Industry
Part of Fraud and brand safety: set a suitability standard early
Fraud and brand safety examples: why it is theft, not vanity
Fraud and brand safety examples: the shapes purchased engagement takes, how each one looks from the buyer's side, and why it is theft rather than vanity.
Fraud and brand safety examples start with a simple transaction: someone sells attention that was manufactured. When the seller has inflated the audience, the campaign buyer pays for something that does not exist. That is fraud against the buyer, not a gray area.
This page sets out the shapes it takes and what each one looks like from the outside. Named cases include Devumi, Newtral, YouTube, Amazon and JPMorgan Chase. What to do operationally is in fraud and brand safety.
What to take away
- The thing being sold is attention. Where the attention is manufactured, the sale is of a fiction, and the buyer is the victim.
- Almost every form leaves the same tracea number that does not fit the numbers around it.
- Structure prevents more of this than detection does, because detection is always a step behind.
The shapes fraud takes, with named cases
Purchased followers. The oldest and easiest to spot. It leaves an audience that does not behave: high counts, low interaction, and comments that do not respond to anything specific in the post.
The FTC's 2019 case against Devumi, announced in a 2019 press release, describes this market. Devumi sold fake followers, likes and views; the FTC treated the trade as a deceptive practice.
Purchased engagement. Likes, comments and saves bought in blocks. More expensive to buy and easier to hide than followers, and it corrupts precisely the metric buyers moved to when follower counts stopped being believed. Facebook sued Newtral Technology in 2019, alleging it sold fake likes and followers on Instagram.
Engagement pods and reciprocal groups. Real people agreeing to interact with each other's posts on schedule. Not automated and not fake accounts, and still not an audience that cares about the content. The interaction arrives too fast, from the same set of accounts, on every post.
Networks of coordinated accounts. Accounts run to look independent while serving one operator. The term for the individual case is a sock puppet account; at scale the same behavior is what people mean by a click farm. The 2019 Facebook complaint against Newtral alleged a network of fake accounts used to sell engagement.
Coordinated messaging presented as spontaneous. Comment sections and reviews organized to look like independent enthusiasm. The general name is astroturfing, and it is a disclosure problem as much as an audience problem, because the connection between the poster and the seller is exactly what is being hidden.
Amazon has sued operators of fake review groups. The FTC announced a final rule on fake reviews in 2024.
Traffic and click inflation. Where a campaign is paid on clicks or traffic, the incentive moves there, and the mechanics are the same ones described under ad fraud. The 2016 Methbot operation, reported by White Ops, generated fake video ad views at scale.
Reported results that were never earned. A screenshot is an image. Where the evidence for a campaign's performance is a picture rather than a live view of the account, it can be anything. Advertisers sued Facebook in 2018 over inflated video metrics.
What each one looks like from the buyer's side
| What you see | What it usually means | What to ask for |
|---|---|---|
| High followers, low interaction | Purchased audience, possibly historic | Native analytics, viewed live |
| Interaction concentrated in the first minutes, then nothing | Pods or purchased blocks | The timing curve on recent posts |
| Comments that would fit any post | Purchased comments | Read fifty of them |
| Audience concentrated in markets the content does not serve | Bought where it was cheapest | Audience location in the account's own view |
| Step changes in follower growth | Bought in batches | The growth chart, not the total |
| Results supplied only as screenshots | Nothing verifiable | A live screen share |
| Sudden strong results after weak ones | Possible purchase to hit a target | Compare against the account's own history |
None of these is proof on its own. Together they are usually decisive, and the reason they work is that manufactured numbers are internally inconsistent: whoever bought one metric rarely bought the others in proportion.
Fraud signals and what to ask
What you see
- High followers, low interaction
- Native analytics, viewed live
- Interaction spikes then stops
- Timing curve on recent posts
- Comments fit any post
- Read fifty of them
- Audience in unserved markets
- Audience location in own view
- Step changes in follower growth
- Growth chart, not the total
- Results only as screenshots
- A live screen share
What to ask for
- High followers, low interaction
- Interaction spikes then stops
- Comments fit any post
- Audience in unserved markets
- Step changes in follower growth
- Results only as screenshots
Why the framing matters
Two arguments get made in defense of this and both are worth answering directly.
The first is that everyone does it and it is just how the market works. That is a claim about frequency, and it does not change what is happening: someone is being paid for reach that was not delivered.
The second is that the numbers are a formality and the work is what matters. If that were true, the numbers would not be worth buying. They are bought because they are what the price is based on.
Another version makes the brand the beneficiary, not the victim. A campaign reported internally on inflated numbers makes its leader look effective. That gives buyers a reason to want real verification, not convenient verification.
Measurement and screening belong to different people where possible. The measurement side is in measurement and ROI.
Brand safety examples: real audience, wrong context
Brand safety failures usually involve a real audience in a context the brand did not want, which is a judgment question rather than a deception question.
In 2017 The Times of London found ads on YouTube next to extremist content. AT&T, Verizon, Johnson & Johnson, L'Oreal and the UK government pulled spending. JPMorgan Chase cut placements from 400,000 sites to 5,000 and saw no change in performance. Procter & Gamble cut $200 million in digital ad spend and reported no reach loss.
It needs a written standard set in advance, not a detection tool. Conflating the two produces a screening process bad at both. Where the standard should be written down is in strategy and objectives, and the terms to act on it belong in contracts and disclosure.
Bottom line
Purchased followers, purchased engagement, pods, coordinated account networks and organized comment campaigns are one transaction. Inflated click reporting is the same: someone is paid for attention that does not exist.
Detect it by looking for numbers that do not fit each other, and insist on native analytics viewed live, not screenshots. Structure deals so payment does not depend on numbers a partner can buy.
Common questions
Is a creator always responsible for a fake audience?
Not always. Accounts attract automated followers without buying them, and old purchases from years ago leave traces. What matters is whether they can show you the account's own analytics and talk sensibly about the anomalies.
How much inauthentic audience is normal?
Some is unavoidable. The question is whether the pattern looks like accumulation or like purchase, and a live view of growth and interaction usually distinguishes them.
Are engagement pods really a problem if the people are real?
They are real people producing interaction that has nothing to do with interest in the content. The buyer is still paying for a signal that means nothing.
Should a brand publicly accuse a partner?
No. Decline the deal, keep your evidence, and take advice before saying anything about anyone. A wrong accusation is a serious matter and a right one is rarely yours to publish.
Does a verification score settle this?
It narrows the field. It does not replace looking at the account with the person who owns it.
