Validate a startup idea with TikTok search data

Idea validation usually means asking people whether they would use your thing. They say yes, you build it, nobody shows up. TikTok search data skips the asking: it shows what people already type into a search bar when the problem is annoying them, how many do it, who they are, and what they use instead.

This is the method we use to validate a startup idea before writing code. Five checks, about twenty minutes, every number from TikTok's own search tool. We ran all five on one idea while writing this, a meal-planning app for food allergies, and left the real output in, including the parts that argue against building it.

Five checks: idea validation in twenty minutes

Market validation on TikTok answers five questions, in order. Stop at the first failure.

1. Is anyone searching the problem? Not the category. The problem.
2. Is the demand steady, or is it this week's news?
3. Who is searching? Age, country, and whether that is your buyer.
4. What do they use today, and what do they hate about it?
5. Is anyone answering them? If not, you have a gap and a marketing channel in one.

The data underneath is Creator Search Insights, TikTok's own search tool: a 7-day search count per topic, a daily chart, an audience breakdown and a video count. Inside TikTok it has no export and no API, so the numbers below came out through a data pull.

Check 1: is anyone searching the problem

The idea: an app that plans a week of meals around a household's exact allergens, so a parent stops reading every label.

The wrong search is recipes. The right search is the phrasing of the pain. What came back on 19 September 2026:

Data: TikTok Creator Search Insights via TokConnect, pulled 19 September 2026. Open the snapshot.

Search popularity is a 7-day global count for a cluster of searches, not one exact phrase.

The signal is the shape, not the size. Substitutions, triggers and children's meals come back as separate problems rather than one blurry category, and autocomplete for food allergy returns food allergy parent. Somebody specific has this problem. Check 1 passes.

Check 2: steady demand, or this week's news

A 7-day number cannot tell a six-month habit apart from a Tuesday. Pull the 30-day and 180-day chart before you trust it.

For gluten free meals for kids, an established topic with 489 videos matched to it, 140 of the 181 days in the 180-day window carry a number. The median day is 99,656 searches, the floor is 49,577, and the peak is 2,453,341 on 29 May 2026. A background hum with one burst is what a permanent problem looks like.

The same pull turned up the pattern to watch for.

Data: TikTok Creator Search Insights via TokConnect, pulled 19 September 2026. Open the snapshot.

That second row would have fooled us. Fourteen times the weekly searches, and a 180-day chart with one day on it. TikTok generates topic clusters continuously, and a new cluster reports a 7-day figure with nothing before it. It tells you people search that way. It does not tell you they have for six months. Run check 2 on an established topic, and save the generated ones for check 5.

Check 3: who is actually searching

The audience pull on that topic, 30 days, returned 5,797 searchers TikTok could classify.

Data: TikTok Creator Search Insights via TokConnect, pulled 19 September 2026. Open the snapshot. Shares are of people TikTok could classify, so read the 17% unknown as unknown.

Six in ten searchers are 25 to 34, seven in ten are women, and almost all are American. That is a parent of a small child, which the autocomplete phrase food allergy parent confirms from another direction. It is the best news in the exercise: a 32-year-old buying safety for a child is an easier sell than a 20-year-old buying convenience. More on the three "audience" datasets in the audience insights guide.

Check 4: what they use today and what they hate

Two pulls. First, the most-liked videos for the product phrase, which tells you who is already selling. Sixteen videos across gluten free app and food allergy app named fourteen products, among them Fig, Find Me Gluten Free, Spoonful, Spokin, JollyGut, GF Scanner, Gluten Dude, Clearplate and MyPlateGuard. A crowded market is good news: somebody already proved people want this.

Second pull: 49 comments under a creator's "top three apps for celiac" video. Twenty-seven of them name a specific app, and the complaints land on two things.

"I deleted Fig when they added the Paywall. It was great until you had to pay to scan items"

"FMGF is kinda hot and miss because it's all crowd sourced with no verification. Gluten dude is more reliable but has far fewer places in it."

"sadly fig is not available in my country"

"I've learned too if you Google the restaurant name and gluten free menu it will show Allergens PDFs that their websites don't always have"

"Currently getting a dog for gluten detection"

Comments pulled 19 September 2026 and quoted without usernames. Open the snapshot. Method for doing this at volume: export TikTok comments.

Paywalls at the point of use, and data nobody trusts. The workarounds are the loudest part: googling a restaurant's allergen PDF, phoning the manufacturer, training a dog. That is what people do when no app is good enough, and it beats any survey answer.

Check 5: is anyone answering them

TikTok reports a video count per topic. A big search number with zero videos means nobody has made the thing that answers it, which is a product gap and a free distribution channel at once: make the video, own the search.

Of the topics in check 1, six carry millions of weekly searches and zero matched videos, including children's meals for food allergies at 4,404,420. Zero means zero videos TikTok has filed under that topic, not zero videos about allergies. The content gap guide explains what counts as a match.

This is the cheapest test in the list. If your idea has ten zero-video topics attached, post ten videos answering them and watch whether anyone clicks through. A weekend's work.

The scorecard

Copy it, fill it in per idea, and be strict about the fail column. Three passes or fewer: keep looking.

What this cannot validate

Search demand is evidence that a problem is annoying. It is not evidence that anyone will pay you to fix it.

• Price. Nobody searches a price. Here the comments got closer than the search data did: one person pays for a rival's better access, another deleted an app the day scanning went behind a paywall. That is a churn point, not a price.
• Trust. Most complaints in this set are about accuracy. No search tool can tell you whether your allergen data would be good enough for a parent to rely on, and that is the real risk in the idea.
• Intent. A topic cluster mixes "I want to solve this" with "this was on my feed and I was curious".

The next test takes a week. Put up a landing page in the wording of your top three topics, since that phrasing is proven to be how people describe the problem. Post three videos answering three zero-video topics, send them to it, and ask for an email. Then call five people who gave you one and ask what they did the last time a label was ambiguous, not whether they like your idea.

Pieter Levels ran a version of this in 2023, posting daily videos about a mockup and watching which idea got clicks. The search data just tells you which videos to make.

That sequence is the point: find ideas in search data, screen them with these five checks, then spend money only on the survivors. For more candidates, start with app ideas, SaaS ideas or the founder's guide to TikTok research.