What People Actually Save From Travel TikToks: 231 Videos, 881 Places

Everyone who plans trips from TikTok has the same folder: a few hundred saved videos, a vague memory of which one had the cliff restaurant, and no idea where any of it is. Dreamfare exists to turn that folder into places on a map, which means every import leaves behind a record of what the video actually contained. We looked at all of them.

This is a small, honest dataset. It covers the 231 TikToks and 71 Instagram Reels that Dreamfare users imported successfully between 1 July and 14 September 2026, plus the cross-user index of 2,626 video-to-place links those imports built. It is what our users save, not what the internet saves, and it skews toward people planning real trips. Read the numbers that way.

The caption is not where the place is

The single most useful number in the set: 52.8% of travel TikToks never name any of their places in the caption. Not one. The venue is on screen for two seconds, or said out loud, or answered in the comments three days later when someone asks "where is this?".

Per place it is worse. Of the 881 places extracted from TikToks, only 42.7% appear anywhere in the caption. The other 57.3% had to be read off the video itself, the transcript, or the comment thread.

Instagram captions are a little more helpful, partly because Reels are more often a single venue with its name in the first line. TikTok captions are hashtags and a hook. Which is why "save it and search later" fails: later, there is nothing to search.

A typical video hides two places. One in twelve hides more than ten

The average TikTok import produced 3.8 places, but the average is pulled up by a long tail of roundup videos. The median is 2.

That last row matters. 7.8% of TikToks are "ten things to do in Lisbon" style roundups, and they carry a disproportionate share of the useful places: eighteen videos produced well over two hundred of the 881. If you only have time to import a few, import the roundups.

Reels behave differently: a median of one place, and 7% of Reels contained no identifiable place at all, usually because the video is a vibe rather than a venue. On the other end of the scale, the three Reddit threads in the sample averaged eighteen places each. A good Reddit thread is worth about five TikToks.

What people save: sights first, food second, almost never the hotel

Travel TikTok is a sightseeing medium. Instagram is where the restaurants are: food is twice the share on Reels that it is on TikTok. Stays barely register on either, which is consistent with what we see in the app: people find the hotel through search with dates and a budget, not through a video.

Nine in ten places are real, findable venues

92.4% of the places extracted from TikToks resolved to a specific venue with an address and a pin. 91.6% on Instagram. The remainder are places that could not be matched with confidence, and Dreamfare leaves those out rather than guessing, because a wrong pin is worse than a missing one. In the cross-user index, 10% of links have no venue id yet, which is the same story from the other side.

Under 1% of extracted places were an area (a neighbourhood, an island) rather than a venue. Creators point at specific things.

Tips are rare on TikTok and common everywhere else

A place is more useful with the sentence that came with it: "go before nine", "cash only", "book the sunset slot". Only 7.9% of TikTok places carried a tip like that. On Instagram it was 31.2%. In the Reddit threads it was 85.2%.

The pattern is the medium. A TikTok shows you a place; a Reddit comment tells you how to do it. A good plan needs both, which is the argument for saving the thread as well as the video.

Where the saves point

The cross-user index links 689 videos from 621 creators to 717 destinations. The twenty with the most saved places:

Two things stand out. Iceland and Interlaken are not cities, they are itineraries: a video about them is a route, and the places come in fives. And Seville and Florence produce a lot of places from very few videos, the roundup effect again. Nusa Penida and Ubud are the opposite: many videos, one place each, the classic single-spot Reel.

About one in five links cannot be read at all

For completeness: 61 of the 292 TikTok links users tried (20.9%) and 32 of the 103 Instagram links (31.1%) failed to import. Private accounts, deleted videos, region locks, and platforms refusing to serve the page all land here. Instagram is the harder platform to read, which matches the caption numbers above. If a link fails, a screenshot of the frame that names the place imports fine.

What this means if you plan from TikTok

• The caption will not save you. Whatever you do with a saved video, do it while the place is still on screen, or use something that reads the screen and the comments for you.
• Import the roundups first. Ten-things videos are 8% of your folder and most of your places.
• Save the Reddit thread too. It has the tips the video does not.
• Do not expect the hotel to be in the folder. Search for it with dates.

If you want the folder turned into a trip, the TikTok trip planner does the reading, the Instagram version handles Reels, and the video location finder answers the "where is this?" question on its own. For the method rather than the tool, start with how to turn TikTok saves into an itinerary.

Method

Counts come from Dreamfare's import records for 1 July to 14 September 2026: successful imports with at least one extracted place (231 TikTok, 71 Instagram, 3 Reddit, 3 YouTube, 1 pasted note), and the anonymised cross-user video-to-place index (2,626 links). "Named in the caption" means the extracted place name appears verbatim, case-insensitive, in the creator's caption; a place named with different spelling counts as not named, so the true caption rate is slightly higher than shown. "Real venue" means the extraction matched a specific place with an address. Per-user data, usernames and trip contents were not used; creator handles are counted, not listed. Small samples (Reddit, YouTube, notes) are reported for contrast only. We will re-run this when the sample is ten times larger.