @-tags and account tags
1,372 (23%)
Caption keyword search · Brand name typed anywhere in the caption
4,788 (80%)
Detection · Scanning the platforms, then watching the creators
6,000 (100%)
Waiting for the tag finds fewer than a quarter of the posts. Careful caption search takes you to about 80%, which sounds respectable until you notice the remainder: 1,212 posts about these brands, with real reach, that no text query could ever return, because the mention lives in the video and nowhere else. The best possible manual tracker misses one post in five, and it misses them silently. You do not get a report of the posts you didn't find.
Step 3: Watch creators, not keywords
The way past the text wall is to stop treating the post as the unit and start treating the creator as the unit.
Text search has one job in this workflow: it is how you discover a creator for the first time. The moment a creator shows up in a search or a tag, they go on a watch list, and from then on you check their profile, not your keywords. A creator who posted about you once, untagged and unprompted, is far more likely to do it again, and their next post about you probably will not carry text either. Watching the handle catches it. Waiting for the keyword does not.
This flips the weekly routine. Instead of re-running searches and hoping, you keep a roster: every creator who has ever mentioned you, in any of the three forms, with a weekly pass over the recent posts of the ones worth watching. The searches still run, but their output is now names for the roster, not mentions for a count. Repeat posters float to the top, and repeat posters are the real finding in any mention data, the difference between ambient buzz and an actual relationship forming. It is the same roster logic that makes competitor social media analysis work when you point it at a rival, and the same reason detected creators make the warmest recruiting shortlist you will ever have.
Step 4: Count reach, not mentions
Most mention tracking ends in a count: 34 mentions this month, up from 28. The count hides the thing you actually want to know, because mentions are not remotely equal, and the tagged ones are not the big ones.
1.31B
views on detected brand posts
6,000 posts surfaced by listening scans across TikTok and Instagram
69%
of those views were untagged
Posts with no @-mention hold 904M of the 1.31B views
3 in 4
creators never tagged the brand
3,470 of 4,572 creators posted without ever typing the @
Sixty-nine percent of the views in our index sit on posts that never tagged the brand. Report mentions from your notifications and you are not just undercounting by two thirds on volume, you are dropping the majority of your actual audience reach on the floor. Weight every mention by its views, and never report a mention count whose untagged share you do not know. If your dashboard says most of your mentions are tagged, that is not a healthy program, that is a tracker that can only see tags.
Once the roster has views attached, the useful questions get cheap: which creators drive reach rather than rows, whether this month's spike was one big post or forty small ones, and what happened after you gifted or hired someone off the roster. That last one is where mention tracking stops being PR hygiene and starts feeding the numbers that justify the program.
When the manual version breaks
The spreadsheet version of all this genuinely works, and I would run it for any brand doing under a handful of mentions a week. It breaks in a predictable place: the searching scales linearly with your creator count, the misspellings multiply, and the watch list outgrows the weekly pass. The failure mode is quiet, too. You do not notice the untagged posts you stopped catching, because nothing tells you they exist.
That is the point where listening software earns its slot: the scans run daily instead of when you remember, every discovered creator goes on the watch automatically, and the untagged fifth stops depending on your Sunday evening. The mechanics are the same ones in this guide. The difference is that they run without you.
The short version
- The standard mention stack reads text. On TikTok and Instagram, a mention has three forms, and only two of them leave text.
- Of 6,000 brand posts surfaced by our listening scans, 23% tagged the brand, about 80% were reachable by careful caption search, and one in five carried no text trace at all.
- Build the free text layer in an afternoon: alerts, notifications, and a weekly in-app search checklist with spacing variants, misspellings and product nicknames.
- Discovery by search, tracking by creator. Every found creator goes on a watch list, and their profile, not your keywords, is what you check from then on.
- Weight mentions by views. The untagged majority held 69% of the reach in our index, and three in four creators never tagged the brand once.
- Go manual until the watch list outgrows the weekly pass, then hand the same workflow to software that runs it daily.
Nora EllisUGCSignal
Nora writes about creator programs and the numbers behind them, drawing on the posts, views, and revenue UGCSignal tracks every day.