Short answer: Spytrend can be connected to an AI assistant directly, over MCP. Then you do not click through filters, you write the task in plain language, and the assistant digs into the database itself, assembles the chain and brings back a finished answer. Several clicks through the interface turn into a single request.

When the assistant does the digging for you

The previous parts of the series showed the service through a user's eyes. You open a section, turn the filters, scroll the feed, look at cards. All by hand. That is fine as long as the task is a narrow one.

But there is a second way to work with the same database, and few people know about it. Spytrend can be connected to an AI assistant directly. Then you do not click through filters, you simply write in plain language: "find slots in Brazil that went from three accounts to a hundred within a week, show the webmasters behind them and their sites". The assistant digs into the database itself, assembles the chain and brings back a finished answer. Several clicks through the interface turn into a single request in words.

Here is what that opens up. Not a retelling of the web interface, but the things that are not there at all or that would wear you out to collect by hand.

De-anonymising a whole network from one tracking parameter

This is probably the most powerful part.

Every affiliate has his own markers baked into his links: sub-codes, pixel ids, all sorts of ch= and the like. Normally you see one webmaster and his domains. Here you can take any of his tracking parameters and ask who else in the database uses exactly that value.

I tested it on a live example. I took a webmaster, pulled a parameter out of his links and ran it against the database. One value (sdmode=4) turned out to be shared with 2,367 other webmasters. On top of that the database returned a list of pixels and other values that overlap between them.

Do you see what that is? You pull the entire connected infrastructure out by a single thread. An affiliate network, its resellers, the accounts of one team spread across dozens of different "faces". In the ad feed you would never glue them together, because from the outside they are separate entities. By a shared tracking parameter they give themselves away instantly.

A pixel gives the buyer away even when he changes pages

An affiliate keeps spawning new fan pages. One gets banned, he raises three more and runs the same thing. From the outside those are different pages, and you will not connect them in the feed.

But his pixel is often the same one, travelling with the bundle. Through it the database pulls out his entire current setup at once. I took one pixel id, asked where else it shows up, and got 52 ads spread across several different pages, but from a single pixel, to a single lander, with identical markers in the links. The person thinks he has multiplied himself and covered his tracks, while through the pixel his whole operation is in plain sight.

It also works ahead of time. Find a competitor, hook his pixel, and after that you simply monitor: every new page he raises under that pixel surfaces in the database on its own.

Telegram and WhatsApp funnels you cannot see in the feed

Part of the traffic goes not to a landing page but straight into a messenger. A person clicks the ad and lands in a Telegram channel or a WhatsApp chat, and the warming up happens there. In the regular feed, without the Hubs filter, that model dissolves among everything else.

The database can cut out exactly those ads that lead into a messenger. I asked how much active advertising currently takes people straight into Telegram and got 69,060 ads. The same separately for WhatsApp.

You immediately see the whole "ads to messenger" layer: who runs it that way, in which niches, to which geo. Useful when you want to understand who builds a funnel through a channel rather than through a site, and to look at the warming mechanics themselves.

A scale detector built to your rules

The web interface has Trends with ready growth buttons: x2, x5, x10. Convenient, but those are somebody else's thresholds.

Through direct access you set your own. Any number, not just the presets. "Show creatives that grew 4.5 times over 3 days, with at least 8 new ads yesterday, at least 6 fresh accounts, and all of it younger than three days." The database returns exactly that slice, with a daily launch history for every creative and its growth multiplier.

I ran that detector across gambling and creatives came up with hundreds-fold growth in a day (one visual went from zero to 246 ads within a day). And you can see how many there were the day before, how many yesterday, the date of the first impression, in which country. You construct your own definition of an "explosion" and catch bundles exactly when they start being scaled.

Separately, the database can catch the moment when the same text script suddenly starts being copied by several teams at once. That is an early signal that an offer has gone around the market.

Bulk export and "the same visual"

Once you have found the creatives that work, you have to collect them for analysis. Through direct access you can export a batch at once, up to five hundred per request, instead of saving them one by one.

And two more things you cannot do by hand.

First: from a single creative the database finds every ad that reuses it, by an exact match of the image. Not "similar by eye", but literally the same visual wherever it surfaces. I checked it on a top slot in Brazil: one creative with 22,288 ads behind it and 5,930 fan pages (Advertisers in the interface). One visual carries half the niche on its back.

Second: a search for visually similar creatives. That one is not an exact match but close variants of the image. Useful when you want to see how different teams redraw one good idea for themselves.

Site recon: advertising against traffic

The database knows not only the advertising but also the sites it leads to. Monthly traffic, where it comes from, the share of paid, what the store is built on and more.

The most valuable thing there, the one you will not see head-on, is the link between advertising and traffic growth. I took the Brazilian site mpljogos.com: 613 thousand visits a month, and the database itself flags that ad growth runs ahead of traffic growth, with the paid share at 97%. The project lives on paid traffic and is accelerating right now. That is how you catch not only bundles but whole growing projects before anyone talks about them.

Ecommerce and dropshipping get their own angle. I ran a query across ecommerce: show growing Shopify stores. First out came guragear.com (camera gear, US): +1013% traffic in a month, a jump from ten thousand visits to 118,944. Next to it the Brazilian-Venezuelan beauty store galeriaavanti.com at around +866% growth. And for each you can see the number of backlinks and the paid traffic share.

Effectively this is reconnaissance on other people's stores. You find whoever is accelerating right now and see straight away what works for them.

The whole path in a single request

Taken separately, all of this can be assembled in the web interface by clicking. But the point of direct access is that the assistant walks the whole chain for you.

Here is what it looks like on a real run through Brazilian gambling.

I ask what is taking off. The database returns fresh slots and crash games with sharp growth over the day. I take the most re-uploaded visual. I drill into it: 22,288 ads, 5,930 fan pages, 285 domains behind one creative. I look at who exactly runs it, take a fan page, and there are exact numbers on its ads and active count, the destination domains, the page's founding date. Then I go one level up, to the webmaster: the top affiliate network in Brazil runs 153,611 ads. I pull a tracking parameter from it and roll out the whole connected network. And finally I go into the sites and look at whose traffic is genuinely growing alongside the advertising.

That used to be an hour or two of manual digging across six sections. Here it is one conversation with an assistant in normal language. You formulate the task, it digs.

On the money side: requests spend tokens - 1 token per row of Meta data (search, de-anonymisation, trends, site statistics), while creative media and TikTok data cost more. Plans and limits are on the pricing page.

The web interface or direct access over MCP

They cover different jobs.

The interface is for looking with your eyes: you come in, scroll, catch the general picture, take apart a specific ad. Fast and visual when the question is narrow.

Direct access through an assistant is for digging deep and wide. When you need to walk the whole funnel from a trend to a site, glue a network together by tracking parameters, export hundreds of creatives at once or set your own threshold for an "explosion". Everything that would take half a day by hand, the assistant assembles in one request.

If a quick look at the niche is enough, open the web interface. If you need someone to systematically work through the database and bring back a finished chain with de-anonymisation, take direct access over MCP.

FAQ

What is direct access to Spytrend through an assistant?
It is connecting the database to an AI assistant directly. Instead of clicking through filters you write the task in words, and the assistant searches, joins the data and brings back a finished answer.

What can the assistant do that the interface cannot?
Set your own explosive-growth threshold instead of the x2/x5/x10 presets and export up to five hundred creatives at once. Gluing a network together by a shared tracking parameter, finding all of one buyer's pages by his pixel, isolating ads leading into Telegram and WhatsApp and finding every reuse of a visual are possible in the interface too, across different sections; the assistant assembles them in one request.

How much does it cost?
Requests spend tokens: 1 token per row of Meta data - search, de-anonymisation, trends, site statistics; creative media and TikTok data cost more. A free account gets 500 tokens for all time, paid plans get 40,000 a month.

Can I track a competitor by his pixel?
Yes. Once you hook a competitor's pixel, you simply monitor the database: every new page he raises under that same pixel surfaces on its own.

When is the web interface better and when the assistant?
The web interface when the question is narrow and you want to look with your eyes. The assistant when you need to walk the whole funnel from a trend to a site, glue a network together by markers or export hundreds of creatives at once.

The Spytrend section walkthrough