Anti-bot policy in plain language

Most agencies in our niche have a page like this. Most of those pages are theatre. They say "we don't use bots" in 40 vague words and then sell you a Telegram-based "boost" the next day. We wrote this one to be useful instead — a list of five concrete things we refuse to do, why, and how each refusal maps to a written rule from a platform or regulator. If you ever catch us doing any of these, you have grounds for a full refund under our refund policy.

We are a small team in Minsk. We work mostly with Polish and EU clients who entered a brand contest and want professional traffic to their entry page. We run paid ads. That's almost the whole product. Read our services page if you want the full scope.

The short version

We refuse to sell, broker, integrate, or wink at these five things:

1. Automated voting scripts (bots that click the vote button)
2. Emulator farms (phone farms that simulate human devices)
3. Purchased SIM or email batches (for mass account creation)
4. Account-rental services (real accounts you "borrow" to vote)
5. Sponsored fake-engagement networks (paid pods, comment rings, click exchanges)

Everything below is the long version: what each one actually is, why people sell it, and the specific written rule it violates.

Why this matters more than it used to

Google's Misrepresentation policy was last revised on 24 May 2024 and it now treats "coordinated inauthentic behavior" and "concealed identity" as a single category of strike-worthy violation. Translation: if we run a Google Ads campaign that points to a landing page whose vote count is being inflated by scripts, Google can suspend the ad account that paid for the traffic. Not the bot operator's account. Ours. Sometimes yours, if the landing page or destination URL is on a domain you own.

Meta's Advertising Standards have an equivalent rule under "Unacceptable Business Practices," and the EU's Digital Services Act (Regulation 2022/2065, applicable to all VLOPs since 25 August 2023, see eur-lex.europa.eu) makes "manipulation of recipient behavior" through deceptive interface or automated means a named risk category.

So when we say no to the five things below, it isn't moral posturing. It's that we want to still have ad accounts in twelve months. Our business model only works if our Google Ads MCC stays alive. Bot-adjacent work is, in a literal sense, suicide for an agency. We don't understand the agencies that still do it.

1. Automated voting scripts

This is the obvious one. A script — Python, Selenium, a Chrome extension, a "vote tool" sold on a sketchy forum — opens the contest page and clicks the vote button. Usually it rotates IP addresses through a residential proxy pool to look like different users.

People ask us for this maybe twice a month. Last month a participant in a Żabka-adjacent product contest (we won't name it more precisely) asked if we could "just press the button 4,000 times" because she'd been told that's what the leader was doing. We said no, explained why, and she went with a competitor who said yes. Two weeks later her entry was disqualified by the brand running the contest, who had logging on their voting endpoint and noticed 3,847 votes from a /16 subnet in Bulgaria over six hours. We didn't get the work. We also didn't get the chargeback or the legal letter.

Mapped to Google's Misrepresentation policy: this falls under "coordinated activity that misleads users... about the source of content the user is interacting with." The vote count is content. A bot-generated vote misrepresents its source. Done.

2. Emulator farms

Slightly more sophisticated. Instead of one script clicking from one server, an emulator farm runs hundreds of virtualized Android devices, each with a unique device fingerprint, each "voting" once. The output looks more human because the device IDs differ.

You can rent these. There's a market for it, mostly out of Southeast Asia. The going rate when we last checked (March 2026) was roughly __DRAFTS_PLACEHOLDER__.04 per "vote" for low-protection contests. We will not broker this, integrate with operators who offer it, or "introduce you to a partner who can help with the manual side." That last phrase is the dodge most agencies use. We refuse the dodge.

Mapped to Google's Misrepresentation policy: same clause as above, plus the separate "Unreliable claims" section, which covers "inflating engagement metrics through inauthentic means." Mapped to Meta's Inauthentic Behavior standard (last substantive update referenced in their changelog 14 February 2025): emulator farms are explicitly listed.

3. Purchased SIM or email batches

This one is upstream of the actual voting. To run an emulator farm or operate fake accounts, you need phone numbers and email addresses that pass the contest platform's verification step. There's a wholesale market for "warmed-up" SIM cards and pre-aged Gmail/Outlook accounts. A batch of 500 Polish-numbered SIMs runs about €1,200 last we checked.

We don't buy these. We don't accept them from clients. If you've already bought a batch and want us to "use" them in a campaign in any way, we won't take the project. This is the boundary we hold most firmly because it's the one with criminal exposure under Polish law — specifically the obligation under the Act on Combating Money Laundering of 1 March 2018 to verify customer identity for prepaid telecom services, which bulk-SIM resellers routinely violate.

This is one of the things we will not do. Even if you signed the contract. Even if the deposit cleared.

4. Account-rental services

The cleanest-looking option from the buyer's side. You don't create fake accounts; you rent real ones. Someone offers their Facebook or Google account for a few złoty and you vote from it. The accounts are real, the people exist, and the vote technically passes the platform's checks.

It still violates the contest's terms of service in every case we've reviewed. The standard clause — and we've read maybe 47 contest regulamins in the last year — says something like "one vote per natural person, voting on behalf of others is prohibited." Renting an account is voting on behalf of someone else. Doesn't matter that they consented.

It also violates Meta's Terms of Service section 3.1 ("you will not buy, sell, or transfer any features of your account") and Google's account terms equivalent. We've seen the "account rental" pivot rebranded as "micro-influencer voting campaigns" — same thing, different wrapper. We don't do either.

If a vote requires a real human making a real choice, and you're paying someone to skip the "choice" part, you're buying a vote. We don't broker vote-buying regardless of how it's described.

5. Sponsored fake-engagement networks

The most subtle and the fastest-growing category. These are private Telegram groups, Discord servers, or invite-only platforms where members trade engagement: I vote for you, you vote for me, the broker takes a cut. The technical term is "engagement pod" but on the vote-promotion side they're usually called "voting circles" or "wzajemne głosowanie."

The votes are from real humans. The humans are not interested in the contest. They're paid or coerced (via reciprocity rules) to vote. The result looks organic until you look at the IP geography, time clustering, or referrer data — which is exactly what contest organizers increasingly do.

Vyper, one of the larger contest-platform vendors used by EU brands, shut down its hosted product on 12 September 2024 and the replacement vendors (KingSumo, Gleam, the SaaS layer behind a lot of Polish brand contests) all ship with referrer-based bot detection now. Pod votes get flagged. Entries get disqualified silently — meaning you won't even know until the winner is announced and it isn't you.

Mapped to Google Ads Misrepresentation policy: "Coordinated deceptive practices" subsection, which was the specific addition in the 24 May 2024 revision. Mapped to the EU Unfair Commercial Practices Directive (2005/29/EC, Annex I clause 23a as amended by Directive 2019/2161): "stating that reviews of a product are submitted by consumers who have actually used or purchased the product without taking reasonable and proportionate steps to check that they originate from such consumers." A vote is a form of review. Coordinated fake votes are coordinated fake reviews.

What we do instead

We buy ads. That's the boring, legal, sustainable answer. Google Search ads targeting people who searched for the brand running the contest. Meta ads to lookalike audiences of the brand's existing customers. Sometimes TikTok if the demographic fits. We write the copy, build the landing page, and report what happened.

If a real person sees your ad and chooses to vote because the entry actually appealed to them, that vote is legitimate, durable, and (frankly) the only kind worth having. The contrarian opinion most agencies in this space don't want you to hear: most contests are won at 600–2,000 real votes, not at the inflated 40,000-vote totals you see at the top of the leaderboard. Those leaderboard whales almost always get disqualified in the audit phase. We've seen it five times this year alone.

The process page describes how we actually run a campaign. The pricing page shows what it costs. Neither involves any of the five things above.

What we will not do, restated

We will not run scripts, rent emulators, buy SIMs, broker accounts, or join pods. We will not introduce you to a third party who does any of these. We will not "look the other way" on traffic you generated through these channels and then layered our ads on top of. If you've already used any of the five before contacting us, tell us honestly during the intake call — sometimes we can still help by running clean ads in parallel, sometimes the contest is already compromised and we'll decline the work. Either way you get honesty about it.

If you suspect a competitor is doing this

Document it. Screenshots of the leaderboard with timestamps, network logs if you have them, the contest organizer's contact email. Most contest regulamins include a reporting mechanism and a sanity-check phase before prize distribution. Use them. The 24 May 2024 Google policy update also enabled third-party reporting of misrepresentation through the Ad Center, though response times are uneven (we've seen anywhere from 4 days to 6 weeks).

We won't do the reporting for you — it isn't our place and it would create conflicts in cases where the reported party is also a prospective client of ours. But we're happy to help you understand what you're looking at on a call. See the contact page or the FAQ for the questions we hear most often.

One last thing

We've been doing this since 2018. We've turned down work worth, conservatively, €38,400 over that period because the request crossed one of these five lines. We don't regret any of it. The accounts we still have, the clients who come back, and the fact that we can write a page like this without our lawyer rewriting it — all of that is worth more than the deals we said no to.

If this page made you reconsider what you're looking for, good. If it made you want to work with us, even better. Read the about page for who we are, and the privacy page for how we handle the data you share with us during a project.