How to build a TikTok swipe file that stays current

A swipe file becomes stale for a simple reason: saving is easy and reviewing is optional.

The fix is not a better folder. It is a record that keeps the watched opening, source, date and performance together, plus a rule that removes old evidence from the current view.

Your TikTok swipe file should answer one production question: which opening patterns deserve a test in the next batch?

What belongs in the file

Save openings, not attractive screenshots.

A screenshot can be attached, but it should never be the whole record. It loses motion, sound, the source and usually the date.

Pick a tool you will maintain

Use the smallest system that supports the fields and views you need.

• Spreadsheet: easy to start, filter and share. Good for one researcher and a few hundred rows.
• Notion or Airtable: useful when examples need richer cards, attachments and linked pattern records.
• Bookmark manager: fast capture, but often weak for structured comparison.
• Database or internal tool: worthwhile when collection is automated or several teams query the evidence.

Do not spend the first week designing a perfect schema. Create the required fields, add 20 watched examples and see which questions remain hard to answer.

Separate examples from patterns

One video is an example. A pattern is a repeated opening structure across relevant videos.

Create two linked views:

Example view

One row per video with the complete source and observation.

Pattern view

One row per reusable formula, linked back to the examples that support it.

For example:

Pattern: show the finished result, put the limiting number on screen, then explain the constraint.

The pattern card should link to several videos, show which niches used it and record whether you have tested it. This stops a copied line becoming the organising unit of the file.

Practitioners describe useful swipe files in similar terms: save what stopped the scroll, then study why the opening worked rather than collecting text to paste (discussion). That is the difference between a research file and a quote folder.

Use tags that change a decision

Tags are only useful when they help you filter a shoot or compare outcomes.

Start with:

• platform;
• named niche;
• creator type;
• opening pattern;
• proof type;
• date window;
• performance band;
• research state;
• tested result.

Avoid subjective tags such as “great,” “inspiring” or “viral vibe.” They are hard to apply consistently and do not tell the next person what to make.

Keep the taxonomy small. If two tags always appear together, merge them. If a tag never changes a search or decision, delete it.

Add a freshness window

“Current” needs a date rule. Start with seven days for fast-moving niches. Widen to 14 or 30 days when the category publishes too slowly.

Do not delete older examples. Move them out of the current view.

Never edit the original publication date to make an old example look fresh. If a pattern returns, the return is new evidence. Keep both moments.

Set a kill rule

The kill rule prevents the file growing forever.

Archive an example from the current view when any of these is true:

• the source is gone and no useful observation remains;
• it falls outside the freshness window and has not recurred;
• it does not fit the named niche;
• the opening was saved from a caption but never watched;
• it duplicates a stronger example from the same creator;
• its pattern cannot be described in a way you could shoot;
• the performance number has no date or source.

Archiving is not failure. A smaller current view is the product.

Capture in under a minute

Long capture forms encourage people to save now and “finish later.” Later is where swipe files go to become decorative.

Use two passes.

Capture pass

Save the URL, creator, platform, publication date, views and a one-line opening description.

Review pass

Watch the opening twice, complete the visual/text/spoken fields, assign a pattern candidate and decide whether it belongs.

Items that never survive the review pass expire automatically after a short queue window. An unfinished inbox should not leak into the trusted file.

Attach your own results

Market examples select ideas. Your own posts tell you whether the pattern transfers.

When you test a pattern, link the result back to the card:

• version and publication date;
• what changed in the opening;
• early retention or closest available metric;
• later retention;
• business outcome when relevant;
• what you would keep or change next.

Your own winners deserve a separate filter. A practitioner can often learn more from openings that repeatedly worked on their account than from another enormous public outlier (discussion).

Do not erase losses. A pattern card with three weak tests is useful evidence that the public example did not transfer under your conditions.

A weekly review that stays small

Run the same sequence:

1. Clear or reject the capture inbox.
2. Move examples outside the freshness window to stale.
3. Find candidate patterns with new independent repetition.
4. Add recent performance observations where needed.
5. Promote only the patterns you can describe as a sequence.
6. Pick a small number for the next shoot.
7. Attach results from the last tests.

The review is finished when the production shortlist is clear, not when every saved video has been admired again.

Migrate an old swipe file without importing the mess

Do not copy every screenshot and bookmark into the new structure. Start with a quarantine view.

For each old item, ask:

1. Can you recover the source video?
2. Can you date the post or observation?
3. Did you watch the complete opening?
4. Does it fit a niche you still research?
5. Can you name the sequence and viewer promise?
6. Is there a performance note worth preserving?

If the answer is no to the first two, archive the item as inspiration rather than evidence. If the source survives, rebuild the row by watching it again. Do not infer the spoken opening from a screenshot.

Migrate pattern cards before individual examples. Pick the few formulas your team still uses, move their strongest verifiable sources, then add current examples. This produces a useful file quickly and leaves the archaeological project optional.

Keep the old folder read-only for a while. People stop trusting a new system when a familiar reference disappears, but duplicating everything makes the new file stale on arrival.

After the migration, compare the current view with the next production brief. If the creator still has to search through archives or ask where an example came from, the file is organised around storage rather than use. Change the default view until the next decisions sit at the top.

Give collaborators a saved view for the current niche and test queue. They should not need permission to edit the underlying taxonomy just to find the three patterns scheduled for production.

Common swipe-file failures

PDF packs

A PDF of 100 hooks has no live sources, dates or performance. Treat it as writing inspiration, not current research.

Screenshot folders

They preserve one frame and lose the opening. Add the source and watched description or discard them.

One giant “viral” tag

It mixes unrelated niches and account sizes. Keep the audience and problem visible.

Saving only winners

Keep a small counter-sample of relevant ordinary posts. It helps reveal which supposed patterns also appear when nothing special happened.

No owner

If nobody archives stale evidence, the file will lie by accumulation. Assign the review to a person or automate the expiry state.

Frequently asked questions

What is a TikTok swipe file?

A TikTok swipe file is a structured collection of source videos and watched openings used for research. A useful file keeps dates, views, niche, visual and spoken details, pattern labels and test results.

How often should I update a hook swipe file?

Review it on the same cadence you plan content. Fast-moving niches may need a weekly current view; slower categories can use a 14- or 30-day window.

Should I delete old hooks?

Usually no. Move them out of the current view and retain the history. If the pattern returns, record a new observation rather than changing the old date.

Can I build a swipe file in a spreadsheet?

Yes. A spreadsheet is enough when it keeps the required fields and can filter by niche, date, state and pattern. Move to a richer tool only when the workflow demands it.

Turn the file into the next shoot

Choose a few current patterns with independent examples and write the first controlled test for each. If the file cannot produce that shortlist, the answer is not more saving. Tighten the fields, tags and kill rule until it can.