Thinking · Newsletter
Where the line is on brand safety, and who actually draws it
Brand safety tools control where an ad runs, not what it says. Agree what the brand will never claim or show, and who approves exceptions, before concepting.
TL;DR. The brand safety tools on the ad networks control where an ad appears, not what it says or shows. That second line is drawn by people, and it is cheapest to draw before concepting: what the brand will never claim, what it will never depict, and who by name can approve an exception. Agree those three things in writing and most brand-versus-performance fights disappear before they start.
The tension is real. Performance wants the ad that works. Brand wants the ad that is defensible. Both are right, and the argument usually happens too late.
Two problems that share one name
"Brand safety" covers two separate jobs, and teams often assume the tools handle both.
The first is placement: what content your ad runs next to. The networks have built real controls for this. Meta's developer documentation describes publisher block lists, which "stop your ads from appearing with publishers you don't consider suitable for your brand or campaign", content block lists, which "allow advertisers to block their ads from appearing adjacent to content they deem unsuitable", and delivery reports that "provide transparency into where ads appeared."
The second is the ad itself: what it claims, what it depicts, and what it promises the product will be. No setting in Ads Manager reviews that. That line is drawn by the people who own the brand; the only question is when.
The expensive version
Creative is made to the performance brief, shown to brand, and cut apart. What ships is a compromise nobody chose. The performance team concludes brand killed it; the brand team concludes performance ignored them.
The cheaper version
Establish the line before concepting. Not a list of banned words: a description of what the brand cannot be seen to say. Those are different. A banned-word list is a filter applied after the work is done. A description of the line is a constraint you can design around, and good creative is often made inside tight constraints. Our casino teardown makes the same point about a genre where the compliance line is the brief.
What the line has to cover
Three outside rules make this more than taste.
Claims. A claim made by an actor or creator in a UGC script is still the brand's claim, so it needs the same proof as any other line of ad copy.
Depiction. In 2020 the UK Advertising Standards Authority ruled that two puzzle game ads built around a pin-pulling minigame were misleading, even though each opened with the disclaimer "Not all images represent actual gameplay." The ASA found players would have to play a significant amount of content in a different style to reach the gameplay shown, and told the publisher to make sure future ads "presented gameplay which was representative of the games they advertised." The ruling applies in the UK, but it shows a disclaimer is not a substitute for a decision.
Design. Platform policies also govern how an ad looks. Google's misleading ad design policy lists, among other things, "Ads that resemble system or site warnings or error messages" and "Ads with features that don't work, such as close buttons." Cheaper to rule out in a brief than in a policy review.
| What to agree before concepting | Example of a line | Who should own it |
|---|---|---|
| What the brand will never claim | No health, earnings or "best" claims without proof on file | Brand or legal, named |
| What it will never depict | No gameplay the player will not meet early; no competitor products | Brand lead, named |
| How the ad may look | No fake system dialogs or dead buttons | Media lead, checked against each network's policy |
| Who approves an exception | One person, with a turnaround time | Agreed by both sides |
What the public game ads show
The depiction question is not hypothetical in games. In the 12,150 public Meta Ad Library ads from 23 game brands that Flixr Studio reads (ads launched between October 2024 and November 2026, most of them since September 2026), 30.4% of the 12,122 ads read contain stretches of fake gameplay: footage that is not the game's core loop as a new player meets it, such as invented puzzles or minigames. 21 of the 23 brands run at least some. The spread is wide: two brands run none, nine keep it under 10% of their ads, and four use it in half or more. Among the 860 ads that have run 90 days or more, 25.3% contain fake gameplay.
We report this neutrally. The readings are made by a model (it agrees with a second model's blind labels on fake versus real gameplay about 95% of the time), long-running is a signal and never proof of performance, and we have no spend data for any of these brands. What the numbers do show is that this is a deliberate line, and each brand has drawn it in a different place. The point of this piece is that it should be drawn on purpose, in a brief, by someone with the authority to draw it.
What we ask for
Three things, in writing, before the first concept:
- What the brand will never claim. Specific, with the proof that exists for the claims it will make.
- What it will never depict. Including the edge cases: competitor products, gameplay that is not representative, situations the brand does not want to be associated with.
- Who has the authority to say yes to an exception, by name.
Why the third one matters most
Most brand-safety delays are not disagreements; they are escalations. Nobody in the room can approve the one line near the edge, so it goes up a chain and waits. Knowing in advance who can approve an edge case turns a two-week loop into an afternoon.
Where Flixr stands
We hold to the original argument: the line should be drawn before production, and it should describe what the brand cannot be seen to say rather than list words to avoid. The evidence above adds two things.
First, the networks' brand safety tools are necessary and they are not the answer to this problem. They protect adjacency. The ad's own content is the brand's responsibility.
Second, fake gameplay is a choice, not a default and not automatically wrong. Most of the game brands Studio reads use some; a few use a lot; two use none. Our view is that it belongs in the "never depict" conversation at the brief, decided with the ASA's representativeness test in mind, rather than discovered by brand in the edit. That is a view from reasoning, not from a test we can show.
How we work makes this practical: a creative lead sits inside the client's account, so the three answers are written into the first brief and logged once, and every weekly loop (brief, make, launch, read, log) starts from them instead of rediscovering them. If your account has ever had an ad pulled for policy reasons, our note on AppLovin suspensions covers the platform side.
Questions people ask
Do Meta's brand safety controls check what my ad says? No. Inventory filters, block lists and delivery reports govern where an ad appears and what it sits next to. What the ad claims and shows is reviewed by the brand and by each network's ad policies, not by those controls.
Is a "not actual gameplay" disclaimer enough? Not in the UK case we know of. The ASA ruled two ads misleading despite that disclaimer, because the gameplay shown was not representative of the games.
How common is fake gameplay in game ads? In the public ads Flixr Studio reads, about 30% of read ads contain some, across 21 of 23 brands, with a very wide spread between brands.
Sources: Meta for Developers, Brand Safety and Suitability (developers.facebook.com/docs/marketing-api/brand-safety-and-suitability); ASA ruling on PLR Worldwide Sales Ltd, 30 September 2020 (asa.org.uk); Google Ads policy, Misleading ad design (support.google.com/adspolicy/answer/15937463); Flixr Studio, public Meta Ad Library game ads, queried 7 November 2026.