"Find content that's already hot, then put your app in it": a French growth marketer's chat toxicity checker app

Toxik AI is a French app: drop in a chat screenshot and it gives you a 0–100 toxicity score, a few red flags, and a comeback you can send straight away. The features aren't new; the English market had this over a year ago, and the French market has rivals three times its size. It got onto the French Lifestyle Top Grossing chart six weeks after launch because its maker is a growth professional who had already written the playbook up on LinkedIn. Below: what it is, where the traffic comes from, where the money comes from, who's competing, and what you can take away.

One LinkedIn post

On 2026 Jul 8, Paul Soubigou posted on LinkedIn, opening with: "If you build an app first and think about distribution later, you're already behind." He went on: "Most people build an app and then pray it spreads. I flip it: find what's already spreading, then build the app into it." He said he found a "clearly already hot TikTok content format," built a Gen Z app over a weekend in 2 hours, repeatedly tested onboarding and pricing until it converted, then automated the content so that "posting now takes 5 minutes a day." Result: over $1,000 in a month, #88 on the French App Store Lifestyle chart, "100% organic, ad spend 0."

Going by timing and description, that app was likely Doro, an dating reply coach he released in April (inferred). Two and a half months later, Sep 24, he shipped Toxik.

Who Paul is: his personal site says he was growth at Rocapine (a consumer app studio that raised a $13M Series A), ran two UGC creator programs totaling 50M organic views, operated a French Instagram network of 2.5M followers over 2021–2024 years, and built an AI tool for batch-generating app short videos. He lists a dozen-plus apps under his own name, writing "my apps combined bring in $5,000 a month, peaking at $2,000 in a week." That's the total across all apps, not Toxik alone.

What it is

The first line of the store page is: "On te manque de respect ? Envoie le screen." (Someone's disrespecting you? Send the screenshot.) The flow has four steps:

1. Upload a chat screenshot, or paste in the messages you received;
2. Get results in seconds: a toxicity score of 0–100, identified red flags (guilt-tripping, gaslighting, hot-and-cold, controlling behavior, playing the victim), a one-line verdict, and "who's driving the conversation";
3. Pick a tone and length. Three tones: Chill (calmly set the record straight), Énervé (firm, telling them enough), Méchant ("cash, tu fermes la conversation": direct, ends the chat); three lengths: one line, a few lines, a big paragraph;
4. Tap to copy and paste it back into WhatsApp, Instagram, Snapchat, Tinder, or Hinge.

No sign-up required; the store page says screenshots are "not saved." There's also a "talk to Toxik" venting mode. It uses "tu" throughout, giving it a young French voice.

Free to download, but the store page says "all features require a paid subscription": weekly €7.99, monthly €14.99, yearly €29.99 or €59.99 (French App Store, 11-06). Among the few French written reviews, one is an all-caps "PAYANT!" complaining there's not even a free tier.

How big it is now

A couple hundred ratings was enough to chart on the French Lifestyle Top Grossing list. One reason is that the bar for that chart isn't high; the other is the hard paywall plus weekly billing, so each paying user brings in more money (inferred). Lots of ratings with almost no written reviews usually means the app prompted for a rating (inferred, not verified).

Where the users come from: the content format was grown by someone else

We couldn't find Toxik's own TikTok account. But the traffic it's tapping into is easy to find: under #relationtoxique on French TikTok, everything is sock-puppet accounts like "xxx_privateaccount," each video feeding a chat screenshot to AI to judge how toxic the other person is.

For example, @shana_privateaccount's bio reads "ambassadrice the cringe" (The Cringe ambassador). From Oct 27 to Nov 4 she posted almost daily, with captions that are one exclamation each ("alors là je suis choquée" (I'm shocked), "mais pq personne m'en a parlé avant ??" (why didn't anyone tell me before??)), plus the same hashtag string, ending with "publicité*" tags. Her Nov 2 post "c'est chaud franchement on voit tout !!" (honestly brutal, you can see everything!!) got 90k views, while most others sat between a few thousand and twenty thousand. Another account, @douae_privateacco, used Moroccan Arabic catchphrases for the same content, peaking around 12k on a single post.

In other words, the "screenshot toxicity check" format had already been pushed and validated on a sock-puppet network by The Cringe before Toxik launched. What Paul did is exactly what his post describes: find what's already spreading, then build the app into it. Which accounts he used, or whether he drew on his studio's creator resources, isn't visible from outside.

Three side threads:

• Store copy in 44 languages. The French title is "Détecte & Réponds" (detect and reply), the US title switches to "Toxicity Checker," with keywords swapped per store. He describes himself as doing "up to 50 languages of ASO."
• Spillover into Francophone Africa. Senegal, Mali, Niger, and Madagascar all charted, and in these stores a few ratings can get you onto the Lifestyle free chart, a natural spillover from French content and French store copy (inferred).
• Paid ads: no evidence found. He publicly emphasizes organic traffic, but since paid acquisition is his day job, it can't be ruled out.

Why this result page works

"Is it me, or is he manipulating me?" has been a recurring emotional topic on French TikTok. Toxik's result page is built to be filmed: one big number (like 87/100), a row of red flags, and one harsh line. A screen recording is already a video, and users share it themselves.

Its difference from rivals is putting two things on one screen: The Cringe leans toward "get a chat report," the English-market Toxic AI toward "help me write a reply," while Toxik judges toxicity first and then replies in three tones. But honestly, that's not why it took off: pasting the screenshot into ChatGPT does the same thing. The real reason is distribution: pick a content format that's already hot, pick a language market English-first products have ignored, then use automation to pile on the content volume.

Who it's competing with

This category is crowded and not new. Toxik isn't the first, and it isn't the biggest. Almost nobody outside is paying attention to it: no press coverage, no discussion on X, and the developer only posted the combined revenue of their portfolio.

If it were you

One-line verdict: you can follow this, but don't copy this app, copy this person's pipeline. Find a TikTok content format where someone is already running ads and the view counts are steady, build an app in one weekend that produces the same result page, pick a language the English-language products haven't seriously worked, use a hard paywall with weekly subscriptions, and bet on several apps at once.

Who it suits: 1–2 people who can build a single-feature app with SwiftUI plus a large-model API in a few days, who can read TikTok in a non-English market, manage a batch of alt accounts, and automate posting. Ideally a native speaker of that language.

Who it doesn't suit: anyone hoping to win on features (free ChatGPT can do it); anyone who wants to build one app and expect it to earn long-term (Doro's rating has stalled around 160, and apps like this are mostly one-wave); anyone who doesn't want to make content and is just waiting for organic store traffic; anyone who cares about the reputation and review risk of "teaching people how to clap back."

How to do it:

• Pick a content format, not a feature. Search emotional topic hashtags in the target language on TikTok (romance, workplace, family, appearance ratings), and look for formats tagged "publicité" or "ad", posted daily from one account, with single posts that consistently hit tens of thousands of views. Someone is paying for it, which means there's downstream conversion.
• Design the result page to be "shootable": one big number, a set of tags, and one line people can copy directly.
• Don't go head-to-head with The Cringe on language: it already has two or three hundred to five hundred ratings each in German, Portuguese, Italian, and Spanish. Better gaps are languages where no one has seriously done the same format (Arabic dialects, Turkish, Indonesian, etc.; similar Arabic-language videos on TikTok have hit 60–700k views per post), or switch to a different emotional topic and skip "toxicity judging."
• Monetize like Toxik: free download, hard paywall, mainly weekly subscriptions; store copy in dozens of languages, with headline keywords adjusted per store.

How to get users: run 5–10 alt accounts, each with its own persona, using the same copy formula (an exclamation + screenshot + result-page screen recording) and the same hashtag set, one post per account per day; turn content production into a template so daily upkeep takes about ten-odd minutes. After two weeks, keep the accounts and scripts that got past 10,000 views and cut the rest. French content like this can also pick up the French-speaking Africa charts on the side; check other languages for similar "adjacent markets."

Timing: how long an app like this lives depends on how long a content format stays hot, usually a few months. From the moment you spot a format catching fire to your launch, ideally no more than two or three weeks; if a language already has two or three similar apps with a hundred-plus ratings each, move on to the next format.

What would mean you read it wrong:

• Toxik's daily new ratings in France fall to single digits by late November and it drops out of the Lifestyle Top Grossing chart, which shows that even the people who do distribution only ride one wave, so budget for "one or two months at a time";
• The Cringe or Toxic AI rolls its alt-account matrix out into the language you picked, and the gap is gone;
• You post on one format for three weeks with not a single post past 10,000 views: you picked the wrong format, so switch formats, don't change the product.

---

Data and sources: App Store French storefront listing and in-app purchase list, French-store and French-speaking-Africa chart positions and rating counts (10-28 to 11-06); App Store search results in the German, Spanish, Italian, and Brazilian stores (11-06); Paul Soubigou's LinkedIn post (2026-07-08) and personal site; videos and captions from TikTok accounts @shana_privateaccount and @douae_privateacco (see 11-06). Where it says "inferred," it's an inference.