Fraud and brand safety: set a suitability standard early

Fraud and brand safety get bundled together in vendor decks and on agency slides, and they describe opposite problems.

Fraud means the audience is not there. You paid for reach that does not exist, or for a response manufactured by something other than interest. The damage is financial and it is measurable in principle.

Brand safety means the audience is real and the context is wrong. The people are genuine, the views are genuine, and your product is now adjacent to something you would not have chosen. The damage is reputational and it is much harder to quantify.

They need different checks, different contract terms, and different responses when they go wrong. Treating them as one category is why brands buy a "verification" tool that solves one of them and assume they have covered both.

What to take away

Fraud: what is actually being faked

Purchased followers. The oldest and least interesting version. An inflated follower count on its own costs you only if you priced on followers, which is a reason not to. It is worth being plain that this is a market with an enforcement record rather than only an industry grumble: in FTC v. Devumi, LLC, the FTC alleged that Devumi sold fake social media indicators, including followers. The FTC described the matter in a 2019 press release about fake indicators of social media influence. Someone buying the campaign is paying for reach that was never delivered.

Purchased or manufactured engagement. Bought likes and comments, or reciprocal pods where a group of accounts agree to engage with each other's posts on a schedule. Pods are harder to spot than bought engagement because the accounts are real people.

Inflated view counts. Bot traffic, or view farms. Varies a great deal by platform and by how views are defined.

Fabricated analytics. A screenshot of a dashboard is an image. Numbers in a media kit are typed. Neither is evidence.

Affiliate and click fraud. Where the payment is performance-based, the incentive shifts. Self-clicking, cookie stuffing, code leakage onto coupon aggregators, and traffic bought elsewhere and passed through as organic referral all inflate what you pay for.

Impersonation and misrepresentation. A lookalike account, or someone claiming to represent a creator they do not. This one costs you the fee and produces no content at all.

Recycled or unlicensed content. Content presented as original that is not, which is both a fraud problem and a rights problem.

How to check, in rough order of usefulness

Ask for platform-native analytics, viewed live. Not a screenshot, not a PDF. A screen share where the creator navigates their own analytics, or an export generated through the platform's own tooling. Meta Business Suite, TikTok Studio, and YouTube Studio are examples. This single step removes most of the crude fabrication, and creators doing real business will not object to it.

Look at the shape of the audience, not the size. Follower geography that has no relationship to the creator's language or subject matter is worth a question. So is an age distribution that does not match the content.

Look at the shape of engagement over time. Genuine performance is volatile: some posts do much better than others. A flat line of near-identical engagement across every post is a pattern that is worth asking about, because organic distribution is not that consistent.

Read the comments. Generic praise, unrelated emoji strings, and the same handful of accounts appearing under every post are the visible signature of pods and bought engagement. This costs five minutes and catches a lot.

Check follower growth history. Step changes with no corresponding piece of breakout content deserve an explanation. There often is a benign one (a feature, a viral moment on another platform, a press mention), and asking gets you that explanation.

Compare against your own delivered results. After a campaign, compare actual view and click delivery against what the audience data implied. Repeated large gaps in one direction are informative even when no single campaign proves anything.

Use verification tools with their limits in mind. Third-party audience-quality tools such as HypeAuditor and CreatorIQ are useful as a first filter. They also disagree with each other, they infer rather than observe, and a low score is a prompt to ask a question rather than a verdict. For broader digital-ad terminology, the Media Rating Council's Invalid Traffic Detection and Filtration Standards can help frame questions, but they are not a creator-specific fraud test. Where a tool's output would cost a creator a deal, tell them what it said and let them respond.

Throughout, distinguish deliberate deception from an artifact. A creator whose account was bot-followed during a spam wave did not commit fraud. So ask before concluding, and note that a genuine explanation usually arrives quickly and with evidence.

Contract and payment structures that reduce exposure

Structure removes more risk than detection does.

Pay against outcomes you can observe, not numbers you are told. Tie any performance element to your own tracked data, not creator-reported figures.

Where a fee is fixed, often appropriate, accept you are buying an attempt; manage risk by choosing partners, not auditing after the fact. A payment structure is only as good as the campaign brief that states which numbers count as delivery.

Require that the creator's analytics be shared post-publication, in a stated form, within a stated period. Make it a deliverable. For sponsored endorsements, the FTC's Endorsement Guides (16 CFR Part 255) explain how the FTC Act applies; material connections should be disclosed clearly and conspicuously.

Warrant against artificial inflation. A clause could say: “Creator represents that they have not knowingly purchased followers or engagement for the account and will not do so during the campaign. If the brand reasonably substantiates artificial activity affecting campaign deliverables, the parties will review the evidence and apply the agreement's remedies.” Adapt the wording to the deal and have counsel review it. The clause gives you a remedy if the representation turns out to be false.

For affiliate arrangements, define what counts as a valid conversion, set a hold period before commission is paid, and state that traffic from coupon aggregation sites, or self-referred purchases, does not qualify. Watch for codes leaking onto deal sites, which is common and often not the creator's doing.

Verify the payee independently of the contact route. Payment redirection is a well-worn fraud against marketing departments, and a request to change bank details mid-campaign should be confirmed through a channel you established earlier.

Brand safety: real audience, wrong context

The risks here divide by when they occur.

Past. Everything the creator has published or said before you found them. Check it, and check it before you make an offer, because a withdrawn offer travels further than a declined pitch. How far back to look is a judgment call, and so is what counts as disqualifying, but making that judgment explicitly, in advance, is better than making it under pressure after a screenshot circulates.

Present. The immediate context of the sponsored content. What else is on the channel around it. What is visible in the frame. What the comment section does within the first hours. Whether the content sits next to something in a feed you cannot control.

Future. What the creator does after your content is live, and while your association is visible. This is the risk that cannot be checked in advance, only planned for.

Brands forget the reverse: creators take reputational risk from brands. A failed product, a company in the news for the wrong reason, or a campaign audiences reject also lands on the creator.

Contracts that give the brand a reputational exit and the creator none are one-sided. Experienced representatives increasingly push back.

Deciding what is out of bounds, before you need to

Write the standard down before a specific case forces the question. A brand suitability standard is the brand's written criteria for which creator conduct and content contexts are acceptable, unacceptable, or require review; it should state how those criteria apply to past, present, and future content and who makes the decision. A list drawn up in the abstract is a policy. One drawn up during an incident is a rationalization, and it will read as one. The standard also has to serve something, which is the aim set in strategy and objectives rather than whatever the newest complaint is about.

The IAB Tech Lab's Content Taxonomy offers shared categories that can help brands describe content, but it is not a brand-specific suitability policy and was not written specifically for creator partnerships. The general shape of the term is set out under brand safety; borrow the structure, but do not assume the categories map cleanly.

Be honest about the fact that these are editorial judgments, not compliance findings. A brand that pretends its brand-safety standard is objective ends up applying it inconsistently and cannot explain either the inclusions or the exclusions.

Keyword blocklists deserve a specific warning. Applied to influencer work, crude blocklists reliably defund exactly the creators covering the subjects that most need funding: health, news, and communities discussed using the same words that appear on the list. If you use one, review what it is actually excluding.

When something goes wrong

Speed matters less than most people think, and accuracy matters more. A response issued in the first hour on incomplete facts is frequently the thing that has to be walked back.

Establish the facts first, from primary sources, and write down what is confirmed and what is reported. Decide whether the issue concerns the sponsored content, the creator's other conduct, or something the audience has attributed to you incorrectly: those need different responses.

Know in advance what your options are and what they cost. Pausing amplification is instant and cheap. Removing content depends entirely on what the contract says, and this is exactly the clause people discover they do not have. Ending a relationship publicly is a decision with its own consequences, including for creators who turn out to have been wrongly accused.

Say who speaks. One named person, one channel, and a rule that nobody else comments. Multiple partial statements from different people is how a manageable problem becomes a story about the response.

Afterward, write down what the early signals were and whether any of your checks could have caught them. Most incidents are not predictable. Some are, and those are the ones worth changing the process for.

What this connects to

Fraud checks belong before the offer, alongside the identity verification described in creator outreach. Brand-safety standards belong in the contract as defined conduct terms rather than as open-ended discretion. Detection of delivery shortfalls happens through the same data as measurement, which is why a measurement plan that only reports favorable numbers also hides fraud.

Bottom line

Keep the two problems separate. Fraud is checked with live platform data, engagement patterns, comment sections, and payment structures that do not reward inflated numbers. Brand safety is handled with a written standard set before you need it, a contract that defines conduct terms in both directions, and a rehearsed response that prioritizes getting the facts right over responding fast.

Common questions

Does a verification tool prove an audience is genuine?

No. These tools infer from observable signals and they disagree with each other. Use one to prioritize where to look, then ask the creator directly about anything it flags.

Is buying followers common enough to worry about?

Common enough that a check is worth the time, and cheap enough to check that there is no reason not to. The more important point is to avoid pricing on follower count in the first place, which removes the incentive.

How far back should we review a creator's content?

There is no correct answer, only a decision you should make deliberately and apply consistently. Whatever you choose, write it down, and expect to defend it either way.

Can we require a creator to delete old content?

You can negotiate for terms about conduct during the campaign. Asking someone to remove unrelated past work is a significant demand and it is often refused, reasonably.

Who is responsible if delivery falls short of what the audience data implied?

Depends on what was promised. If you contracted for a fixed fee against deliverables, you got the deliverables. If you want delivery guarantees, negotiate them explicitly and expect a different price.