We hear almost every day about new advancements in model intelligence and agentic capabilities, like the new Astra model or Fable 5.1, and open source is catching up faster than we thought.
While almost all of them can do powerful research and dive into the web, one thing they can't do well is look inside platforms like TikTok, Instagram or LinkedIn and pull out real numbers.
Take an example: say you run a small brand that sells coffee gear and want to know which home-espresso videos actually took off on TikTok this month: how many views, what the first few seconds looked like, and what viewers wrote underneath. The AI basically has a few core options:
- Use web search, which is based on indexed data (e.g. from the ChatGPT crawler). You get "trending coffee hashtags" articles and "how to go viral" guides, written weeks ago, about someone else's niche, with no view counts at all.
- Use agentic web browsing (a Chrome extension like I talked about in a previous post, or computer use). This emulates you scrolling TikTok and opening comment threads one by one. It works, but for a hundred videos and thousands of comments it is slow and expensive.
With a couple of videos you can certainly do that by hand. But if you want the whole picture of your niche, this is a dead end.
What you need instead is a TikTok scraper: a tool that reads the videos for a search or hashtag and hands you views, likes, captions and comments as a table.
Luckily there is a way to do this without writing any code, using a data scraping platform called Apify. If you don't know Apify, check them out here.
In a nutshell, they offer ready-made scrapers for almost the entire web, including an official TikTok Scraper and TikTok Comments Scraper, and provide them through a single interface, which you can even connect to your AI.
I have already talked about this in many posts in the past, but frankly it was always still too difficult for some folks who are not very tech-savvy to understand the concept and connect Apify to their AI.
But recently Apify decided to go their own way and not just offer a connector (MCP) but provide their own AI, which knows, given a simple or complex question, how to:
- select the scrapers that can help
- configure these scrapers automatically
- run and monitor them
- efficiently sift through the scraped data to answer the question
The interface is as simple as ChatGPT or Claude, with one huge difference: access to almost the entire web.
And if you don't consume too many scrapes, you can even use it entirely for free (if you stay within the $5 monthly free scraping credit) (apify.com).
Let's take the example. Say I sell coffee gear and want to know what home-espresso content is working on TikTok right now.
Once you have signed up, simply head to https://console.apify.com/ and open the Dashboard.
By default it will do what most other AIs do: it will search the web.
The result is what you would expect: articles about "trending coffee hashtags" and general tips for going viral. Nice to read, but no view counts, nothing from this month, and not a single real comment from real viewers.
So you ask the same AI to go to the source: scrape TikTok directly and read the comments. You can also give it a budget so it never goes over the free tier.
Under the hood it picks the right scrapers from the Apify Store, the official TikTok Scraper and TikTok Comments Scraper, runs them, waits for them to finish, and evaluates the results. One of them, the profile scraper, came back empty; that happens, and the AI simply works with what the others found.
And after a few minutes you have a completely different picture. Instead of generic hashtag tips, it has the actual top videos of the month in your niche, sorted by views, with what they have in common, and it read the comments, so you know what viewers keep asking about. That is a content plan and a product-FAQ list in one.
(Creators in the screenshot are described, not named.)
What did it cost? The TikTok Scraper is priced per result (about $0.0037 per video on the free tier, per the Apify Store) and the Comments Scraper per comment (about $0.00125), so a run like this stays well inside the $5 of free monthly credit.
Before you copy anyone: use this to learn what formats and questions resonate, not to copy other creators' videos. Stick to public posts and respect TikTok's terms.
This was of course just one example. Swap "home espresso" for "skincare routine", "budget travel Japan" or "standing desks"; the technique is the same. The point is: web search tells you what people wrote about TikTok; a TikTok scraper tells you what is actually happening on it.
Top comments (0)